Sentence openings in academic economics articles in English and Danish

Scandinavians have always had to use a second language to communicate with their peers. In hard science developments have long been international and it has been important for results to reach an international community of academic peers. In the last fifty years globalisation and professionalisation have affected many other academic fields (such as medicine and economics) in such a way that publication directed to a local community, including those involved in practice, has been downgraded relative to those which reach an international academic community (Kærgård 1998, Petersen and Shaw 2002) . Simultaneously academic leadership has passed to the US along with leadership in other fields. The consequence is that all over the world academics feel pressure to write in English, as the language of the US , or be ignored.

Scandinavians have always had to use a second language to communicate with their peers.In hard science developments have long been international and it has been important for results to reach an international community of academic peers.In the last fifty years globalisation and professionalisation have affected many other academic fields (such as medicine and economics) in such a way that publication directed to a local community, including those involved in practice, has been downgraded relative to those which reach an international academic community (Kaergård 1998, Petersen andShaw 2002).Simultaneously academic leadership has passed to the US along with leadership in other fields.The consequence is that all over the world academics feel pressure to write in English, as the language of the US, or be ignored.This pressure causes much less anguish in Scandinavia than in countries where it has traditionally been possible to write in the national language and be read internationally.This may be because Scandinavian scholars have always had to use another language or because English is typologically close to Swedish, Danish, and Norwegian.In any case, the evidence of international comparisons is that the language is no handicap.Sivertsen and Aksnes (2000), for example, quote figures that show Finland, Denmark, Sweden and the Netherlands ranking with the US at the top of the tables for publication "impact" in the natural and social sciences as measured by citation (RCI), relative to expenditure.By contrast, for example, Japan spends a higher proportion of its GNP than the US on research but achieves a much lower impact.Other factors 2 The Danish expression "valby-engelsk" is composed of the name Valby -part of Copenhagen -and the word for English.It means "naive danicised English" and is analogous to Chaucer's "Srratford-ar-Bow French" for provincial Anglo-Norman.
clearly weigh heavily, but ease of access to the language and rhetoric of the US, as the research community by whom it is crucial to be cited, is likely to be among those that influence such figures.

