Semantic prosody in a cross-linguistic perspective [1]

Signe Oksefjell Ebeling
University of Oslo

Abstract

1. Introduction and aims

The semantic prosody, i.e. the evaluative meaning or communicative purpose, of cause has previously been explored both on the basis of monolingual data (Stubbs 1995) and on the basis of comparable cross-linguistic data (e.g. Berber Sardinha 2000). [3] The present paper picks up the thread of both of the aforementioned studies and aims to further explore the semantic prosody of cause in a cross-linguistic perspective, on the basis of translational data, viz. the English-Norwegian Parallel Corpus (ENPC).

The study follows Stubbs in taking both the noun and the verb cause as its point of departure. On the basis of the ENPC, an analysis of the semantic prosody and lexicogrammatical patterns of cause will be carried out; its Norwegian translation will be recorded in each case. The top three translational correspondences will then serve as the starting point for a similar analysis in order to establish whether the most commonly used correspondences share the negative semantic prosody of cause. [4]

Berber Sardinha studied the semantic prosody of the English and Portuguese cognates cause and causar on the basis of comparable monolingual corpora for English and Portuguese, and concluded that they “share a negative semantic prosody” (2000: 97). For the language pair English-Norwegian there are no obvious cognates such as cause and causar, and a pilot study in the ENPC confirms this; in 83 verb translations of the verb cause 19 different verbs have been used in Norwegian. Another observation worth exploring further is the fact that the lemma cause (noun and verb uses) occurs twice as frequently in the original texts as it does in the translated texts. Thus, it seems fair to say that there is a lexical gap between English and Norwegian, triggering a cross-linguistic investigation of a negatively loaded lemma of a different kind than the one carried out by Berber Sardinha.

The main aim of the present investigation is to examine how stable semantic prosodies are across languages, also where non-cognates are involved. This will be done in a case study investigating English cause and its Norwegian correspondences in a bidirectional translation corpus. Thus, another important aspect of this paper is to explore the method of using a bidirectional corpus for the purpose of investigating prosodies cross-linguistically, more specifically the semantic prosody of the Norwegian correspondences of cause. If different from the negative prosody of English cause, what are the cross-linguistic implications for contrastive studies and the translation of items or units with particular prosodies? As pointed out by Berber Sardinha, this is an important question because:

[…] the shift in semantic prosody quality may be unintentional. When this happens in translation, it gives rise to a different connotation to that intended by the writer or speaker of the source language text. This should be avoided in translation because it may create misunderstandings with respect to tone or content of the original message (Berber Sardinha 2000: 96).

Morley and Partington (2009: 140) also stress the fact that an awareness of prosodies is essential for translators, so that they are better able to distinguish between potential translation equivalents.

2. Background

2.1 Semantic prosody

‘Semantic prosody’ has been studied by corpus linguists for almost two decades. Still, Hunston (2007: 249) refers to ‘semantic prosody’ as a “contentious term”, and disagreement on what it in fact refers to has led to recent debates on the topic (e.g. Hunston 2007 and Stewart 2010). Points of disagreement include the questions of (a) whether the prosody resides in the lexical item or in the discourse, (b) whether semantic prosody is connotational in nature or not, and (c) whether, or how, semantic prosody is different from semantic preference. See e.g. Morley and Partington (2009), Hunston (2007), Partington (2004) for a discussion of one or more of these issues. Highly critical views of semantic prosody have also been voiced (Whitsitt 2005, in particular).

The concept of semantic prosody is very much anchored in the Sinclairian tradition, and the present paper draws on work within this tradition, in particular Sinclair (e.g. 1987, 1991, 1996) and Stubbs (e.g. 1995, 2009). In what follows, I will outline the main features of semantic prosody on which this study relies. (For other and more comprehensive overviews of the study of semantic prosody, see e.g. Xiao & McEnery (2006) and Stewart (2009).)

The term ‘semantic prosody’ was first used by Louw (1993), who defined it in the following way: “a consistent aura of meaning with which a form is imbued by its collocates” (Louw 1993: 157). Louw attributed the concept to Sinclair who developed it in later work, notably so in his model of extended lexical units, where semantic prosody is the last of four levels: (1) collocation, (2) colligation, (3) semantic preference, and (4) semantic prosody (Sinclair 1996, 1998). This model assumes that “meaning cannot be said to belong to a single word, but to the phraseology as a whole” (Hunston 2002: 141). It is based on quantitative analysis, where systematic collocational patterns emerge “to such a degree that some words prefer, or even require, a semantic profile of the words with which they combine” (Dam-Jensen & Zethsen 2008: 206). This is what is termed ‘semantic preference’ within Sinclair’s model.

The fourth component of an extended lexical unit is known as ‘semantic prosody’, or ‘discourse prosody’, to borrow a term from Stubbs (2007); it refers to relations that involve evaluative meaning (Dam-Jensen & Zethsen 2008, Morley & Partington 2009, Stubbs 2009). It is also important to stress that semantic prosody refers “not to simple co-occurrence but to the consistent discourse function of the unit formed by a series of co-occurrences: the ‘unit of meaning’” (Hunston 2007: 257), as in the case of units of meaning of which the verb budge is the ‘core’:

The core gives us the starting point, in the case of budge one that anticipates the prosody fairly clearly; the optional patterns of collocation, colligation and semantic preference bring out relevant aspects of the meaning, and the prosody can then be searched for in the close environment. (Sinclair 1998: 22)

With reference to the same example with budge, Stubbs (2009: 124) explains that the verb budge often occurs in the vicinity of a negative element, and has a semantic preference for e.g. doors, lids of jars, obstinate people, relating to the topic of the text. A large set of examples (typically in the form of concordance lines) reveals that, in extended lexical units with budge, “the speaker has tried repeatedly to do something, has failed, and is now annoyed. This overall evaluative ‘semantic prosody’ is the communicative function of the whole unit” (Stubbs 2009: 124).

Thus, “the semantic prosody of an item is the reason why it is chosen, over and above the semantic preferences that also characterise it” (Sinclair 1998: 20). In other words, the semantic prosody has to do with the pragmatic function of an extended lexical item; “[w]ithout it, the string of words just ‘means’ – it is not put to use in a viable communication” (Sinclair 1996: 88).

