Is "den forste att gå" den förste som gick?

Like all other European languages and a good many others Swedish has been exposed to the linguistic influence of English for a long time. T h e first wave of loanwords reached Swedish as early as the 17 t h century, bringing a trickle of different loans, such as dogg [dog], tobak [tobacco], puritan and kväkare [Quaker]1. The following centuries saw an accelerating influx of English loans, most of them linked to the introduction of new goods and techniques, for example words denoting new means of transportation such as skonare [schooner], kutter [cutter], brigg [brig], terms for new types of food like potatis [potatoes], rostbiff [roast beef] and käx [cakes], for new trends in fashion, for instance schal [shawl], keps [cap], väst [waistcoat], and new developments in sports, among them finish, hockey, match. By the end of the 19 t h century, the trickle of loans had developed into a steady stream, prompting the early 20 t h century linguist Nils Bergsten to remark in 1915 that "Som ett allmänt intryck kan fastslås att det engelska inflytandet sannolikt ännu ej på länge nått sin kulmen" [It may be stated as a general impression that, in all probability, the English influence will not have reached its peak for a long time yet']. (Bergsten 1915). Today, when the process that started as a trickle has developed into a virtual torrent, it is easy to see that his remark was highly prophetic.

Like all other European languages -and a good many others -Swedish has been exposed to the linguistic influence of English for a long time.The first wave of loanwords reached Swedish as early as the 17 th century, bringing a trickle of different loans, such as dogg [dog], tobak [tobacco], puritan and kväkare [Quaker]1 .The following centuries saw an accelerating influx of English loans, most of them linked to the introduction of new goods and techniques, for example words denoting new means of transportation such as skonare [schooner], kutter [cutter], brigg [brig], terms for new types of food like potatis [potatoes], rostbiff [roast beef] and käx [cakes], for new trends in fashion, for instance schal [shawl], keps [cap], väst [waistcoat], and new developments in sports, among them finish, hockey, match.By the end of the 19 th century, the trickle of loans had developed into a steady stream, prompting the early 20 th century linguist Nils Bergsten to remark -in 1915 -that "Som ett allmänt intryck kan fastslås att det engelska inflytandet sannolikt ännu ej på länge nått sin kulmen" [It may be stated as a general impression that, in all probability, the English influence will not have reached its peak for a long time yet'].(Bergsten 1915).Today, when the process that started as a trickle has developed into a virtual torrent, it is easy to see that his remark was highly prophetic.
As the above examples indicate, some of the loans were adapted to Swedish spelling and pronunciation, while others were taken over more or less wholesale from English.What they all have in common is the fact that they are direct loans, i.e. they are all instances of lexical borrowing from the foreign language.
Most studies of borrowing -especially those taking a historical perspective -focus on direct loans.This emphasis on direct lexical borrowing is understandable from several points of view, not least because such loans are highly visible and offer incontrovertible evidence of their foreign extraction.Not all foreign linguistic influence is direct, however: studies of linguistic borrowing have long recognised the existence of translation loans (also known as "caiques"), which usually involve literal translation of compounds and phrases in the foreign language, and semantic loans in which the meaning of a native word is extended in accordance with usage in the donor language, as when Swedish köpa [buy] can also be used in a phrase like köpa ett argument Unlike what is the case with direct loans, the evidence for the loan status of caiques and semantic loans is always inferential and dependent on the interpreter's knowledge of the donor language.When for example a Swedish journalist recently wrote (about golf star Annika Sörenstam) "Om Annika spelar bra öppnar hon en burk med mask" (Expressen May 22, 2003, p. 4), there is a strong likelihood that he was translating2 the American English expression open a can of worms ("create a complex and difficult problem") literally into Swedish öppna en burk mask, an expression that must have struck many readers as extremely odd.
However, a strong likelihood that something is a loan is not the same as evidence that it is, and it is obvious that in (purported) semantic loans and caiques, the link between the Swedish expression and its alleged foreign original is much more tenuous than in the case of direct loans.
That link is even more tenuous in another category of loans, i.e. the category known as construction loans or structural loans.In these, a syntactic or morphological pattern traditionally used to express a certain meaning in the borrowing language is assumed to have been replaced by the pattern preferred by the donor language.In a Swedish construction like Den största svenska segern någonsin [the greatest Swedish victory ever], for example, the placement of någonsin represents a fairly recent development in the language, a development, furthermore, that is arguably an instance of construction or structural borrowing from English.
But structural borrowing may also involve what has been called "the preferred way of saying things" , i.e. the default choice of realisation of a linguistic variable.The aim of the present paper is to examine a Swedish construction in which English seems to have affected Swedish usage in such a variable.The variable in question is the choice between attinfinitives and finite relative clauses with som [who/that] as postmodifiers in NPs involving one of the ordinals förstal förste [first].Traditionally the preferred -and indeed the only recognised -postmodifier in such NPs has been and still is a finite relative clause introduced by som, as in (1) and (2) However, there appears to be a fairly widespread feeling among native Swedish speakers that in recent Swedish usage the relative clause postmodifier in constructions like (1) and ( 2) is increasingly being replaced by postmodifiers using infinitival /^-constructions as in (3) and (4).
[The Englishman Adrian M. was the first to set a European record in Strasbourg] (DN87).
4. Hon blev den första svenska ryttarinnan att vinna en tävling ... (DN87) [She became the first Swedish woman rider to win a competition] Many speakers of Swedish deplore this development and feel that the use of NPs with infinitival postmodifiers in sentences like (3) and (4) goes against the norms of Swedish, and that a better way of expressing the meanings of these sentences would be to use NPs postmodified by finite relative clauses as in (1) and (2)5 .
Against this background, the aim of the present paper can now be made more precise: it is to test the hypothesis that the relative clause postmodifiers in sentences like (1) and (2) are increasingly being replaced by AE#-infintives as in (3) and (4).However, before we address this question, it should be pointed out that this suspected ongoing change in Swedish affects not only constructions with första/förste, but in fact all postmodified NPs in which the head is premodified by a superlative or an ordinal, as for instance in the recently attested examples (5) and ( 6 For practical reasons, however, the present study is concerned only with constructions involving infinitivally postmodified NPs beginning with one of the combinations den förste, den första, de första, det första, either on their own or followed by one or several nouns serving as head(s) of the phrase.In the remainder of the paper I will refer to such NPs as "förste/första-NPs".