Scandinavian English?
Texts in one language influenced by another arise in various ways.Sometimes the influence comes from a specific text, as in translations where a number of studies have shown that the frequency of particular mother-tongue features is influenced by their frequency in the source text (and by implication the source language) (Altenberg 1998, Hasselgård 1997).Sometimes the whole code of the recipient language is influenced by that of the source language, as when vernacular languages have constructions originally calqued on academic Latin, or when usages appear in Scandinavian languages which are based on English (such as construction loans like Swedish han är en läkare for han är läkare "he is a doctor"(Ljung 1985: 80), where the indefinite article is ascribed to English).Sometimes, finally, writing or speech in a second language can be expected to show, along with developmental features, transfer of the first language and source culture.This could be called learner language or interlanguage, although this article looks at skilled writing for which the terms are not appropriate.
In a survey (Petersen & Shaw 2002) of one group of bilingual academics -Danish applied economists -respondents disagreed as to whether academic economics articles in English written by Danes differed consistendy from their equivalents written by people from the US, Britain, Australia, etc ("Anglos").About half the 82 questionnaire respondents thought they did not, 40% thought they did, and 10% "didn't know".Suggested reasons for there being no difference were: that internationalisation has meant that "Anglo rhetoric/style" is not meaningful; that the processes of reviewing and editing erase differences; that the field uses mathematics and verbal formulae; and that all participants in the field are immersed in the same style and register of English.Suggested reasons why there might be a difference were: that that limited language proficiency leads to ineffective "style and flow", or as one respondent put it "man ser ofte dålige artikler formuleret på ubehjaelpeligt "valby-engelsk"" ("you often see bad articles formulated in clumsy Danish-English") 2 ; that Danes have a different style or rhetoric due to a national habit of thought or expression; and, more generally, that there are differences in disciplinary cultures among nations.The notion that different languages imply different audiences also arose, in discussions of the purpose of writing in Danish or English.We could summarise these points on three levels: articles in English by Scandinavians might be structured differently or use different patterns of rhetoric, there might be non-standard usages in Scandinavian English, and finally such writing might exhibit a different or reduced range of vocabulary and sentence structure.
On the first level, the evidence suggests (Melander et al 1997, Shaw 2003) that there are characteristic Scandinavian rhetorical features in much Scandinavian academic writing in English, but that they are within the acceptable range of variation of writing in English across the globe and so go unnoticed in international publications.On the second, there are studies of the typical errors of Scandinavian writers (for Swedish, for example, Köhlmyr 1999Köhlmyr , 2003;;Karlsson 2002, Warren 1982, Thagg Fisher 1985) and one can assume that many of the errors that still exist in academic articles are indeed ironed out by correction services, reviewers, editors, etc. (though this standardisation process does not in fact erase all idiosyncrasies).On the third level, it seems likely that much of the individual flavour of varieties of English derives not from absolute differences from other varieties, but from relative ones -a greater or lesser tendency to use certain types of construction for example (Leitner 1994).A sense of indefinable idiosyncrasy which might be ascribed to rhetoric or incorrect English might actually be due to such relative differences.Code features of Swedish learners' writing in English which differ in frequency from those found in the English written by learners from other mother tongue backgrounds have been investigated in learner corpora such as SWICLE (the Swedish component of the International Corpus of Learner English).Boström-Aronsson (2001), for example, found that clefts and pseudo-clefts were more frequent in Swedish advanced student writing than in comparable English native-speaker writing (see also Eriksson 2001, Hagglund 2001).Increasingly such investigations are done with reference not only to L2 native texts (English written by native speakers in this case) but also to LI native texts (Swedish-language texts in this case: Borin and Priitz 2003, Altenberg 1998, Altenberg and Tapper 1998).The texts examined so far have mainly been written by learners.Mature writing in a foreign language, such as the English written by Scandinavian academics, has not been much investigated at the level of syntax or vocabulary, and it would be useful to know whether it shows analogous transfer features.

