An English-Norwegian contrastive analysis of downtoners , more or less

In this paper we present a contrastive study of the downtoners more or less and mer eller mindre. After establishing that these two binomials have developed along similar paths in English and Norwegian, can be found to occur in the same syntactic environments in the two languages and, indeed, have the same meanings, we seek explanations as to why they are not used as translations of each other in the E nglish-Norwegian Parallel Corpus to the extent that would be expected of two such seemingly perfect matches. The picture that emerges is a complex one. First, more or less is much more frequent than mer eller mindre, both in source (original) and target (translation ) texts. Next, more or less corresponds to many more Norwegian expressions in t he data, than vice versa, signalling a wider repertoire of use than its Norwegian counte rpart. Finally, mer eller mindre is used as a premodifier and as a degree adjunct in approxi mately the same number of instances, while the degree adjunct use is dominant in the Eng lish data for more or less . The paper introduces the reverse mutual correspondence measur e to complement the more established mutual correspondence measure, introduc ed by Altenberg (1999), to add another dimension to contrastive analysis based on bidirectional translation corpus data.

1 Introduction This paper reports on a contrastive study that takes the English downtoner more or less as its starting point.Our attention was drawn to this pattern after having extracted 3-gram lists from the extended fiction part of the English-Norwegian Parallel Corpus, the ENPC+.With an expected frequency of six, more or less actually occurs 41 times (31 pmw) in the English original (source) texts. 2 Our interest was further piqued when it was revealed that its frequency turned out to be much higher in the translated English material (70 occurrences, or 49 pmw). 1 The authors wish to thank two anonymous reviewers for their invaluable and extensive feedback on an earlier version of the article. 2 The expected frequency was calculated by multiplying the number of times more occurs by the number of times less occurs and dividing the sum by the total number of s-units (orthographic sentences) in the corpus.Note that we only include the content words more and less in the calculation, not or.See Ebeling & Ebeling (2013: 70) and Gries (2008: 6, n3) for a discussion of how to calculate expected frequencies of multi-word words.
Moreover, when looking at the cross-linguistic data, we find that the Norwegian cognate binomial, mer eller mindre, is not as frequent as more or less, nor are the two found to correspond to the degree one might expect in a parallel corpus.
With regard to the meaning and use of more or less, the Oxford English Dictionary (OED) establishes the following: a. more or less (also †less or more, †more or min, etc.): in a greater or lesser degree; to a greater or lesser extent; to all intents and purposes, virtually, essentially; nearly, almost.Also in negative contexts: not at all, hardly at all.Also occas.as an adjective phrase (esp. in Linguistics): designating a question which is answerable only in terms of degree.
The OED entry is interesting in at least two respects: (i) although more or less is an irreversible binomial today, it seems to have been reversible in earlier times, and (ii) that its now obsolete equivalent more or min is etymologically related to present-day Scandinavian mer(e) eller mindre (cf.entries for more and min in the OED).
In a wider cross-linguistic context one can note that similar, and even cognate, expressions are attested across many Germanic and Romance languages, e.g.: With a view to their historical background, diachronic and crosslinguistic development and also similarity of form, it is surprising that English more or less and Norwegian mer eller mindre are found as each other's translations in less than half of the cases in the ENPC+ material.
English more or less has mer eller mindre as its translation in 22 of the 41 instances in original texts (approx.53%; see further Table 3).Example (1) shows an instance where the two correspond, while in example (2) a different Norwegian correspondence is used, namely på sett og vis ('in a way'). (1) One of them still looked more or less intact -… [TaFr1E] Et av dem så fortsatt mer eller mindre intakt ut -… [TaFr1TN] (2) It more or less lets Danny Boy out -if he wanted the whole thing, … [TaFr1E] Det utelukker på sett og vis Danny Boy -hvis han hadde villet ha hele greia, ... [TaFr1TN] Furthermore, of the 70 instances of more or less in the English translated (target) texts, only 11.4% had mer eller mindre as their source.In fact, 26 different Norwegian source expressions were recorded.In the opposite direction of translation, we can note that Norwegian mer eller mindre was translated into more or less in around 42.8% of the cases, while more or less was used as its source in around 70% of the cases (see Section 5 for details and tables).With these cross-linguistic observations as our starting point, we will investigate this apparently skewed relationship between the cognate patterns in more detail.Downtoners may, according to Quirk et al. (1985: 445, 601), have at least two main (syntactic) functions, viz.that of premodifier or subjunct adverbial.It will therefore be of interest to survey the different uses of the cognates in contrast in order to establish whether function plays a role in the choice of cross-linguistic correspondence.Equally important will be the study of other correspondences, as these may give indications as to the extent of the semantic field to which the downtoners belong.This leads us to the overarching question of why two seemingly perfect cross-linguistic matches should reach a mutual correspondence of only around 50%.In this context, we will introduce the concept of a reverse mutual correspondence and explore what insights such a measure can add to the contrastive analysis.
More specifically, the questions we seek answers to in this paper include: • Is it possible to determine what role the syntactic function plays for the downtoners to correspond to each other?• What can the contrastive analysis uncover about the use and semantics of the downtoners under study here?
The outline of the paper is as follows.Section 2 gives a very brief overview of the corpora consulted, while Section 3 provides some theoretical background to the study of intensifiers in general and the downtoner more or less in particular.Section 4 contains some pertinent observations on the basis of monolingual data.Then we move on to the English-Norwegian contrastive case study, where we will give a detailed analysis of the cross-linguistic data (Section 5).Section 6 offers some concluding remarks.