In some of the earlier papers on semantic prosody, including Sinclair (1987, 1991) and Stubbs (1995), semantic prosody was typically “regarded as a rather vague indicator of the attitudinal connotation of a word: something good/desirable or bad/undesirable” (Stubbs 2009: 126). This has led to a confusion between ‘semantic preference’ and ‘semantic prosody’, since they, in some cases, may seem to refer to the same phenomenon. However, they should be seen as “two distinct yet interdependent collocational meanings” (Xiao & McEnery 2006: 107), and I believe Stubbs manages to distinguish between the two in a clear manner: “it might be helpful to use different terms here, in order to distinguish between semantic relations [i.e. ‘semantic preference’] (to the topic of the surrounding text), and the pragmatic function [i.e. ‘semantic prosody’] (of the whole phrasal unit)” (Stubbs 2009: 126). [5] Similarly, Partington observes that “semantic preference is a ‘narrower’ phenomenon – relating the node item to another item from a particular semantic set – than prosody which can affect wider stretches of text” (Partington 2004: 151).

Nevertheless, the good vs. bad dichotomy need not be abandoned. In their article on frequently asked questions about semantic prosody, Morley and Partington (2009: 141) discuss the relationship between semantic prosody and evaluation. They suggest that “evaluation at its most basic is a two-term system”, viz. good vs. bad, or positive vs. negative. [6] The fact that the goodness or badness may come in different forms can be seen in relation to different communicative or pragmatic purposes, such as danger, difficulty, uncertainty, all of which are subcategories of bad, or negative, prosody. To arrive at such specific prosodies, we need to look at an item’s semantic preference, and Morley and Partington (2009: 142) suggest that “the overall semantic prosody of an item is both shaped by and expressed in its semantic preferences”.

This paper takes a relatively early publication on semantic prosody as its point of departure, viz. Stubbs (1995), and reference is made to words having “distinctive semantic profiles or ‘prosodies’” (Stubbs 1995: 23). These words with their distinctive semantic prosodies represent ‘the core’ of an extended lexical unit, in Sinclair’s terms, and as such function as the starting point of an investigation of semantic prosodies. [7] Thus, in the following when reference is made to the “semantic prosody of cause”, it is the semantic prosody of units including cause, where cause is the core, that is implied. Moreover, in my analysis, communicative purposes such as the ones mentioned above, viz. danger, difficulty, uncertainty, will not be considered. I will mainly focus on positive vs. negative prosody (including neutral), in order to make my study as comparable as possible to that of Stubbs (1995).

2.2 The semantic prosody of cause

In Stubbs’ 1995 article on the semantic prosody of cause “it is shown that the lemma cause occurs in predominantly ‘unpleasant’ collocations” (Stubbs 1995: 166). The present study, like Stubbs’, takes both noun and verb uses of cause into account, as in the following examples.

In his study, Stubbs investigates instances of the lemma cause in several corpora, first by looking at relatively small corpora: the 500,000-word London-Lund Corpus (LLC) and the 1-million-word Lancaster-Oslo/Bergen Corpus (LOB). In LLC and LOB he found cause to occur with negative prosody in around 80% of the cases (neutral: 18%, positive: 2%). Similar figures were found in the more substantial 120-million-word Cobuild corpus. In addition, his results have later been substantiated in studies by Xiao & McEnery (2006) and Dam-Jensen & Zethsen (2008).

The findings with regard to the semantic prosody of cause referred to above are all based on monolingual (English) data. As the present study takes a cross-linguistic approach, a few observations on cross-linguistic semantic prosody will be presented in the next section.

2.3 Cross-linguistic semantic prosody (of cause)

Collocation and semantic prosody have recently attracted much interest from researchers studying English, but little work has been done on languages other than English and still less work has been undertaken contrasting the collocational behaviour and semantic prosody of synonyms in different languages. (Xiao & McEnery 2006: 103)

This observation is to some degree still a valid one; however, more and more studies, mainly on the basis of comparable cross-linguistic data, have emerged over the last ten years, including e.g. Berber Sardinha (1999, 2000) for the language pair English-Portuguese, Tognini-Bonelli (2001, 2002 [8]) and Partington (1998) for the language pair English-Italian, Xiao & McEnery (2006) for the language pair English-Chinese, and Dam-Jensen & Zethsen (2006) for the language pair English-Danish. [9] Most of these have direct relevance for the present investigation since they all draw on Stubbs (1995) and his work on the prosody of cause: Berber-Sardinha (1999, 2000), Xiao & McEnery (2006), and Dam-Jensen & Zethsen (2006). [10]

Even if these studies on three different language pairs differ in certain respects – cognate vs. non-cognate equivalents, related vs. non-related languages, implications for translation vs. language learning – they lend support to Stubbs’ claim that cause has an overwhelmingly negative profile, i.e. cause and some of the equivalents with which it is compared have similar tendencies towards a negative prosody. However, Stewart (2009), with reference to Xiao & McEnery’s and Berber Sardinha’s studies, stresses an important point, namely that “semantic prosodies of near synonyms are unpredictable across the two language pairs, in some cases being quite similar and in others quite different” (Stewart 2009: 32).

In his study of the English-Portuguese cognates cause and causar, Berber Sardinha (2000: 97) “concludes that ‘cause’ and ‘causar’ share a negative semantic prosody”, as exemplified in (3).

Similarly, taking the English-Danish synonymous non-cognates cause and forårsage as their point of departure, Dam-Jensen and Zethsen conclude that “our analysis convincingly shows an extremely negative semantic profile [of forårsage]” (Dam-Jensen & Zethsen 2006: 1620), as in (4).

(4)	forårsage en omfattende miljøkatastrofe (Dam-Jensen & Zethsen 2006: 1618) Lit.: cause a comprehensive environmental disaster

The Danish synonym of forårsage – medføre –, on the other hand, shows that even if it is often used in negative contexts, it is not used in negative contexts to the same degree as forårsage. This is in accordance with what Xiao & McEnery observe for English and Chinese: “near synonyms are normally not interchangeable in either language” (Xiao & McEnery 2006: 125).