Cross-linguistic influences on thematisation
One feature which has been investigated on the basis of translations is thematisation.Hasselgård (1997) has examined sentence openings in original texts in English and Norwegian and their translations into the other language, and Altenberg (1998) has looked at similar features in translations from English into Swedish and vice versa.
It seems to be agreed that in English the first element in the sentence is more often the subject than in Scandinavian languages.In Hasselgård's fiction texts from the English-Norwegian Parallel Corpus 66% of the sentences in the original (i.e.not translated) English sample began with the subject.However, only 55.3% of the sentences in Hasselgård's original Norwegian sample began with the subject.Translations tend to maintain the proportions of the source language.In Hasselgård's Norwegian translations of the English fiction texts 64.2% of sentences had the subject as initial element, while in her English translations of the Norwegian fiction texts the figure was only 57.3%.Around ten percent of sentences in both languages began with objects; in English such sentences were confined to direct speech, while in Norwegian they also occurred in narration.Gosden (1992Gosden ( , 1993)), looking at published hard-science research articles in English by native speakers of English, found that some 67% of sentences began with the subject.He furthermore found that virtually all the sentences in his sample that did not begin with the subject had initial adverbials.Altenberg (1998) reports "a number of studies" of Swedish (not further specified) with results in the following ranges: between 60 and 75% of onsets are subjects, between 20 and 30% are adverbials, from 2 to 14% are objects, and between 0 and 2% are complements.Proportions depend on text type.For example, in Swedish fronted adverbials are more common in formal writing than in informal speech.
Adverbials which can be fronted fall into several classes, including stance adverbials (disjuncts) like frankly, or from this point of view, circumstance adverbials (adjuncts) like in 1956 or outside Stockholm, and linking adverbials (conjuncts) like however and on the other hand (Biber et al 1999).There has been quite a lot of research into the placement of linking adverbials, or conjuncts, with varying definitions of the class.Altenberg (1998) used a broad definition which includes at least and of course as well as however, by contrast, thus, etc He found that a higher proportion of the conjuncts in his generically mixed sample of English texts were sentence-initial than in Swedish (74% in English, 37% in Swedish), so that initial conjuncts in English are often translated by non-initial ones in Swedish and vice-versa.The proportion of connectors that are sentence-initial is, however, dependent on genre.Biber et al (1999:891) found that in English academic writing some 50% of linking adverbials were initial, with 40% medial and 10% final.Holding genre constant, frequency also varies with writer skill.Altenberg and Tapper (1998) found that advanced Swedish students writing argumentative essays in English had three times as many conjuncts clause-initially as clause-medially.This contrasts both with Swedish students writing Swedish, who placed nearly twice as many conjuncts medially as initially, and with English students writing English, showing a reversal of the Swedish pattern of placement frequency, with twice as many initial conjuncts as medial.The high proportion of initial conjuncts in the Swedish students' writing in English suggests that there is something other than simple transfer taking place and the placement of conjuncts in English reflects writer skill.This is supported by my comparison (Shaw 2001) of native Englishspeaker student essays in English literature with published essays.The students had roughly equal numbers of initial and non-initial conjuncts, while the professional academics had more than twice as many in non-initial position.
These differences in occurrences in certain positions have to be distinguished from absolute differences in the number of adverbials in general or of certain types in the texts.Botin and Priitz (2003) found that both original Swedish texts and the English of Swedish students included proportionally more adverbs than original English texts.In particular the Swedish originals included about 50% more conjuncts than the English ones.However their corpora (part of the BNC English-language corpus, the Uppsala Student English corpus, and the Stockholm Umeå Corpus of Swedish) are not closely comparable.
The observation that different languages appear to be characterised by different frequencies of conjunct use can be related to a discussion about overuse and underuse of such adverbials in learner writing.Granger and Tyson (1996) found overuse by francophone students of connectors like actually and as a matter of fact.Similarly Bolton et al (2002) found an overuse of logical connectors by Hong Kong Chinese university students relative to professional academics writing in English in a study based on the ICE Corpus of Hong Kong English (confirming the observation of Crewe (1990) that Hong Kong students overuse and misuse a number of conjuncts relative to NSE writers), but noted that British students also tend to overuse connectors relative to published academic writing.Altenberg and Tapper (1998) found underuse by Swedish-speaking students writing English, compared to their British peers, of some logical connectors, but the significance of this is obscure if British students overuse connectors relative to skilled writers.Granger and Tyson (1996) think that most of the differences they found are due to transfer of collocations and rhetoric from French -the frequency of en effet in French might lead to frequent indeed in English, for example.However, Milton and Tsang (1993), who looked at Hong Kong students, ascribe over-use to over-teaching of conjuncts.Altenberg and Tapper (1998) suggest both interlingual and training factors.The evidence from native-speaker student and professional use above suggests developmental factors as a third possibility.

Aim
It is the aim of this study to look at an aspect of thematisation in published academic English written by Danes and assess how far there are frequency differences in the use of theme, which might make the code they use subdy different from that in writing for an international audience without making it in any way grammatically incorrect or rhetorically different.The type of analysis has already been used on learner texts but has not earlier been applied to texts written by professional writers.The basic research questions are very simple: • What elements occur first in sentences in articles on economics in Danish by Danes, in English by Danes, and in English by Anglos?
• What proportion of linking adverbials are placed in the initial position of the sentence in each of these sets of articles?
Thus the first question refers to all initial elements, and asks what categories they belong to, while the second refers to a category and asks what proportion of its members are placed initially.