The English model
In order for construction borrowing to take place, there has to be a foreign construction serving as a model for the change suspected to be going on in the receiving language.In the case of the postmodifying Swedish attconstructions, we don't have to look far afield to find the construction cast as the villain of the piece: it is obviously the class of English complex NPs with postmodifying ^-infinitive clauses in which the antecedent corresponds to the subject of the infinitive, a construction described in e.g.As both Quirk et al and Kjellmer note, the postmodifying to-infinitives in such NPs correspond to and may be paraphrased as finite relative clauses; Kjellmer suggests that they be called "relative infinitives", a usage that will be adopted here.Formally, the NPs involved in such constructions may be divided into two subcategories.In one of these, first itself serves as the head of the NP -for example in (8).In the other, first is the adjectival modifier of the head of the NP, as in (9) below: 8. Harold Wilson was the first to reply -by return post.(BNC A03 997) 9. I cherished the ambition to be the first (perhaps the only) Italian writer to describe the Yiddish world (BNC A05 1451).
In the rest of this paper I will refer to NPs like that in (8) as "simple first-NPs" and to those like that in (9) as "complex first-NFs".
The relative force of the ^-constructions in ( 8) and ( 9) is easily verified: in both, the infinitives may be replaced by the relative clauses who replied and who described.However, as most grammars intended for teaching point out, the infinitival construction is clearly the preferred one: in the entire 90 million words making up the written part of the BNC, there are 1,349 the first to constructions, but only 5 the first who and 36 the first that constructions.This means that the three constructions occur, respectively, 0.05, 0.41 and 15.46 times per 1 million words.
Semantically, the sentences with infinitive postmodifiers and those with relative clause postmodifiers are identical -both describe the order in which things take place: if Harold Wilson is described as the first person to reply, then, of all the people who replied (to whatever summons), Harold Wilson was literally the first.If he had only been number three among the repliers, he would have been described as the third person to reply, had he been number 15 he would be the fifteenth to reply, etc.
However, not all NPs with to-infinitival postmodifiers are relative and order-describing.In ( 10) -( 12), for example, the to-infinitive cannot be replaced by a relative clause.10.I am the first to admit that it is not that easy. (