Corpora
The main source of data for this study is the English-Norwegian Parallel Corpus+ (ENPC+), which is a bidirectional corpus containing English and Norwegian original fiction texts from the late 1900s and early 2000s, and their translations into Norwegian and English.It is an expanded version of the fiction part of the English-Norwegian Parallel Corpus (cf.Johansson 2007, Ebeling & Ebeling 2013).Each of the four parts of the corpus (English originals, Norwegian translations, Norwegian originals, English translations) contains around 1.3 million running words, altogether amounting to 5.2 million words.
Additional sources of data include the British National Corpus (BNC) for monolingual English material and the Leksikografisk bokmålskorpus (LBK) for monolingual Norwegian material.They are both 100-million-word corpora, "designed to represent a wide crosssection of current British English" (http://www.natcorp.ox.ac.uk/) and Norwegian bokmål.In a few places we also draw on material from the Corpus of Contemporary American English (COCA) and a relatively small corpus of spoken Norwegian (NoTa).
In some cases, and to enhance comparability with the ENPC+ material, only the fiction parts of the BNC (ca.16 mill.words) and LBK (ca.13.8 mill words) will be used.

3
Background Over the years, intensifiers, including downtoners, have been subject to extensive research in general (English) language research (Quirk et al. 1985, Huddleston & Pullum 2002, Stoffel 1901, Borst 1902, Bolinger 1972, Bäcklund 1973, Partington 1993, Paradis 1994, Erman 2014), in historical and diachronic linguistics (Fettig 1935, Peters 1994, Nevalainen & Rissanen 2002, Méndez-Naya (ed.) 2008), in spoken language research (Altenberg 1991, Metsä-Ketelä 2006), in sociolinguistics (Xiao & Tao 2007), in learner language research (Granger 1998, Lorenz 1999), in English as a Lingua Franca research (Metsä-Ketelä 2006), in variation studies (Wittouck 2011), in contrastive analysis (Bardas 2008), and many more.An exhaustive survey of previous research will not be attempted here, but the above-mentioned publications bear witness to a long-standing and wide-ranging tradition in the study of intensifiers in English.However, there seem to be relatively few studies devoted to multi-word intensifiers in general, and to more or less in particular, although there are some exceptions such as Metsä-Ketelä (2006), which we will return to below, and Lorenz (1999: 73-74) in a short section on phrasal items used as premodifying intensifiers.
Unfortunately we cannot do justice to all sides of the extensive research referred to above; however, some observations about more or less, or more broadly downtoners, or even more broadly intensifiers, are in order.
The following quote from Stoffel (1901: 129) very much sets the scene: In the preceding section of this book I have set forth how certain Intensive Adverbs expressing completeness of degree [e.g.very, pure], have gradually come to mean a high or a very high degree of quality.[...] In the present section I intend to treat in the same manner certain other adverbs of degree, which, to distinguish them from Intensives, I have ventured to designate as "down-toning" adverbs.
Adding the dimensions of delexicalisation and grammaticalisation, Lorenz claims that "an intensifier becomes reduced to its modulating, scaling function -upgrading or downtoning the item in its focus without expressing any denotational meaning of its own" (Lorenz 2002: 146).