Xiao & McEnery (2006) focus on the semantic prosody of near synonyms in both English and Chinese, viz. the cause group (including cause, bring about, lead to, etc. for English). They also look at close translation equivalents and conclude that some equivalents “display very similar collocational behaviour and semantic prosodies in both English and Chinese”, e.g. “cause vs. zhi4shi3 (致使)/zao4cheng2 (造成)” (Xiao & McEnery 2006: 120). Additionally, Xiao & McEnery point out that it is also important to note that “the semantic prosodies observed in general domains may not apply to technical texts” (2006: 125).

On the basis of their cross-linguistic findings, Xiao & McEnery (2006: 125) conclude:

While the corpus-based approach can only reveal, but not explain, such cross-linguistic similarity, at least part of the explanation, in our view, can be found in the common basis of natural language semantics – ‘the conceptual system that emerges from everyday human experience’ (Sweetser 1990: 1).

With this cross-linguistic background, it will be an interesting and challenging task to pick up the thread and apply a slightly different method to investigate the semantic prosody of cause in a cross-linguistic perspective.

As mentioned above, the three studies referred to here differ in their focus on the implications of semantic prosody on translation vs. language learning. While Xiao & McEnery (2006) are mainly concerned with the latter, Berber Sardinha (2000) and Dam-Jensen & Zethsen (2006, 2008) are mainly concerned with the former. Since this study relies on translational data, some of the voiced concerns will be presented in the following section.

2.4 Implications for translation

Several scholars have pointed out the importance of semantic prosody in connection with translation and translation studies, because, as illustrated above, cross-lingustic near synonyms and translation equivalents found in the dictionary may or may not have different prosodies across languages. Partington, for instance, concludes that “the pitfalls for translators unaware of such prosodic differences are evident” (Partington 1998: 78). He uses an example from English vs. Italian to illustrate such a pitfall: English impressive typically has favourable prosody, while the “look-alike Italian word” impressionante is used “as often as not to collocate with neutral and unfavourable items” (Partington 1998: 77). This is supported by Tognini-Bonelli:

The evidence of a divergence in semantic preference and/or prosody will be of great help to the translator, for example, and will allow him/her to avoid those instances of rather infelicitous ‘translationese’. (Tognini-Bonelli 2002: 85)

Tognini-Bonelli exemplifies her claim in a study of the English-Italian translation equivalents in case vs. se per caso, and argues that “at the level of semantic prosody it [the translation equivalent – se per caso] could generate a trap for the unaware translator because the correspondence is similar but not as systematic” (Tognini-Bonelli 2002: 90).

Concerns of this kind have also been expressed by Dam-Jensen & Zethsen (2008), Berber Sardinha (2000), Morley and Partington (2009), and Stewart (2009), and I believe that by carrying out more studies on cross-linguistic semantic prosodies, we may gain insight that will be valuable for translators, translator training, and language learning alike. In the following, we will explore a method that may prove useful in the process of gaining more cross-linguistic insight into this area.

3. Material and method

We knew that our approach should be corpus-based as previous studies have shown that a speaker’s intuition is usually an unreliable guide to patterns of collocation and that intuition is an even poorer guide to semantic prosody. (Xiao & McEnery 2006: 103)

Virtually all studies on semantic prosody have been carried out on the basis of corpora, including the cross-linguistic studies referred to above. One reason for this, in addition to the one referred to in the quote from Xiao and McEnery above, might be that the KWIC format, or the use of concordance lines, in corpus linguistics lends itself particularly well to the study of semantic prosody, since it is only on the basis of recurrent lexical items in context that semantic prosody can be identified.

While previous studies on semantic prosody have mainly relied on monolingual (comparable) corpora and dictionaries, the methodology I will apply in this study makes use of a parallel corpus, i.e. a bidirectional translation corpus.

A bidirectional translation corpus is a comparable and translation corpus in one, where “each type can be used to control and supplement the other” (Johansson 2007: 11). One of the first corpora of this kind was the English-Norwegian Parallel Corpus (ENPC), compiled under the direction of Stig Johansson between 1994-1997. [11] The model for the ENPC, as devised by Johansson, includes English and Norwegian original texts and their translations, containing the same amount and same type of texts (to the extent that this was possible), i.e. the different components are matched for size as well as text-type. [12] The ENPC contains text extracts of 12,000-15,000 words from 30 original fiction texts and 20 original non-fiction texts from both English and Norwegian; in total, including the translations, it consists of approximately 2.6 million running words. (See Johansson et al. 1999 for more details.) The model can be illustrated in the following way:

The bidirectional translation model makes it possible to distinguish between language differences and translation effects; hence we can say that it is a corpus for contrastive analysis and translation studies, packed in one. (Johansson 2007: 12)

Thus, the present study differs from the other cross-linguistic studies mentioned in that it has available not only a comparable corpus, but also a translation corpus, enabling the researcher to study actual correspondences rather than to rely on translational equivalents found in dictionaries. The translation correspondences in a corpus such as the ENPC have the advantage that they are choices made by several translators in similar contexts. [13]

There is some disagreement as to what constitutes the best basis for comparison in contrastive studies; however, such a discussion lies outside the scope of this paper, but see Ebeling (2000) for a discussion, and e.g. Lado (1968 [1957]), Krzeszowski (1990), Chesterman (1998) for other views. In connection with the choice of corpus for this contrastive study, it should also be mentioned that it assumes, with James (1980: 178), that correspondence in translation (or ‘translation equivalence’) is the best available tertium comparationis, i.e. basis for cross-linguistic comparison.

To explore the semantic prosody of cause in an English-Norwegian cross-linguistic perspective I will draw on data from the ENPC, both fiction and non-fiction, applying a methodology that exploits both the comparable and translational part of the corpus. The procedure is illustrated in Figure 2.

The bidirectional contrastive method outlined and exemplified in Figure 2 gives a step-by-step guide to the present study. In step 1 (a and b) all instances of cause are recorded – including their semantic prosody – both in the English original and translation parts of the ENPC, while in step 2 (a and b), all the Norwegian translations and sources of cause are recorded, providing examples of the following kind:

In the present paper single examples like (5) and (6) will be used to illustrate different points to do with semantic prosody. However, the prosodies themselves are not identified on the basis of single examples, but on the basis of recurrent instances of lexical units in context, as shown in Figure 3, where caused is Key-Word-In-Context in concordance lines from the ENPC. In order to manage the data, a FileMaker Pro database was created.