Material and method
The texts examined are thirty published articles in economics.Ten texts (called set DD) are in Danish written by Danes who also publish in English, ten texts (called set DE) in English written by the same set of Danish writers, and ten texts (called set EE) in English written by what I will call Anglo writers.These are researchers based in institutions in the US, Britain, or New Zealand (as it happens) and at least one member of each authorial team has an Anglo-Saxon name.Whether or not the third set constitutes "native-speaker" writing it is probably not characterised by a specific non-English substrate.
The Danish texts are from the Nationaløkonomisk Tidsskrift which ranks 152 by weighted citations in Kalaitzidakis et al (2001).Of the English-language Journals mentioned here only Scandinavian Journal of Economics (27) and Oxford Economic Papers (53) rank higher.
The corpus was constructed by finding papers in the Nationaløkonomisk Tidsskrift Wimen as far as possible by scholars who had been interviewed for the survey mentioned in Section 1.1 (Petersen & Shaw 2002), checking bibliographies for publications in English by the same authors, and then checking the journals in which these had appeared (i.e.journals based in English-speaking countries).Consequently all articles have comparable topics and disciplinary orientation, and broadly similar expectations of reader expertise.DD and DE have similar author backgrounds and differ in language and audience.DE and EE have the same language and audiences and differ in author background.
Table 1 shows the journals from which the articles came.It also shows that many articles had multiple authorship and so the paired English-language and Danish-language articles do not always have exactly the same authors.Nevertheless the authors of DE and DD come from the same "discourse community".Separately from the count of all initial elements, all sentence-level conjuncts anywhere in the samples were also counted.Sentence-level conjuncts are those which show the relation between two orthographic sentences rather than between two clauses in a sentence.Thus in He worked hard and therefore passed the exam, therefore is not a sentence-level conjunct What I mean by this may be further illustrated by two examples, with the initial elements highlighted and labeled: This [subject] enables the researcher to test for the non-neutrality of technical change by examining the sign (and significance) of the coefficient on the technology variable, or any other "external" factor.Under the null hypothesis of "neutrality," [stance adverbial] the coefficients would be zero.The mean and standard deviation for each group of ten articles for each parameter was calculated and t-tests were carried out to obtain an indication of which differences were likely to be significant statistically.4

Results
Table 2 shows the average number of sentences in the first hundred sentences of each article which began with subjects, adverbials, or "other" elements.The "others" are all question-words.2 shows that there are significantly more sentences that begin with subjects in the Anglo articles, and correspondingly significantly fewer that begin with adverbials than in the articles by Danish writers.
There is no significant difference between the Danish articles in either language.Moreover there are no significant differences in the small numbers of sentences which begin in other ways.The articles in English by Danish writers are much more like the articles in Danish by these writers than they are like articles in English by Anglo writers.
Another finding shown in Table 2 is that the EE articles were much more different from one another in terms of proportions of particular initial elements than the articles by Danes.It can be seen that the standard deviations of the means were much greater for the EE articles than for DE or DD.Thus, the range in EE was so wide that the article with the lowest number of sentences with initial subjects had 50, while the highest had 84, while for the articles by Danish authors the range was much narrower: the corresponding figures are 45 and 58 for DE, and 45 and 65 for DD, for DD 45 and 65.Nevertheless Table 2 shows there are significantly more sentences that begin with subjects in the Anglo articles, and correspondingly significantly fewer that begin with adverbials than in the articles by Danish writers, and there is no significant difference between the Danish articles in either language There are no significant differences in the small numbers of sentences which begin in other ways.The articles in English by Danish writers are much more like the articles in Danish by these writers than they are like articles in English by Anglo writers.
Table 3 shows the average numbers of initial adverbials of various types per hundred sentences, and also the average number of linking adverbials in all positions.As we would expect, most of the significant differences in averages here are between the two languages.Danishlanguage texts have significantly more sentence-initial adverbial clauses, and more non-sentence-initial linking adverbials (and more linking adverbials overall).There is, however, one highly significant score for a difference on the other dimension: there are considerably more initial circumstance adverbials in the articles by Danish authors in either language than in those by Anglo ones.3 are based on the raw scores, and there are fewer initial adverbials overall in the Anglo texts, so it is no surprise if there are fewer initial adverbials of any given type in these texts.Another way to look at these data is in terms of the percentage of initial adverbials of the various classes, as in Figure 1.In these proportional terms, the difference in initial adverbial clauses vanishes, but there are still relatively fewer initial circumstance adverbials in the Anglo sample.

initial adverbials in various categories
Table 3 also shows that a significantly higher proportion of linking adverbials are placed non-initially in Danish than in English.As Figure 2 illustrates, the DE sample actually had an even lower proportion in this position than the EE one.