In their overview of intensifiers, Quirk et al. (1985: 589ff)  The subtypes provide nothing more than a rough guide to semantic distinction, because (i) the varying effects of intensifiers represent a semantic gradient, which is obscured by a clear-cut division into classes; (ii) some intensifiers are sometimes used for different effects; and (iii) speakers vary in their use of intensifiers.(ibid.: 590) Indeed, downtoners, for instance, are said to have at least two main functions in the clause, namely that of subjunct adverbial or modifier.The term "subjunct" seems to be unique to Quirk et al. (1985), according to Hasselgård's (2010) survey of how different reference grammars have classified adverbials.Quirk et al. (1985: 566) define subjuncts as having a subordinate role compared to other clause elements, e.g. they may be subordinate to the whole clause, or more typically in our context, they "may be subordinated to an individual clause element […] or even to an element forming part of a clause element" (ibid.: 567).Since "subjunct" as a category appears to be a controversial one, we choose to follow Hasselgård, and Biber et al. (1999), in including adverbial intensifiers among the adjuncts. 4We will use the term degree adjunct; according to Hasselgård (2010: 29), degree adjuncts: typically specify the intensity with which something is carried out and may be elicited by the question to what extent.Related to this is the function of intensifier, which includes amplifier and downtoner.
Example (3) shows the degree adjunct use of more or less,5 while example (4) shows its function as a (regular) modifier of an adjective and example (5) as a modifier of another adverb:6 (3) They knew their business, more or less.(BNC:A0N 447) (4) The street was more or less deserted.(BNC: G0P 3303) (5) ... she had been driving more or less continuously since just after nine that morning.(BNC: JYF 239) In their discussion of degree adverbs, Huddleston and Pullum (2002: 723)  to account for a seemingly nice order of intensifiers, this partly breaks down when scrutinising the use of the compromisers, in her case, on the syntagmatic axis.A similar observation is also made by Quirk et al. in an extensive footnote (1985: 446), where they show that e.g.quite can function as an amplifier (She's quite right, i.e. she is absolutely/ completely right) or as a downtoner (That's quite good, i.e. that's fairly/rather good). 7Notice also the (old-fashioned) use of rather in contexts such as 'Did you enjoy the party?' 'Rather!' (ibid.591, note).This is reminiscent of Aijmer's (2007) observation with regard to the development of "semantic and pragmatic polysemies" of actuality adverbs.She attributes such a development to grammaticalisation processes, echoing Lorenz' (1999Lorenz' ( , 2002) ) view that intensifiers are particularly prone to grammaticalisation.Metsä-Ketelä (2006), comparing more or less in native and (English as a Lingua Franca) non-native spoken academic English (MICASE and ELFA corpora), makes several pertinent observations about its use. 8She recognises three sub-functions of more or less; one which she terms minimizing, where more or less is synonymous with simply, only or just, and where it has dismissive connotations.The second use is where more or less combines with the same, or a similar word, e.g.equivalent, and where the whole pattern is used to "compare similarities between two or more concepts or entities" (ibid.: 137).The third use she calls approximating quantities.This is where more or less for instance modifies a quantity such as "all people in the western societies" (ibid.: 139).In the latter two cases, more or less seems to function as approximator in Quirk et al.'s (1985) terminology, i.e. almost or nearly are good paraphrases of more or less.Lorenz (2002) does not discuss more or less as such, but adverbs inly used as premodifiers in the light of delexicalisation.He sees a development where evaluative, comparative and modal adverbs over time may gradually lose or "shed their conceptual meaning" (ibid.147) and become purely scalar.His examples include very, drawn from the modal class of adverbs.This class also includes virtually, which is very close to more or less in meaning in certain contexts.However, Lorenz says that many of these adverbs "are more characteristically known as clause-level modal adverbs" and "not all emphasizers are therefore equally likely to become grammaticalised as adjective intensifiers" (ibid.152).As we shall see, more or less, and to a lesser extent mer eller mindre, is mostly used as a clause-level modifier.