The third step taken in this investigation takes step 2a as its starting point and records all instances of the top three noun and verb correspondences of cause for further study of semantic prosody. One further step that could have been considered is the one indicated by the dotted horizontal arrow, i.e. to look at English translations of the top three Norwegian correspondences. This procedure is seen to be outside the scope of the investigation.

The diagonal arrow indicates the fourth, and final, step where comparable data, as revealed by the parallel corpus, in the two languages are compared. The two vertical arrows may serve as a control for translation effects; the lines are dotted to indicate that this is not one of the main concerns of this study.

We will now turn to the case study, applying the method outlined above.

4. Bidirectional contrastive analysis of cause

Before we move on to the cross-linguistic analysis proper, the data in the first step of the investigation will be presented. Tables 1 and 2 show the raw number of occurrences of cause in the ENPC, English originals and English translations, respectively. [16]

To identify the prosody of extended units of meaning with cause in the ENPC material, concordance lines with cause were scrutinised both in the English original and translated texts (see Figure 3). With the relatively small amount of data used here, few very clear patterns emerge, particularly in terms of collocational and colligational patterns. However, it is worth mentioning that for the verb cause, the data reveals the following colligations: cause+NP, cause+NP+inf., cause+NP+NP. When it comes to semantic preference, more uniform patterns emerge, both for the noun and verb, viz. typically for the verb: a subject (animate or inanimate) causes something undesirable (e.g. cause an explosion) or a subject (animate or inanimate) causes someone (to do) something undesirable (e.g. caused him to flee, caused her such grief), and similarly for the noun: the cause typically reflects an undesirable state, object or outcome (e.g. cause of death).

Based on this information on extended units of meaning with cause, the ENPC material reveals that cause, in the original texts, is found with a negative prosody in 29 out of 41 cases for the noun, and in 96 out of 114 for the verb (Table 1). In the English translations, cause as a noun has negative prosody in 18 out of 34 instances and as a verb in 33 out of 42. Observations that can be made on the basis of these admittedly small figures include a confirmation of the negative semantic prosody of cause with > 80% in originals. Although it only reaches 67% in the translated texts, it still reaches almost 80% for the verb (as compared to ca. 85% for the verb in the original texts). Thus, the negative prosody is more clearly present when cause is used as a verb; this is possibly related to the fact that the “aim or principle” sense of the noun cause is included (example 7), as was also the case in Stubbs’ study (1995: 171).

Finally, it is worth noticing that while the noun use is fairly stable between the original and translated texts in terms of distribution, cause as a verb is used far less in the translations from Norwegian than in the English original texts, viz. 114 (in originals) vs. 42 (in translations).

Moving on to step 2a in the analysis, where the Norwegian translations of cause are examined, we get the following breakdown in terms congruent vs. divergent (i.e. non-congruent) translation (Table 3).

 As shown in Table 3, most of the correspondences are congruent, i.e. overt correspondences that match the English original construction in terms of both form and meaning,  e.g. example 8, where a similar syntactic pattern with fikk ‘got’ is used to translate the pattern with caused. [17] The remainder are divergent, i.e. overt correspondences that do not match the English original ‘cause’-construction in terms of both form and meaning, e.g. example 9, where the Norwegian translation has an agent subject (hun ‘she’), followed by a verb (fornærmet ‘offended’) + direct object (sitt vertskap ‘her hosts’) + manner adjunct (grovt ‘tremendously’). The English original, on the other hand, has no agent, but an infinitive clause as subject followed by monotransitive caused. For the purpose of this paper, and unlike Johansson (2007), I have also included zero correspondences, i.e. cases where there is no overt correspondence in the translation, in the divergent group, as shown in example 10.

The analysis so far can be illustrated and summed up by looking at the bidirectional contrastive model again, adding some preliminary observations (Figure 4).

The congruent correspondences of the verb and noun uses of cause will differ, since the verb cause will correspond to a verb and the noun to a noun. In what follows, and still part of step 2a, we will therefore explore the verb and noun correspondences separately. [18] Step three in the analysis will also be part of the individual verb and noun analyses.

4.1 Verb correspondences

Since one of the aims of this study is to investigate the semantic prosody of Norwegian correspondences of expressions with cause, the divergent translations will be left out of the analysis, since they do not contain elements similar to those found in the original texts.

In the ENPC material, there are 83 congruent verb correspondences of cause (cf. Table 3), and as mentioned in the Introduction, 19 different Norwegian verbs have been used to translate these. The fact that cause can be found in syntactically different constructions may play a role as to this diversity. The ones recorded in the ENPC material include cause+NP, cause+NP+inf., cause+NP+NP, where the first two are the main patterns giving rise to congruent translations, cf. examples 11-13. These were also mentioned as colligational patterns in Section 4 above. However, this also suggests that there exists no one clear Norwegian counterpart to cause. The latter point is supported by the fact that major bilingual dictionaries such as Cappelens store engelsk-norsk ordbok and Engelsk stor ordbok offer a wide variety of translation equivalents. An overview of the three most common Norwegian correspondences in the ENPC material is given in Table 4.

Få (x til å) is typically used in constructions where cause is followed by an NP and a to-infinitive, as shown in example 11. Incidentally, this translation is not given as one of the alternatives in either of the dictionaries mentioned above.

Forårsake and føre til are typically used in contexts where cause is directly followed by an NP only, as in examples 12 and 13.

The next step (3) in the bidirectional analysis is to take the top three Norwegian correspondences and investigate their distribution and prosody in the Norwegian original texts, and in this manner establish to what extent the translational correspondences in fact can be called equivalents of cause, also in terms of semantic prosody. Table 5 presents the relevant data.