The Swedish postmodifying att-som variable
As we saw in the preceding section, there is a stylistic alternative to English relative infinitive constructions like e.g.Harold Wilson was the first to reply, i.e. constructions using a finite relative clause like Harold Wilson was the first who replied.The second type was shown to be rare, but since the possibility of variation exists, postmodification in English y?r#-NPs must be described as a variablethe to-who variable -with two possible realizations: the to-infinitive -which is overwhelmingly the most frequent Clearly in ( 10) and ( 11) the first (person) to admit is not the first (person) who admits, but has an almost performative 7 character: these sentences mean something like "I express my extreme willingness to admit that...".It is more difficult to fall back on a performative interpretation for third person NPs like the local council, but it is still the case that replacing the infinitive with a relative clause would distort the meaning of the sentence.None of the non-relative constructions are about the order in which admitting and recognising takes place.Their meaning is perhaps best characterised as "attitudinal" since their function seems to be to describe a positive attitude or tendency towards the state, feeling or even action denoted by the infinitive.
We can sum up the presentation of the English y?r.tf-constructions discussed above as in the following figure : choice -and the finite relative clause.It was also found that the relative toinfinitive has a non-relative homonymous construction with what was called "attitudinal" meaning.
The same description fits the facts of Swedish: the language has what we may call an att-som variable, i.e. relative postmodification of förste/första-NPs may be realised as either an ^-infinitive or a finite relative clause in som, as shown in examples (1) to (4) in section 1.However, the two languages differ in their choice of preferred realisation of the variable: in Swedish, the preferred realisation of the variable is the relative clause construction with som rather than the alternative realisation with an att-'m£mmve.
The two languages are also alike in that the Swedish att-infinitive sometimes used to realise the att-som variable has a homonymous "double" which is not relative at all, but expresses "attitudinal" meaning.Before going into detail about this I will briefly describe how the att-som variable was defined and how instances of it in the data were identified.
The first step was to make a preliminary list of all instances of the attsom variable used as postmodifiers in förste/första-N¥s in a number of Swedish text corpora from different points in time (the data used will be presented in the following section).The variable was defined as the sum of all cases of firste/firsta-NP postmodifiers consisting of either a relative attinfinitive or a som + relative clause construction in which the relative pronoun is the subject of the relative clause.
The second step of the investigation was to verify the variable status of each example, i.e. to determine in how many cases the constructions were really interchangeable without loss of meaning.Once that has been done, all that remained to do was to calculate the respective proportions of ^-constructions and røwz-constructions in order to obtain a measure of the popularity of each of the two realisations.
In the majority of cases the verification process created no problems.Thus, to offer two examples, it is clear that ( 13) and ( 15) can be successfully reformulated as ( 14) and ( 16) respectively.Is 'den förste att gå' den förste som gick?
16. Alice och hennes man var de första som byggde i kvarteret However, in certain cases the interchangeability of att and som is more difficult to establish.It clearly co-varies with other characteristics of the text, such as the complexity of the predicate in the infinitive/relative clause: thus, for example, the choice of <z#+infinitive rather than rørø+relative clause appears to be somewhat less likely when the predicate contains an auxiliary, in particular if the auxiliary is a modal.There are also differences among the individual modals here: as the examples ( 17 The discussion above focused on difficulties that came up in decisions about the replacing of jøwz-constructions by ^-constructions.As a rule, there are no problems in replacements going in the opposite direction, i.e.Ä«--infinitives can usually be freely converted into sow-constructions.However, just as in English there is one noticeable exception to this, viz. the non-relative -"attitudinal" -infinitives.The Swedish constructions of this kind are identical to the English ones in most respects, compare for instance ( 21) -( 22).21) and ( 22) were identified and excluded from the investigation.Spotting such non-relative ^-constructions is in most instances easy, since they usually contain predicates with "attitudinal" meanings, for example erkänna [admit, acknowledge, recognise], beklaga [regret, deplore], etc.However, sometimes a wider choice of predicates is available, as in ( 23) 8 : 23. Han var den förste att rycka ut till hennes undsättning.
[He was the first to come to her rescue].
Sentences like ( 23) are problematic, since they can be given both a relative and a non-relative ("attitudinal") interpretation, meaning either "He was the first person who came to her rescue" or "He was always willing to come to her rescue (every time she needed it)".On the first reading, ( 23) describes a fact, but on the second it describes an attitude or a characteristic attributed to the subject.In Grünbaum (2003), the author argues that forste-(and siste-) NPs with postmodifying AE#-infinitives like den förste att rycka ut till hennes undsättning have traditionally only been used with non-relative -what I have called "attitudinal" -meaning and insists that in the interests of linguistic clarity they should stay that way.Like many others she feels that there has been an unfortunate increase in the relative use of such NPs in Swedish, and that this is due to the influence from English, a language that admittedly uses the same infinitival construction to express both relative and non-relative meaning.