4
A preamble to the study proper Before we embark on the contrastive study a few observations made on the basis of monolingual corpora are in order.A quick survey of the binomial more or less in the BNC reveals that it is more frequent in the spoken part of the corpus than in the written part, viz.44 vs. 26 occurrences per million words.It is also more frequent overall than in the written fiction part of the corpus where it occurs 21.6 times per million words (see further Section 5.1).
With regard to frequency of use in the fiction part of the LBK corpus we can note that mer eller mindre occurs 20 times per million words, which is roughly the same as the frequency of more or less in the fiction part of the BNC.
When it comes to meaning and how to classify the many instances of more or less and mer eller mindre, it becomes clear that this is perhaps even more challenging than we have been led to believe based on what has been maintained above.It is not, for instance, the case that there is always a difference in meaning between almost (an approximator in Quirk et al.'s terms) and more or less (a compromiser in Quirk et al.'s terms), which lends evidence to Quirk et al.'s words of caution about the semantic subtypes of intensifiers.In the following example, (6), more or less means almost, since a lump cannot be more than intact.
(6) The lump is more or less intact, though crumbs of it stick to Perry's fringe.(BNC: HGL 1608) Pinning down the exact meaning(s) of an expression or phrase is of course made even more difficult when that expression or phrase contains items which on their own have vague meanings, such as more or less.
There is, however, one use of more or less where it is easy to pin down its meaning, and that is when the expression roughly means one more or one less of a quantity, as in the next example.
(7) ..., except that from one year to another a pint more or less milk might be ordered for the teas.(BNC: HA4 688) Such uses will not form part of the current study; also, they are very few and far between.Metsä-Ketelä mentions that more or less can be seen as a hedge and used as a "softener" (2006: 132), especially in face-to-face situations or encounters.Similar uses are attested in the BNC material where more or less can be found after hesitation markers such as er or erm in the spoken material, within brackets, between commas and dashes and at the very end of clauses or sentences in the written component of the BNC.It is also easy to find occurrences where more or less has a dismissive connotation, as in ( 14). ( 14) Ingard shares were still quoted on the Stock Exchange, and we had evidence that they were more or less worthless.(BNC: H0D 2100) The discussion so far has shown that more or less works equally well as an approximator (e.g.example ( 6)), a compromiser (e.g.example ( 12)), and as a diminisher (e.g.example ( 14)).Moreover, when preceded by the word nothing it can act as a maximiser as well, as shown in ( 15).
(15) This is nothing more or less than an abuse of your professional integrity.(BNC: HGJ 2005) [= this is actually an abuse of ...] 9 When surveying the Norwegian data in the fiction part of the monolingual corpus LBK, we find similar uses of the Norwegian cognate to those recorded for English more or less.However, one is struck by the fact that the maximiser function of the binomial mer eller mindre is one of its most frequent uses.In Norwegian the word (h)verken is used in the 9 The expression nothing more or less (than) does not seem to have the same fixedness as more or less on its own.For example you could have only one part of it.This is nothing more than...This is nothing less than..., comparable to Norwegian Dette er intet mindre enn… So the question is if this is the same phrase, or if speakers, perhaps unconsciously, confuse the various ways of saying more or less the same thing.
However, one could perhaps argue that this is actually a different construction altogether where mer eller mindre is an integral part of the pattern (h)verken mer eller mindre.
We have been able to establish, based on monolingual data, that English and Norwegian have a cognate binomial -more or less and mer eller mindre -which seem almost identical in use and meaning.Our null-hypothesis for the cross-linguistic part of the study is therefore that the two binomials are equal in use and meaning.
To sum up the discussion so far we could venture an analysis of the uses of more or less (and mer eller mindre) along the following lines.
If we were to position more or less, together with a few other downtoners, in relation to how they relate to the truth condition of the proposition they tone down, we can argue that what is almost true is not in fact true, but that what is nothing more or less true is indeed true.On such a scale more or less occupies the middle position of not committing to whether the proposition is true or not true.
(hardly) almost more or less (partly) verken mer eller mindre nothing more or less (than) ← not true true → However, specifying the truth-conditional semantics of an expression only tells us part of the story, not least when it comes to downtoners.
Although the use of one downtoner rather than another seems to be a 10 (H)verken, written with or without an intial h means 'neither of two (or more) parts', comparable to 'neither ... nor' in English as in han kan verken lese eller skrive = he can neither read nor write.
purely scalar one (cf.Lorenz 2002: 147-148), the speaker's assessment as to where a downtoner belongs along this scale seems to play a role as well.Consider hardly, for instance, which tells us not only that something is not true, but adds that this is the opinion of someone, i.e. the speaker's evaluation along this scale.