As the numbers indicate, få (x til å) is predominantly neutral, but is also found in negative contexts in about 23% of the cases. Føre til is found with negative prosody in 50% of the cases, while there are very few instances of forårsake overall, only three in total, making me unable to comment further on forårsake on the basis of this material. However, it might be worth noticing that while there are no occurrences of forårsake in the non-fiction original texts, it occurs 25 times in the translated non-fictional texts, 14 of which are translations of cause. This suggests that forårsake is subject to translationese in the ENPC material. [20] It also points to an interesting area for further study; to what extent can forårsake and cause really be said to be translation equivalents, as suggested both by English-Norwegian bilingual dictionaries and by the study by Dam-Jensen and Zethsen (2006) on the Danish cognate forårsage?

We can summarise the verb tendencies as follows:

• The closest Norwegian equivalent in terms of semantic prosody is føre til (‘lead to’).
• The most commonly used correspondence in translation få (x til å) (‘get x to’) does not have the same degree of negative semantic prosody as cause. However, data from the Norwegian original texts do show that få x til å (‘get x to’), even if it predominantly has neutral prosody, does occur with negative prosody (in about 23% of the cases in the ENPC material). [21]
• The third most common correspondence in the translations, forårsake (‘cause’/‘bring about’), is rare in the Norwegian original texts. This is substantiated by a search in a 28 million-word monolingual corpus, viz. Leksikografisk bokmålskorpus where forårsake occurs with a frequency of 22 per million words, as compared to the verb cause in the British National Corpus, occurring with a frequency of 204 per million words. Does this discrepancy in use make them less suited as direct counterparts?; it may at least suggest that the higher frequency of forårsake in translation makes it seem marked in certain contexts.

4.2 Noun correspondences

In this section, the noun cause and its Norwegian correspondences will undergo the same type of analysis as the verb in the previous section.

The noun has a congruent translation in 36 of the 41 instances in the ENPC material, and as shown in Table 6 there are three main congruent correspondences found in the material, årsak, grunn, and sak.

Årsak, the noun counterpart of forårsake, is by far the dominant correspondence and is used in more than half of the instances, suggesting that it has a broader set of meanings than sak and grunn, but not as rich as cause, since årsak does not include the ‘aim/principle’ meaning that is present in sak. An example of each of the top three correspondences is given in examples 14-16.

It could be argued that the use of cause as exemplified in 16 should have been left out of the analysis altogether, since it represents a non-causative meaning of cause. It has also been observed (cf. Stubbs 1995: 171) that this ‘aim/principle’ use of cause may contribute to a less clear-cut negative prosody of cause, since it often occurs in more favourable contexts than is generally the case for cause; see example 17. [22] It could also be argued that this extended meaning of cause is in fact not part of the same lexical item as the other instances of cause investigated here. However, as this use is part of Stubbs’ (1995) study, serving as the starting point for the present investigation, I decided to include it.

Analogous to the verb analysis, the use of the three most common correspondences are investigated in the original texts in order to establish their semantic prosody.

A revealing finding that emerges from Table 7 is that årsak is predominantly found with negative prosody in Norwegian. Grunn is the most commonly used noun of the three in the Norwegian original texts, and reveals a versatile prosody, but mainly neutral and negative. Finally, sak is a noun with a highly polysemous nature in Norwegian, and after having browsed through 252 occurrences of the lemma in the ENPC, I was left with a handful of instances where it had the meaning that overlaps with cause, i.e. “aim/principle to fight for”, as in example 18.

As already noted in the observation box in Figure 4, the most common noun sources of cause in the ENPC coincide with the top 3 congruent translations, i.e. if you search for cause in the English translated texts, årsak is the noun that most commonly occurs as its source, as shown in example 19.

In a cross-linguistic perspective this indicates a higher degree of mutual correspondence between årsak and cause than was the case with any of the verb correspondences, thus implying a tighter relation – and a stronger case for equivalence – between the two nouns in terms of form, meaning and use. [23] In order to test the mutual correspondence of the two nouns we need to go outside the scope of this investigation (cf. the dotted horizontal arrow in Figure 2), and look at the English translations of the Norwegian counterparts of cause. In this process it becomes clear that the mutual correspondence of årsak and cause may not be as strong as indicated above, since the typical English translation of årsak is reason rather than cause; see example 20. Incidentally, reason is also the preferred translation of Norwegian grunn.

The noun tendencies can be summarized as follows:

• The closest Norwegian equivalent of cause in terms of semantic prosody is årsak (‘cause’); årsak is also the most commonly used correspondence in the translations.
• Grunn (‘reason’) is predominantly neutral, but is also found in negative contexts; in 31 of the 87 instances with negative prosody, grunn has been translated into reason (i.e. 36%) and only 3 times into cause. The same tendency is seen for årsak (most commonly translated into reason).
• The ‘aim/principle’ reading is relatively rare both in English and Norwegian text (represented by sak). Stubbs found this use to contribute to a less negative count for cause overall (1995: 171). It is tempting to draw a similar conclusion for sak, but there is too little material to base this on.

What implications, both with regard to translation and contrastive analysis, do the cross-linguistic tendencies have when both verb and noun uses of cause and their Norwegian correspondences are taken into account? This is a particularly important question in the light of Tognini-Bonelli’s view that “only when functionally complete will a unit of meaning be available as a possible choice to the translator or for comparison to the contrastive linguist” (Tognini-Bonelli 2002: 79).

4.3 Cross-linguistic implications

Even if the Norwegian correspondences do not have the same degree of negative prosody as cause, data from the Norwegian original texts show that negative environments are not unnatural for e.g. føre til or årsak. Thus, in these cases it can be argued that the prosody is naturally preserved in translation by the use of Norwegian non-cognates of cause. However, as pointed out by Dam-Jensen & Zethsen (2006), even in cases where there are “clear negative evaluation tendencies, […] they are not so overwhelming that they prevent the verb from being used neutrally or even positively”, and in the case of Danish medføre ‘lead to’, the negative implications “are not as devastating as are those of forårsage” (Dam-Jensen & Zethsen 2006: 1620), or indeed as those of cause. These observations also apply to all the Norwegian correspondences of cause studied here. This, in turn, may point to a gap between the two languages; there are clear counterparts to cause in Norwegian, i.e. there are correspondences (as shown in the corpus data), but no real equivalent of cause. This is also supported by the fact that cause is clearly less used in English translations than in original text, i.e. there is no one good Norwegian source that would give rise to cause in the translation.