Data and results
It is now high time to address the question asked at the beginning of the paper, i.e. the question whether there is reason to believe that there has been an increase in the use of relative infinitive constructions in Swedish in recent years.
In order to do that we have to find a number of comparable Swedish texts strung out along the time axis in a reasonably regular manner and compare them with regard to their use of första/förste constructions.Such a comparison is now possible thanks to the efforts of Språkdata at the University of Göteborg, an organization that has assembled a number of Swedish text corpora from different points in time and made them available on the Internet.The main advantage of the above choice of corpora is their positions along the time axis.Press 65 and GP 03 represent the beginning and the end, respectively, of the full time range covered by the Språkdata newspaper corpora.As a result of this, the selection of texts decided on here makes it possible to compare texts over a time span of almost 40 years.In addition, the time intervals separating the first four corpora are of exactly the same length, i.e. 11 years.For the final corpus -GP03 -it was obviously no longer possible to maintain this regularity.However, the above advantages were bought at a price.The price in this case involved the size of the corpora, which varies considerably as Table 1 indicates: While such huge differences in corpus size are obviously deplorable, they do not so far as I can see, invalidate the results of the comparison, since what will be compared is not absolute but relative frequencies.Thus what we will be comparing is the different realisations of the relative att-som variable, i.e. the choice between att and som constructions in relative postmodifiers in NPs of the förste/första type, in cases where both constructions are deemed to be possible.
Let us now turn to the results of the comparison.Table 2 presents the frequencies of occurrence for the relative att and som constructions in the five newspaper corpora (the non-relative AEAf-constructions have been omitted).The column for the ^constructions also contains percentage figures revealing the proportion of the att-som variable represented by the relative ^constructions.Regarding the question whether there has been an increase in the use of relative «»-infinitives, it does look as if the use of att has been gaining ground at the expense of som but only moderately so.Thus with one exception -Press 76 -the proportion of /^-realisations calculated as a percentage of the total number of realisations has gone up from 10.8% in 1965 to 17.6% in GP03.It must be admitted, however, that the actual number of occurrences is low and that the total picture that emerges can hardly be said to support the view that there has been an explosive increase in the use of relative Ä#-infinitives.In addition, a chi square test of the figures in Table 2 reveals that the differences are not statistically significant at the .05level.
In fact, if we relate the number of occurrences to the size of the corpora involved, it looks as if there has been a decrease in the use of the entire att-som variable, i.e. the use of NPs of the type den förste soml att, de första soml att, etc. has actually gone done in the period 1965 to 2003.The total number of att-som realisations in Press 65 was 46.If the ratio between number of instances and corpus size had remained constant, the total number of realisations in Press 98 should have been 9.32 times higher, i.e. roughly 429, since Press 98 is 9.32 times larger than Press 65 (9,239,336 / 990, 989 = 9,323).In order to reach that value, the reported value for Press 98 would have to be multiplied by a factor of 2.6.
The same point can also be made in yet another way i.e. by comparing the number of occurrences per 1 million words across the corpora.This has been done in Table 3.It is probably wise to take the results in Table 3 with a pinch of salt, since in this type of comparison, the small size of Press 65 and Press 76 will tend to skew the outcome of the comparison.However, these results provide a certain amount of further support for the idea that the ^-constructions have been gaining ground at the expense of the jørø-constructions: counting from DN 87 downwards we notice a decline in the som figures, while the att figures remain more or less stable.