5
The study 5.1 Distribution and comparison with monolingual corpora One of the things that drew our attention to more or less and its Norwegian counterpart mer eller mindre was, as noted in the introduction, the difference in frequency between the two items in comparable amounts of source data and the difference in frequency between source (original) and target (translation) texts.Table 1 shows that more or less overall is more frequent in the English part of the corpus than mer eller mindre is in the Norwegian part.There is also a marked difference between source and target texts in both languages.We cannot perform an in-depth and full-scale study of more or less and mer eller mindre in other corpora here.However, a few comparisons and observations are in order, and could be used as a starting point for further studies.In the BNC, more or less is more frequent in spoken than in written English and more frequent in academic writing than in fiction.In the Corpus of Contemporary American English (COCA), more or less is less attested in the spoken part of the corpus than in the written part, but again it is more frequent in academic writing than in fiction.The Norwegian LBK corpus shows the same tendency, with mer eller mindre being more frequent in academic writing than in fiction.The number per million words for the spoken Norwegian (NoTa) material is lower than for the BNC, but higher than for COCA. 14According to Lorenz (2002: 143) the lexico-grammatical category of intensifiers is mainly employed to achieve expressivity.As such, it thrives on novelty, i.e. on innovation and semantic change.Such change is typically heralded in the more dynamic text-types, occurring in spoken rather than written languages, in informal rather than formal conversation, between younger rather than older speakers.
Table 2 seems only partly to support Lorenz's observation if we contend that academic writing is less dynamic, containing less dialogue than fiction.On the other hand, our data seem to support Lorenz's statement in that we find more occurrences in the ENPC+, which contains mainly crime fiction, arguably containing a larger portion mimicking informal conversation, than in the fiction part of the BNC and COCA.Obviously, the matter needs to be investigated further.

Mutual correspondence
In Altenberg (1999) a measure, termed mutual correspondence (MC), was introduced to quantify the strength of attraction between two items in a bidirectional translation corpus.When we apply this measure to more or less and mer eller mindre (Table 3), we get an MC of 50%, which is considered fairly high, but much lower than we would expect.Our null-hypothesis, which was that the two items were equivalent, would predict an MC of 100%.This is, however, seldom the case when dealing with translation of fiction, or indeed, translation in general, since there nearly always will be cases where a clause or phrase is left untranslated.Still, an MC of only 50% seems low based on what we have learned about the meaning and use of the two binomials.Based on these numbers it seems as if the Norwegian translators more readily go for the cognate item than the English translators do.However, since we have twice as many instances of more or less than of mer eller mindre, this can only be a tentative explanation for this translation bias.
When we look at the instances where the two binomials do not correspond to each other it is noteworthy that many of these other correspondences are multi-word synonyms and not simplexes.That is, the translators do not resort to synonymous, single-word adverbs such as almost, nearly or approximately when translating mer eller mindre into something other than more or less.Correspondences of more or less other than mer eller mindre are ikke noen saerlig ('not something special'), på sett og vis ('in a way'), stort sett ('by and large'), så godt som ('as good as'), så å si ('so to speak'), sånn noenlunde ('just about') and til en viss grad ('to a certain degree').
A similar, albeit not so well attested, tendency can be noted when going from Norwegian into English.The following two multi-word downtoners are found: so to speak and to some degree or other.
Multi-word correspondences were also chosen for the two instances of verken mer eller mindre, but this was to be expected.The attested correspondences are neither more nor less and no less.
The single-word correspondences of mer eller mindre are basically, just, and more when going from Norwegian to English and egentlig ('really') and nesten ('almost') when going the other way.There are also a number of instances where the downtoners have not been translated, and one could argue that in these cases important bits of the original are lost in the translation.
In ( 17), where just is used, we can note the dissmissive connotation, while it is debatable whether the use of nesten ('almost'), in example ( 18) captures the meaning of the original.On the other hand, since more or less undid has been translated by slo meg nesten ut ('knocked me almost out'), the use of nesten ('almost') in the Norwegian translation is of course acceptable, even though it alters the meaning of the original slightly.(19) They sit down and talk, and if I can more or less remember the gist and flow of their conversation, it's because I asked Katya to play the scene again after the movie was finished.[PaAu1E] De setter seg og snakker sammen, og grunnen til at jeg Ø husker hovedtrekkene og gangen i samtalen deres, er at jeg ba Katya spille scenen om igjen etter at filmen var slutt.[PaAu1TN] Gloss: They sit down and talk together, and the reason to that I Ø remember ...
(20) Amputasjonen fins dokumentert på videoer i en serie eksklusive samlinger; klipp fra disse videoene har dessuten vaert vist, copyright-belagt og lukrativt, på tv-stasjoner mer eller mindre verden over, men ingen -ingen, som Bella har understrekethar sett lillefingeren live.[PeRy1N] The amputation is documented on videos in a series of exclusive collections; clips of these videos have also been showncopyrighted and raking in the cash -on TV stations Ø all over the world but no-one -no-one, Bella emphasises -has seen the little finger live.[PeRy1TE] Gloss: ..., on TV stations more or less all over the world ...