Norwegian and Danish forårsake/forårsage seem to share a negative prosody with cause, based on monolingual data for Norwegian (see Section 4.1) and on Dam-Jensen & Zethsen (2006) for Danish. However, evidence from the bidirectional translation corpus suggests that, in Norwegian, forårsake is not used to the same extent as cause. If this functionally matching item, also in terms of prosody, is overused in translation, it would be marked as translationese, and as such not be a perfect match after all. Again, we are left to infer that Norwegian does not have an item that matches cause at all four levels of the extended lexical unit, and that we in fact have to do with a kind of ‘lexical unit gap’ between the two languages. This should come as no surprise; both Tognini-Bonelli (2001, 2002) and Partington (1998) observe that there exist very few perfect cross-linguistic equivalents. Tognini-Bonelli further argues that a unit is only truly functionally complete when the contextual and functional dimensions are merged, i.e. units where

collocational and colligational patterning (that is lexical and grammatical choices respectively) are intertwined to build up a multi-word unit with a a specific semantic preference, associating the formal patterning with a semantic field, and an identifiable semantic prosody, performing an attitudinal and pragmatic function in the discourse. (Tognini-Bonelli 2002: 79)

The most common Norwegian correspondence of the verb cause, få (x til å), and the second most common correspondence of the noun cause, grunn, were typically found in neutral contexts in the Norwegian original texts. Whether these should be investigated as cases of translationese – on the grounds that both få (x til å) and grunn are overused in negative contexts in translation – is hard to determine on the basis of the ENPC data. It should also be stressed that items that are overused (or underused) in translation are not necessarily incorrect translations, as such, but they represent an unusual distribution of linguistic features in translation as compared to original text (influenced or not by the source language). Nevertheless, it is important that translators are made aware of the more neutral use of e.g. få (x til å) and grunn as compared to cause (x to) and cause.

5. Concluding remarks and future research

In this study, bidirectional translation corpora have been shown to offer a different kind of “real” data than monolingual (comparable) corpora, dictionaries, and introspection. As a result, it has been possible to tease out cross-linguistic tendencies involving the semantic prosody of units containing cause. The bidirectional contrastive method has been explored to study semantic prosody in English and Norwegian and clear advantages of using this model have emerged; here we can study cross-linguistic issues within one framework, such as:

• ability to study correspondences as they appear in contexts produced by several professional translators
• ability to study the distribution of cause and its correspondences in the two languages (in the same amount and type of data)
• ability to study the semantic prosody of cause and its Norwegian correspondences
• ability to study Norwegian counterparts of cause and compare their semantic prosody with that of cause
• ability to find out to what extent cause and its Norwegian correspondences in fact match in terms of prosody

However, a word of caution is in order; bidirectional corpora are, in general, smaller than most monolingual corpora. In other words, the results presented in this paper are tendencies emerging from a relatively small set of data. Thus, even with a model including both comparable and translation data, very often large-scale monolingual corpora will have to be consulted in order to arrive at more robust conclusions. Similarly, dictionaries and introspection may serve to complement studies carried out on the basis of bidirectional corpora.

Important questions for future research have emerged from the present study. For example, can the Norwegian verb føre til ‘lead to’ and the noun årsak ‘cause’ be said to be part of extended lexical units with a negative semantic prosody, as suggested by the ENPC data? In order to investigate whether this is a stable characteristic of these two items, large monolingual corpora need to be consulted. Similarly, another issue worth pursuing involves the fact that, in the ENPC, føre til is typically translated into English lead to; what will a study of lead to in a large monolingual corpus reveal with regard to semantic prosody? This is in fact something other scholars have looked into and Xiao & McEnery show that lead to does not have a clear semantic prosody (cf. Xiao & McEnery 2006: 117), a finding that is substantiated by Dam-Jensen & Zethsen, who found that it tends to have a negative profile, “but to a lesser degree than cause” (Dam-Jensen & Zethsen 2008: 210). In other words, the choice between lead to and cause is, at least partly, one that concerns semantic prosody; a speaker’s reason for using the extended lexical unit with cause is to stress the negative emotion/attitude. This is in accordance with what Sinclair observes for the use of budge vs. move:

We consider why people use this word [budge], why they do not use the common verb move, with which any use of budge can be replaced. Something does not budge when it does not move despite attempts to move it. From the perspective of the person who wants something moved, this is frustrating and irritation, and these emotions may find expression, because this is the ‘semantic prosody’ of the use of budge. (Sinclair 1998: 20)

With regard to forårsake as a translational correspondence of cause, where there seems to be a case of translationese, will a study of forårsake in a large monolingual corpus substantiate this claim? What will be revealed in terms of when and how it is used? And finally, what determines the choice between the nouns cause and reason in English? Both the ENPC and larger monolingual corpora may be used for this purpose.

With these questions for future research in mind, it may be argued that comparable monolingual corpora are better suited for cross-linguistic studies of semantic prosody than bidirectional translation corpora. However, without exploring data of the kind used in the present study, the tendencies, observations and questions for future research might never have emerged.

Although translations are generally true to the original text, also in conveying the semantic environment of lexical items, this investigation has shown that when we use the bidirectional method to look into cross-linguistic correspondences in more detail, these correspondences are revealed to be used with a less clear-cut (negative) prosody when they occur in an arguably more natural environment, i.e. in original text. A case in point is the favoured Norwegian translation of the verb cause: få (x til å), which in the original texts in the ENPC shows a tendency towards a neutral prosody, rather than a negative prosody corresponding to that of cause.

Notes

[1] I would like to thank Hilde Hasselgård, Magnus Huber, and an anonymous reviewer for valuable comments on earlier versions of this paper.

[2] The base form of a word in italics is in this paper used to refer to the lemma, e.g. cause, føre til (‘lead to’).

[3] Morley and Partington (2009) equate semantic prosody with the evaluative meaning of linguistic units, while Stubbs (forthcoming), with reference to Sinclair, equates it with evaluation and communicative purpose of extended lexical units.

[4] In this paper correspondence rather than equivalence is used to refer to “forms which are observed to correlate” (Johansson 2007: 23), including both sources and translations, since what we obeserve in a corpus are correspondences and not necessarily equivalents. “Analysing the correspondences we may eventually arrive at a clearer notion of what counts as equivalent across languages” (Johansson 2007: 5).