Last words
The aim of the research on which the present paper is based was to find out whether there has been an increase in the use of the relative attconstruction in Swedish NPs premodified by första or förste like for example Hon blev den första (svenska författaren) att motta priset.The results of the analysis indicated that there appears to have been a certain increase in the use of relative att in the almost 40 years that have passed since 1965, the date of the earliest newspaper corpus available.
That there has been an increase would have been even clearer if we had concentrated on the "simple" type of firsta/firste-NF i.e. those NPs in which första/första is itself the head of the noun-phrase.This becomes apparent from a study of the results in Table 4, which breaks down the results into separate columns for the simple and the complex NPs.Clearly it is the simple type of förstalförste-NF that is primarily gaining ground at the expense of the røwz-construction.But for all this, there remains a feeling that there is just too little data for us to draw any firm conclusions.It has to be said that it is somewhat disappointing to find such scarcity of examples in what must be regarded as a fairly substantial collection of data.Furthermore this very scarcity of relevant examples calls into question the often reported impression among native speakers that these days, Swedish is positively teeming with relative ^-constructions.One possible explanation for the apparent scarcity of data may of course be that the newspaper corpora used here are simply not representative and that picking another batch of newspaper corpora would yield different and more definitive results.Such an explanation may well be correct.However, there is also another explanation that suggests itself, i.e. that for all reports to the contrary, a rate of occurrence of about 3 times per 1 million words is in fact reasonable for such a construction.
On such a view, the perceived ubiquity claimed for the relative attconstructions by many native speakers of Swedish would have to be explained in terms of the salience of these constructions: it may be that a rate of occurrence of 3 per 1 million words is enough to make the relative ^-constructions seem to be all over the place.
At the same time, there can be no doubt of the presence of the relative ^-construction in present-day Swedish.Thus a Google search of the Internet for the phrases den första att, den förste att, de första att and det första att yielded the following results 9 : ' I wish to thank Solve Ohlander from Göteborg university for suggesting such a search.Obviously these figures are approximations and there are many possible sources of error in this kind of search.10Thus, to mention just one of these, these reported figures clearly also include all instances of non-relative attconstructions like Jag är den förste att erkänna att... which would have to be dropped from the data.But even if the figures in Table 5 should perhaps be reduced by some 25%, the results of the Google search are highly convincing and suggest that by now the relative ^-construction has carved out a place for itself in the Swedish language.

7I
owe this suggestion to Peter Alberg-Jensen of the Slavic department at Stockholm university.106 13. Bli den förste i Ert kvarter att äga sista skriket i Popkonsten!. (Press 65) ["Become the first on your block to own the latest in pop art"] 14.Bli den förste i Ert kvarter som äger sista skriket i Popkonsten! 15.Alice och hennes man var de första att bygga i kvarteret.(GP 03) [Alice and her husband were the first to build a house in their area']

21.
Tyvärr är jag själv den första att betvivla detta.(Press 65) [Sadly, I am the first person to doubt that] 22. Att städarbetet ändå tycks ha klickat ibland är han den förste att erkänna.(DN 1987) [That the cleaning still sometimes appears to have been neglected, he is the first to admit] In the course of what I have called the verification process, sentences like (

From
among the texts available, I chose the following newspaper corpora as the data on which to base my study: Press 65, Press 76, DN 87, Press 98 and GP 03.The first two of these are made up of texts from various Swedish newspapers from the years 1965 and 1976, respectively.DN 87 consists of articles appearing in Sweden's biggest daily Dagens This example is from Grünbaum 2003 Nyheter in 1987, Press 98 contains texts from different Swedish dailies published in 1998, and GP 03 contains the full contents of the Göteborg daily GöteborgsPosten from the first six months of 2003.

Table 1 :
The size of the Swedish newspaper corpora.

Table 2 :
The distribution of postmodifiying relative att and somconstructions in five Swedish newspaper corpora What is perhaps most noticeable about the results in Table2is how few realisations of the att-som variable there are in a total corpus of 23.5 million words.Overall, there are 486 realisations of the variable, which works out at 20.7 realisations per one million words.There are almost exactly three relative att-NPs per 1 million words and roughly 17.7 instances of the relative rørø-construction in the same amount of words.

Table 3 :
Occurrences of relative att and som per 1 million words in the five newspaper corpora

Table 5 :
Results of a Google search for första/förste NPs