Reverse mutual correspondence
The MC measure has so far, to our knowledge, only been used going from original to translation, i.e. how often corresponding items are used as translations of each other.Given the potential of the corpus at hand, i.e. its bidirectional structure, the MC seems to report only half of the story.Thus, we were curious to see what happens if we go the other way and calculate the mutual correspondence based on the translations and their sources, viz. the reverse MC (rMC), i.e. when we calculate a percentage based on the number of times our items have each other as source.
In the case of English, more or less is attested 70 times in the target texts, but only 10 of these have mer eller mindre as their source.For Norwegian the corresponding numbers are 31 and 22.This yields a much more skewed picture than the MC measure (Table 3), and it reveals that more or less to a greater extent than mer eller mindre has many sources other than its cognate.On the other hand, the numbers seem to support the tentative explanation offered above about the difference between the languages in that Norwegian mer eller mindre is more readily used as a correspondence of more or less, than vice versa.
Of the nine occurrences of mer eller mindre in the Norwegian target texts that do not have more or less as their source correspondence (33 -21;Table 4), two have Ø correspondences (sources) and another two have been rewritten to an extent where it is difficult to establish a clear source.In three of the cases synonymous donwtoners are the source, viz.all but, pretty much, and half as in the following example, (21).Gloss: ... and everybody has more or less forgotten how it was before The remaining two instances are examples of the Norwegian pattern verken mer eller mindre and the English sources are exactly as specified and no more and no less.
Turning to the sources of more or less in the Norwegian original texts, we find that more or less corresponds to no fewer than 26 different items, if we discount Ø correspondences.This shows the versatility of more or less as a downtoner in that the translators find that it can be used to translate a large number of Norwegian downtoners and other hedging devices.If we disregard mer eller mindre some of the more frequent sources of more or less are: • noenlunde ('to some degree') • naermest ('practically', lit.: nearest) • omtrent ('approximately') • stort sett ('for the most part', 'by and large') • temmelig ('pretty') It is also worth noticing the several Norwegian expressions that contain speech verbs among the sources: naer sagt (lit.: near spoken), så å si ('so to speak') and praktisk talt ('practically speaking'), and the frequent use of the modifier sånn ('kind of') which further strengthens, or intensifies, the meaning of the downtoner.Sånn is very frequently used with noenlunde, as in (22).In fact, it seems to be the case that noenlunde on its own is restricted to the premodifying use, and that it requires sånn to function as a degree adjunct.Similarly, temmelig seems to be restricted to premodifier use.
(22) Folkene fra flyselskapene ser bare at navnet og bildet i passet stemmer sånn noenlunde.[JoNe2] Airline staff only make sure the name and photo match more or less.
[JoNe2TE] Gloss: The people from the airline companies see only that the name and the picture in the passport match kind of to some degree.
We believe that the reverse MC measure should be employed whenever possible (a bidirectional translation corpus is of course needed), since it adds another dimension to the contrastive analysis.It shows to what extent a linguistic item in one language can have as its source one, two or a multitude of items in the other language.This relationship between target and source should be compared to the inverse relationship, that between source and target, since it will say something about the extent to which the translators have used the full palette of possible correspondences (translations).Even more importantly, it will indicate if such a palette exists at all, which in turn can tell us something about a possible lexical or phrasal gap in one of the languages. 15