[5] See also Tognini-Bonelli (2002).

[6] Thompson and Hunston, as quoted in Morley and Partingtion (2009: 141), also argue for this binary divide of good vs. bad.

[7] “… having started with a node as a core, they [extended units of meaning] have incorporated other words in the co-text that appeared to be co-selected with it and form a regular pattern. They are, therefore, multi-word units in that they are defined by the strict correlation existing between a node and its context” (Tognini-Bonelli 2001: 19).

[8] Tognini-Bonelli points out that ideally she would also have used a translation corpus if it had been available; since it was not, she relies on translational information “taken from reference books or intuition” (Tognini-Bonelli 2002: 82–83).

[9] Stewart’s (2009) study on translating semantic prosody from English into Italian should also be mentioned here, although it differs from the other studies mentioned in that it does not make use of comparable data and in that it is a pedagogical experiment in the teaching of translation.

[10] In a later article, Dam-Jensen & Zethsen (2008) are concerned with the prosody of cause and its near synonym lead to, but it is not a cross-linguistic study as such, since it rather tests Danish translators’ intuition with regard to the prosody of English expressions including cause and lead to.

[11] See Johansson (2007: 12) for an earlier and similar, but unrelated, project on Serbo-Croatian and English.

[12] The translations are published translations by different translators and were not commissioned especially for this project.

[13] For a discussion of other advantages with this model, see Aijmer & Altenberg (1996), and for a discussion of problems related to the use of translated texts for language research, see e.g. Johansson (1998: 6).

[14] Step 2b will not play a major role in this investigation; it will merely serve as a cross-linguistic check of how and to what extent the Norwegian sources of cause match the translations.

[15] As can be seen in Figure 3, the concordance lines in the ENPC browser stop at s-unit boundaries. In cases where this was not considered sufficient context, the extended context function in the ENPC search interface was used (cf. Ebeling 1998).

[16] Raw figures are used, as comparable amounts of text are found in each of the subcorpora in the ENPC, i.e. around 650,000 words in each of the four components: English originals, English translations, Norwegian originals, Norwegian translations.

[17] For an overview of classification of correspondences in terms of congruence, see Johansson (2007: 25).

[18] As already mentioned, step 2b, concerned with the Norwegian sources of cause, will not be part of the main analysis. Suffice it to say that there is a discrepancy between translation and source for the verb use (cf. bottom right-hand observation box in Figure 4), i.e. the Norwegian sources of cause in the ENPC are not the same as the translational correspondences of cause.

[19] Cf. Dam-Jensen & Zethsen’s (2006: 1617ff) study on the Danish cognate forårsage.

[20] Translationese is a term used to mean the “systematic influence on target language (TL) from source language (SL)” Gellerstam (1986: 88). Johansson (2007: 32) uses “translation effects” to refer to the same phenomenon of what he calls overuse and underuse in translations as compared to original text.

[21] Although not part of the present investigation, it is also interesting to note that the typical translation of Norwegian få (in the construction få x til å) is make; cause is hardly used at all.

[22] When cause has an ‘aim / principle’ reading it would be logical to suggest that it occurs in patterns that are different from the more common reading: ‘a thing that makes things happen as a consequence of something’. In the material investigated here, it is hard to draw any conclusions, but it is tempting to suggest that attributive adjectives tend to occur more often, as e.g. “the good cause”. In terms of semantic preference, however, it is clearly used in more favourable contexts, thus contributing to a more positive prosody.

[23] The term ‘mutual correspondence’ was first used and defined by Bengt Altenberg as “the frequency with which different (grammatical, semantic and lexical) expressions are translated into each other” (Altenberg 1999: 254).

Corpora

BNC = The British National Corpus, version 3 (BNC XML Edition). 2007. Distributed by Oxford University Computing Services on behalf of the BNC Consortium. http://www.natcorp.ox.ac.uk/

ENPC = The English-Norwegian Parallel Corpus. 1994–1997. Compiled by Stig Johansson (project leader), Knut Hofland (project leader), Jarle Ebeling (research assistant), Signe Oksefjell (research assistant). Dept. of British and American Studies, University of Oslo. http://www.hf.uio.no/ilos/english/services/omc/enpc/

Leksikografisk bokmålskorpus (‘Lexicographic corpus of Norwegian bokmål’). http://www.hf.uio.no/iln/tjenester/kunnskap/samlinger/bokmal/veiledningkorpus/

References

Aijmer, K. & B. Altenberg. 1996. “Introduction”. Languages in Contrast: Papers from a Symposium on Text-Based Cross-Linguistic Studies, Lund 4–5 March 1994, ed. by K. Aijmer, B. Altenberg & M. Johansson. Lund: Lund University Press.

Altenberg, B. 1999. “Adverbial connectors in English and Swedish: Semantic and lexical correspondences”. Out of Corpora: Studies in Honour of Stig Johansson, ed. by H. Hasselgård & S. Oksefjell, 249–268. Amsterdam & Atlanta, GA: Rodopi.

Berber Sardinha, T. 1999. “Padrões lexicais e colocações do português”. Paper presented at the symposium Processamento Computacional do Português, 9, InPLA, PUCSP, Brazil. http://www2.lael.pucsp.br/~tony/temp/publications/1999padroes_inpla.pdf

Berber Sardinha, T. 2000. “Semantic prosodies in English and Portuguese: A contrastive study”. Cuadernos de Filología Inglesa 9(1): 93–110.

Chesterman, A. 1998. Contrastive Functional Analysis. Amsterdam & Philadelphia: John Benjamins.

Dam-Jensen, H. & K.K. Zethsen. 2006. “Pragmatic patterns and the lexical system – A reassessment of evaluation in language”. Journal of Pragmatics 39: 1608–1623.

Dam-Jensen, H. & K.K. Zethsen. 2008. “Translator awareness of semantic prosodies”. Target 20(2): 203–221.

Ebeling, J. 1998. “The Translation Corpus Explorer: A browser for parallel texts”. Corpora and Cross-Linguistic Research: Theory, Method, and Case Studies, ed. by S. Johansson & S. Oksefjell, 101–112. Amsterdam & Atlanta, GA: Rodopi.