Discussion
The Anglo economists in this sample placed a very high proportion of their linking adverbials sentence-initially, unlike the literary scholars described in Shaw (2001).This reminds us that the features examined here depend very much on genre and topic.The findings probably also depend on the exact definition of linking adverbials.The only linking adverbials counted here were those that linked sentences, whereas other studies have included adverbials linking clauses within sentences, and, like Altenberg and Tapper (1998), have classified stance adverbials like of course as conjuncts.A further proviso arises from the finding that the standard deviations were higher for the Anglo sample than for the two Danish samples.The writers of the Danish samples are all members of a single smallish national academic community, and probably all know each other personally while the Anglo writers come from different countries and represent a larger and much more diffuse collection of scholars.Hence it is not surprising that the Danish texts are more homogeneous, and it is possible that we are looking at the idiosyncratic usage of a particular group of Danes, rather than anything that can be generalised.A final warning should be that in the absence of Danish corpus studies it has been tacitly assumed that work on Swedish was relevant to Danish.
Nevertheless, the findings comparing original Danish with English written by Anglos confirm those of previous workers on English and Scandinavian languages in comparison with English.Sentences in Danish by Danes are more likely to begin with an element other than the subject than sentences in English by Anglos.In academic writing this element is likely to be an adverbial.More sentences in the Danish articles than in the English begin with a circumstance adverbial.Furthermore a higher proportion of all linking adverbials are initial and a higher proportion of all initial adverbials are linking adverbials in English than in Danish, although Danish has more linking adverbials overall.
The data for the English written by Danes are somewhat surprising.Even though the English material is grammatical, edited, and completely unexceptionable, it is more like Danish than like English written by Anglos on two dimensions: the proportions of sentences which begin with subjects, and numbers or proportions of adverbial introductions which are circumstances/adjuncts.This means that the English spontaneously written by experienced professional academics has some of the characteristics of the grammatical and completely unexceptionable translationese written by experienced professional translators.
It would be ridiculous to treat these Danish academics as if they were learners of English, given that they can obviously write articles publishable in journals which in some cases are of very high status, a level of communicative proficiency higher than that of a large majority of native speakers of English.A "World Englishes" (Kachru 1983) perspective would say that what has been investigated here is Danish Standard English, a perfecdy respectable variety with its own characteristics, not to be confused with "valby-engelsk" (which would be Danish Non-Standard English presumably).
Two arguments could be given for such a position.One is that the articles in English examined here have indeed emerged from a Danish academic community who both read English and use English in their professional writing.For example conference papers, obviously drawing largely on international literature in English, will be written and discussed, often largely by Danes or other Scandinavians, in English, converted into working papers in English, discussed internally in Danish and only then submitted to English-language journals.They will thus have had a long life in a community using "Danish Standard English" before they emerge into an "International English" forum, and one can reasonably speak of a Danish speech (or writing) community using English in some registers.The other argument would be that the kinds of reasonably consistent frequency differences noted here also characterise bona-fide variety differences in English.Thus Tottie (2002) found that one feature which contributes to the Britishness of British English and the Americanness of American English is the relative frequency of perhaps and maybe, and Leitner (1994) found, for example, that start is more frequent than begin in fiction in the British LOB corpus, while in the US Brown and Indian Kolhapur corpora begin predominates.By analogy one could claim that it was a characteristic of Danish English that circumstance adverbials were more often placed initially than in British English.This would imply that Danes would want their own "national" variety of English to express their identity, as Singaporeans are said to express their identity through Singapore English (Tongue 1974).It is my impression, however, that the Danish respondents to our questionnaire were completely exonormative: Danish was their language and they had no need for an English of their own.Their aim was to write the foreign language as closely as possible to native norms.If we accept this point of view then we have to look at the data from a second-language-acquisition perspective.The question is whether the differences from Anglo frequencies are due to transfer from the first language or are a developmental feature, or are in fact a combination.The Danish writers said in the survey that they had never had any instruction in academic writing in English or Danish, so the third possibility, over-use of a feature due to over-teaching, is excluded.
If the Danes are not placing adverbials in this way because it is a norm of their local variety (since they do not see themselves as having a local variety), why are they doing it?The most obvious explanation is that they are transferring norms for fronting adverbials from the first language to the second, giving them a slight "foreign accent" in their writing.The transfer explanation is supported by the finding that it is exactly the circumstance adverbials that are more frequent initially in both DE and DD.However, it appears that some things are transferred and others are not.Specifically, although initial placement of circumstance adverbials in particular is transferred from writing in Danish to writing in English, noninitial placement of linking adverbials is not.In fact Table 3 shows that the linking adverbials which are non-initial are an even lower proportion of all linking adverbs in DE than in EE (as in Altenberg and Tapper's Swedish case).
Why should the tendency to place adjuncts initially be transferred while there is not a similar tendency to place conjuncts non-initially?The reason may not be simple transfer of norms but a general tendency to place adverbials peripherally in order to avoid cognitive overload.Second language acquisition studies show that early learners of a variety of languages tend to place negative markers and adverb(ial)s peripherally (outside the subject-verb-object core) in the second language (no he do it) even when the marker is placed centrally in the clause in both the first language and the one being acquired (Johnston 1985, cited in Ellis 1994).Perhaps it is also "easier" to place adverbials peripherally.Comparison of student essays and published papers (Shaw 2001) suggests that linking adverbials placed other than initially are more frequent in more skilled native-speaker English, suggesting that it is easier to place them initially for native speakers as well.It seems possible that this is a type of developmentally-guided transfer: the tendencies which are simpler to transfer are transferred, those which are more difficult are not transferred.
But we must avoid treating sophisticated multiply-published writers as learners for whom adverbial placement causes cognitive overload.These usages are more likely to be fossilized remnants from an earlier developmental stage.The writers, one could argue, have never adapted their style in this area to the level at which they are now writing.And in fact it is hard to see how they could have noticed any need to do so; they could not be aware that their average proportion of sentence-initial adverbs is higher than that of Anglos, because they have models of Anglo writing which conforms to their mean frequencies; plenty of Anglo writers (two out of ten in this case) have adverbial-placement proportions close to or even below the "Danish-English" mean.
It is my assumption that Danish writers want English to be a foreign language for them -that their language loyalty is to Danish and they do not want to express Danishness through English.In that case they want to write like Anglos and might like to know that they are likely to seem to have developed more as skilled writers if they place their subjects initially (where they have the choice), so as to approach more closely to the Anglo norm.If they choose not to do that, it seems unlikely that anyone will notice.