Syntactic function and cross-linguistic correspondence
We have seen how more or less, in addition to its core function as a compromiser, can be an approximator conveying a meaning synonymous with almost, and, when modifying negatively loaded adjectives, can have dismissive connotation.This versatility of more or less makes it accessible as a translation of a large number of expressions.
In example (23), for instance, more or less is indeed used to translate nesten ('almost').
(23) -Greit for oss, sa jeg, og rullet stolen over gulvet i resepsjonen, som nesten var tom.[AnHo2] "That's fine," I said, rolling my chair through the lobby, which was more or less empty.[AnHo2TE] Gloss: -Fine by us, said I, and rolled the chair across the floor in the reception, which almost was empty.
The same is the case with the following example, (24), where the Norwegian source has bortimot ('almost', lit.: away-towards). 16 15 Dyvik (1998Dyvik ( , 2004) ) 25) and ( 26), or to be used to modify a verb phrase, as in ( 27).(27) I den grad de hadde forlatt rommene sine, hadde de stort sett holdt sammen eller sittet fordypet i hver sin bok.[AnHo2] Whenever they had left their rooms, they had more or less stuck together , or sat alone buried in a book.[AnHo2TE] Gloss: In the sense that they had left their rooms, they had for the most part kept together ...
We started the discussion of more or less with reference to Quirk et al.'s (1985) claim that downtoners may have two main functions, that of an adverbial and a premodifying adverb, where in the former case its English more or less are in fact originally made up of two or more words.This applies e.g. to both noen/lunde and bort/i/mot (or borti/mot).
function is to modify or tone down a clause element or a complete clause or utterance (proposition).
With regard to the distribution of these functions, it can be observed that English favours the adverbial function, while the two functions are more evenly distributed among the Norwegian occurrences in the source data.In the cases where the two are translated by each other, we can observe in Table 6 that the preferred function of the downtoners in the original texts is reflected in the translations.This means that both the English and Norwegian translations distributionally follow the tendency noted for more or less and mer eller mindre in original text, i.e.English more or less is typically found as a translation of mer eller mindre in adjunct function, while mer eller mindre is most typically used as a translation of more or less in premodifying function.More specifically, when more or less is used as a premodifier, mer eller mindre is chosen in 70% of those cases, while when more or less is used as a degree adjunct, mer eller mindre is chosen in only 48% of those cases.The distribution of degree adjunct vs. premodifier uses reported in Tables 5 and 6 suggests one additional reason for the unexpectedly low MC of the downtoners.In other words, frequency of preferred syntactic function seems to play a role along with the wider semantic repertoire of more or less discussed above.
This distribution of uses is also substantiated in the 70 English translations where more or less is found as a degree adjunct; the picture is less clear in the 31 Norwegian translations while mer eller mindre seems to be subject to a translation effect, having adjunct more or less as its source in many of the cases.

6
Conclusion Although the two items under discussion seem to have followed the same paths of development in that they can occur in the same position in the clause, take on the same syntactic functions and have the same meanings available, this does not relate the whole story.What we have shown in the contrastive analysis is that: • More or less is more frequent in our data than mer eller mindre; • More or less has a greater number of sources than mer eller mindre; • More or less has a wider semantic repertoire than its Norwegian counterpart; • More or less prefers adjunct use, while mer eller mindre (marginally) prefers the premodifying use; • The 'overuse' of more or less in the English translations may point to a common strategy among the English translators to go for a "safe translation", disregarding potential meaning nuances in the various Norwegian sources; • Translations are good indicators of the (intended) meaning of polysemous patterns, such as more or less/ mer eller mindre; • When preceded by nothing in English and (h)verken in Norwegian more or less and mer eller mindre do not function as approximators or compromisers, but quite the opposite, and can be paraphrased by no more (and) no less and exactly; • When studying the many facets of degree adjuncts, multi-word expressions or phrases should also form part of the study, as they carry similar meanings and are used in similar co-texts to the more traditionally researched single-word adjuncts.
This list serves to illustrate that the method of using a bidirectional corpus of original and translated texts may play a crucial role in teasing out a (cross-linguistic) semantic web that is not easily detectable when looking at monolingual data only.As pointed out by Aijmer (2007: 118): The contrastive or translation method has the effect to sharpen the description of the polysemy or multifunctionality of the adverbs and to show how they are related to each other in terms of shared or different meanings.
In a similar vein, we have shown that more or less, in particular, shares meaning and function with a range of Norwegian items.What emerges is a complex semantic field of downtoners.Altenberg's (1999) MC measure and the rMC introduced here add further evidence to the cross-linguistic complexity revealed by bidirectional translation data.It seems obvious that more research is needed to investigate how single-word and multi-word modifiers and degree adjuncts operate, e.g. to what extent they have the same distribution and function in the clause and indeed how they operate in different text-types and genres.Furthermore, and as we have hinted at for Norwegian at least, there seems to have been a diachronic development involving phonetic reduction where degree adjuncts originally written as two or more words, are now written as one word, e.g.noenlunde ('just about'), omtrent ('approximately') bortimot ('well-nigh').Noenlunde is a good example in this respect, as it seems as if speakers of Norwegian now feel that this word needs to be further modified by sånn ('kind of') and så ('so'), to be able to function felicitously as a degree adjunct.
It would also be of great interest to portray a larger portion of the semantic field revolving around more or less, by for instance taking some of the most frequent (multi-word) correspondences as starting points.For instance, what would a contrastive study of stort sett ('for the most part', 'by and large') or så å si ('so to speak') have yielded?