Ebeling, J. 2000. Presentative Constructions in English and Norwegian: A Corpus-Based Contrastive Study. Oslo: Unipub.

Gellerstam, M. 1986. “Translationese in Swedish novels translated from English”. Translation Studies in Scandinavia: Proceedings from the Scandinavian Symposium on Translation Theory (SSOTT) II, ed. by L. Wollin & H. Lindquist. Malmö: Gleerup.

Hunston, S. 2002. Corpora in Applied Linguistics. Cambridge: Cambridge University Press.

Hunston, S. 2007. “Semantic prosody revisited”. International Journal of Corpus Linguistics 12(2): 249–268.

James, C. 1980. Contrastive Analysis. London & New York: Longman.

Johansson, S. 1998. “On the role of corpora in cross-linguistic research”. Corpora and Cross-Linguistic Research: Theory, Method, and Case Studies, ed. by S. Johansson & S. Oksefjell, 3–24. Amsterdam & Atlanta, GA: Rodopi.

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

Johansson, S., J. Ebeling & S. Oksefjell. 1999 (2002). The English Norwegian Parallel Corpus: Manual. Department of British and American Studies, University of Oslo. http://www.hf.uio.no/ilos/english/services/omc/enpc/ENPCmanual.pdf

Johansson, S. & K. Hofland. 1994. “Towards an English-Norwegian Parallel Corpus”. Creating and Using English Language Corpora, ed. by U. Fries, G. Tottie & P. Schneider, 25–37. Amsterdam & Atlanta, GA: Rodopi.

Krzeszowski, T.P. 1990. Contrasting Languages: The Scope of Contrastive Linguistics. Berlin & New York: Mouton de Gruyter.

Lado, R. 1968 [1957]. Linguistics Across Cultures: Applied Linguistics for Language Teachers. Ann Arbor: The University of Michigan Press.

Louw, B. 1993. “Irony in the text or insincerity in the writer? The diagnostic potential of semantic prosodies”. Text and Technology: In Honour of John Sinclair, ed. by M. Baker, G. Francis & E. Tognini-Bonelli, 157–175. Philadelphia & Amsterdam: John Benjamins.

Louw, B. 2000. “Contextual prosodic theory: Bringing semantic prosodies to life”. Words in Context: A Tribute to John Sinclair on His Retirement, ed. by C. Heffer & H. Saunston. www.revue-texto.net/docannexe/file/124/louw_prosodie.pdf

Morley, J. & A. Partington. 2009. “A few Frequently Asked Questions about semantic – or evaluative – prosody”. International Journal of Corpus Linguistics 14(2): 139–158.

Partington, A. 1998. Patterns and Meaning: Using Corpora for English Language Research and Teaching. Amsterdam & Philadelphia: John Benjamins.

Partington, A. 2004. “‘Utterly content in each other’s company’. Semantic prosody and semantic preference”. International Journal of Corpus Linguistics 9(1): 131–156.

Sinclair, J. 1987. “The nature of the evidence”. Looking Up: An Account of the COBUILD Project in Lexical Computing, ed. by J. Sinclair. London & Glasgow: Collins ELT.

Sinclair, J. 1991. Corpus, Concordance, Collocation. Oxford: Oxford University Press.

Sinclair, J. 1996. “The search for units of meaning”. Textus IX: 75–106.

Sinclair, J. 1998. “The lexical item”. Contrastive Lexical Semantics, ed. by E. Weigand, 1–24. Amsterdam & Philadelphia: John Benjamins.

Stewart, D. 2009. “Safeguarding the lexicogrammatical environment: Translating semantic prosody”. Corpus Use and Translating: Corpus Use for Learning to Translate and Learning Corpus Use to Translate, ed. by A. Beeby, P. Rodríguez Inés & P. Sánchez-Gijón, 29–46. Amsterdam & Philadelphia: John Benjamins.

Stewart, D. 2010. Semantic Prosody: A Critical Evaluation. New York & London: Routledge.

Stubbs, M. 1995. “Collocations and semantic profiles: On the cause of trouble with quantitative studies”. Functions of Language 2(1): 23–55.

Stubbs, M. 2007. “Quantitative data on multi-word sequences in English: The case of the word world”. Text, Discourse and Corpora: Theory and Analysis, ed. by M. Hoey, M. Mahlberg, M. Stubbs & W. Teubert, 163–189. London: Continuum.

Stubbs, M. 2009. “The search for units of meaning: Sinclair on empirical semantics. Memorial article: John Sinclair (1933–2007)”. Applied Linguistics 30(1): 115–137.

Stubbs, M. Forthcoming. “Sequence and order: The neo-Firthian tradition of corpus semantics“. To appear in Corpus Perspectives on Patterns of Lexis, ed. by H. Hasselgård, J. Ebeling & S.O. Ebeling. Amsterdam & Philadelphia: John Benjamins.

Tognini-Bonelli, E. 2001. Corpus Linguistics at Work. Amsterdam & Philadelphia: John Benjamins.

Tognini-Bonelli, E. 2002. “Functionally complete units of meaning across English and Italian: Towards a corpus-driven approach”. Lexis in Contrast: Corpus-Based Approaches, ed. by B. Altenberg & S. Granger, 73–95. Amsterdam & Philadelphia: John Benjamins.

Whitsitt, S. 2005. “A critique of the concept of semantic prosody”. International Journal of Corpus Linguistics 10(3): 283–305.

Xiao, R. & T. McEnery. 2006. “Collocation, semantic prosody, and near synonymy: A cross-linguistic perspective”. Applied Linguistics 27(1): 103–129.

Zethsen, K.K. 2006. “Semantic prosody: Creating awareness about a versatile tool”. Tidsskrift for sprogforskning 4(1–2): 275–294.

Dictionaries

Cappelens store engelsk-norsk ordbok. 1988 [2nd ed.]. Edited and revised by Herbert Svenkerud. Oslo: Cappelen.

Engelsk stor ordbok. 2008. Edited by Vibecke C.D. Haslerud & Petter Henriksen. Oslo: Kunnskapsforlaget.