Figure
Figure 2: Percentages of linking adverbials which are non-initial

Table 1 :
Articles investigated: journals of publication and authorship Nat 0k with P and one in Oxford Economic Papers with S.The first hundred sentences in each of thirty articles were examined.Sentences are treated as bounded by full stops or semi-colons, so we are looking at sentence-initial, not clause-initial elements.Initial elements are divided first of all into subjects, adverbials, and others.The category "others" would include any objects or complements that occurred (none did, even in Danish), question words, imperative verbs in jussive sentences, and elements in minor sentences of the the more, the merrier type.The adverbials are then divided into finite clauses and other adverbials.
Biber et al (1999)ents an author.For example authors Q and R wrote an article in Nat 0k and another in Labour, and Q co-wrote one article in (Finite clauses were introduced by because, if, when, etc., or by the verb in Danish conditionals analogous to Had this been done) The non-clausal elements are then divided, using the criteria ofBiber et al (1999)into linking adverbials (broadly conjuncts like however), stance adverbials (broadly disjuncts or prefaces like as seen from Table3, or to illustrate this result) and circumstance adverbials (broadly adjuncts, typically space or time adverbials like in the energy demand literature or until the late 1980sbut this is the default class).

Table 2 :
Average number of sentences in the first hundred with certain initial elements

Table 3 :
sentence-initial adverbials and all linking adverbials