( 8 )
That was er more or less the compartment that I worked in.(BNC: K7G 87) (9) It means that public awareness has been raised and that professionals are (more or less) willing to entertain the possibility of its occurrence.(BNC: CFE 461) (10) Most people know when they are hungry and will eat, more or less, accordingly.(BNC: CEE 1270) (11) As a result of Theodora's attentions she was now able to decipher it more or less.(BNC: H8B 305) Another use, also mentioned by Metsä-Ketelä, is when the speaker does not want to commit fully to what has been claimed or maintained either by the speaker him-/herself or by others.(12) As far as the clothing is concerned, it's all in the reports, but to save you time I can tell you more or less what we found and it isn't a great deal.(BNC: GW3 1971) (13) ..., and when stripped of all complications this was more or less what Bonar Law intended all along.(BNC: EW1 1442)

( 17 )
Klaerne virker mer eller mindre slengt på ham.[AnHo1N] It looks like his clothes were just thrown on.[AnHo1TE] Gloss: The clothes seem more or less thrown on him.(18) The articles were hardly worth the trouble, but the thought of my daughter spending untold hours typing up all those ancient pieces of mine -for posterity, as she put it -more or less undid me, and I didn't know what to say.[PaAu1E] Artiklene var knapt bryet verdt, men tanken på at min datter hadde tilbrakt utallige timer med å skrive alle disse eldgamle innleggene mine inn på PC-en -for ettertiden, som hun uttrykte det -slo meg nesten ut, og jeg visste ikke hva jeg skulle si.[PaAu1TN] Gloss: ... knocked me almost out, ... Finally, (19) and (20) illustrate cases where the original downtoner has been left out of the translation altogether, and where the compromising, or hedging, function encoded by the adverbs in the original is lost in the translation.
11It reflects "the speaker's willingness to reduce his or her commitment to the propositional content of the utterance"(Metsä-Ketelä 2006: 130).A further factor that plays a major role is the meaning, or effect, of what the toned down item brings into the equation, i.e. the phraseological role of the whole unit including the downtoner.The reason why more or less is sometimes seen as synonymous with almost, sometimes with approximately, sometimes has a dismissive connotation, etc. is the item following or preceding more or less.
12This is clearly illustrated by intact in example (6) and by worthless in example (14).Finally, when more or less is negated, it becomes a maximiser, a function that is very frequent in Norwegian with mer eller mindre being preceded by (h)verken.13

Table 1
Distribution of more or less and mer eller mindre in the ENPC+ (raw figures and frequency per million words)

Table 2
Distribution per million words of more or less and mer eller

Table 3
Of the 41 instances of more or less in the English source texts, 22 are translated by mer eller mindre, while 9 of the 21 instances of mer eller mindre in the Norwegian source text were translated by more or less.

Table 4
explores word semantics in a similar fashion by going back and forth between source and target texts, and creating, what he calls, semantic mirrors based on bidirectional translation corpus data.With regard to the Norwegian sources of more or less, it seems as if other expressions, e.g.(sånn) noenlunde, which occurs 15 times as a source of more or less and stort sett, which occurs 40 times in the original Norwegian part of the corpus (compared to 21 occurrences of mer eller mindre), compete with mer eller mindre in its two main (syntactic) functions as premodifier and as a degree adjunct.Stort sett, for example, can be found to constitute a more or less (!) complete turn in a conversation, as in (

Table 5
Distribution according to function in the original texts