46
Informational density: a problem for translation and translation theory 1 CATHRINE FABRICIUS-HANSEN Abstract The present paper is concerned with theoretical and practical aspects of paraphrasing or translating (German) texts showing a relatively high degree of syntactic complexity and informational density into (Norwegian) texts characterized by a less complex, more paratactical style, and vice versa. The theoretical setting is (segmented) discourse representional theory (Kamp and Reyle 1993; Asher 1993), which allows a whole family of informationally equivalent texts to be represented by one discourse repre- sentation structure and thus opens the way for a theoretical explication of the notion of (relative) informational density. It is shown that paraphrasing/translating a "hypotactical" into a "paratactical" text is governed by two principles information splitting and discourse structure fidelity — that are, to a certain degree, in conflict with each other. The more information splitting is done, the more difficult it will be to reconstruct the segmented discourse representational structure (SDKS) of the original text, that is, the overall discourse/text structure in the more traditional sense. Translation from "paratactical" into "hypotactical" texts calls for information collecting instead of information splitting; the main difficulty lies in assigning a SDKS to the text and determining which part of the information given in the text should be syntactically downgraded, and how that should be done. Introduction Viewed from a foreign language like Norwegian, written varieties of modern German often seem rather heavy, loaded with information, difficult to read, and difficult to translate adequately into the mother tongue. In particular, this holds of academic and various other kinds of expository prose, which — so it seems tend to show a much higher Linguistics 34 (1996), 521-565 0024-3949/96/0034-0521 © Walter de Gruyter Brought to you by | University of Oslo in Norwa Authenticated | 129.240.0.83 Download Date | 11/30/13 4:52 PM

Informational density: a problem for translation and translation theory

  • Upload
    uio

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Informational density: a problem fortranslation and translation theory1

CATHRINE FABRICIUS-HANSEN

Abstract

The present paper is concerned with theoretical and practical aspects ofparaphrasing or translating (German) texts showing a relatively high degreeof syntactic complexity and informational density into (Norwegian) textscharacterized by a less complex, more paratactical style, and vice versa.The theoretical setting is (segmented) discourse representional theory(Kamp and Reyle 1993; Asher 1993), which allows a whole family ofinformationally equivalent texts to be represented by one discourse repre-sentation structure and thus opens the way for a theoretical explicationof the notion of (relative) informational density. It is shown thatparaphrasing/translating a "hypotactical" into a "paratactical" text isgoverned by two principles — information splitting and discourse structurefidelity — that are, to a certain degree, in conflict with each other. Themore information splitting is done, the more difficult it will be to reconstructthe segmented discourse representational structure (SDKS) of the originaltext, that is, the overall discourse/text structure in the more traditionalsense. Translation from "paratactical" into "hypotactical" texts calls forinformation collecting instead of information splitting; the main difficultylies in assigning a SDKS to the text and determining which part of theinformation given in the text should be syntactically downgraded, and howthat should be done.

Introduction

Viewed from a foreign language like Norwegian, written varieties ofmodern German often seem rather heavy, loaded with information,difficult to read, and difficult to translate adequately into the mothertongue. In particular, this holds of academic and various other kinds ofexpository prose, which — so it seems — tend to show a much higher

Linguistics 34 (1996), 521-565 0024-3949/96/0034-0521© Walter de Gruyter

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

522 C. Fabricius-Hansen

degree of Informationsdichte — informational density — than correspond-ing texts in, for instance, Norwegian. But what exactly does that mean?And of what kind exactly are the translation problems caused by differ-ences in informational density between source and target language? Theseare the questions the present paper is concerned with. In section 1, Iaddress the first, more theoretical question, trying to define or explainthe notion of informational density rather informally within the frame-work of discourse representation theory (DRT).2 Section 2 illustrates howa DRT-inspired technique may be put to use when translating informa-tionally compact German texts into less compact Norwegian — that is,at least as a means to make the translation problems transparent. Takentogether, the two sections represent a first attempt at exploring theusefulness of DRT for translation theory and practice. In section 3 it isshown ho'w the DRT extension outlined in Asher (1993) can be used toaccount for important aspects of text and text translation beyond theinformational content alone, viz. the so-called rhetorical relations, infor-mation structure, illocutionary structure, and the like. Section 4 containsa brief summary.

1. Informational density and discourse representation theory

1.1. What do we mean by stating that a sentence or a text is information-ally dense? How can we measure its informational density? With a viewto these questions, let us consider sentence (1) and the sequence in ( ).

(1) Frankreich trauert über den Tod eines sehr bekannten französischenSchauspielers.Trance mourns the death of a very famous French actor.'

( ) Ein französischer Schauspieler ist gestorben. Er war sehr bekannt.Frankreich trauert über seinen Tod.

French actor has died. He was very famous. France mournshis death.'

(1) and ( ) are both possible answers to a question like Was ist dennlos? Was ist denn passiert? 'What is going on? What has happened?',which one might utter seeing, for example, flags at half-mast all over,streets crowded with weeping people, etc.; that is, taken as texts ordiscourses they are compatible with the same leading question, the same"quaestio" (Klein and von Stutterheim 1992). And relative to that quaes-tio, they apparently convey the same information, thus being semanticallyequivalent in a certain sense.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 523

Having developed a dynamic notion of meaning, discourse representa-tion theory (DRT) is well equipped to handle this kind of informationequivalence between texts of any length. In DRT, a semantic representa-tion — a so-called discourse representation structure (DRS) — is assignedto the discourse by a specific syntax-driven construction algorithm thatproceeds sentence by sentence: "If j sentences have been processed toyield the DRS Kj [K: 'context of interpretation'], then the processing ofSj + 1 will yield a DRS that combines with Kj to form an extended DRSKj + 1" (Asher 1993: 63). That is, for each sentence being processed, theDRS constructed so far will be updated with the information conveyedby the sentence in question relative to the given context (DRS). A DRSconsists of a set of discourse referents and a set of conditions on thosereferents. It may be viewed as a kind of mental or cognitive representationof the information expressed by the discourse, intermediate betweenlanguage on the one hand and (a model of) "reality" on the other, wherethe discourse referents are "conceptual objects" and the conditions areascriptions of concepts to those objects (Asher 1993: 65). In a secondseparate step of interpretation, DRSs are given a model-theoretical inter-pretation, being true in a model (e.g. of the "real" world) if and only ifthe discourse referents can be associated with entities in the model sothat each of the conditions in the DRS is verified in the model, that is,if there is a mapping function that allows the DRS to be properlyembedded into the model.

As will be demonstrated below, the intermediate DRS level of represen-tation allows one to describe processes of anaphoric resolution, definite-ness, and various other discourse phenomena in a simple and efficientway.

1.2. Now let us return to our one-sentence discourse (1). What thespeaker states when uttering (1) and the recipient is informed of whenprocessing (1) is something like this: at the time when (1) is produced(«), the (collective) entity (x) called France is in the state (s) of mourningover an event (e) that took place before the time of production, andwhich is characterized as the dying of a (male) individual (y) with theproperty of being an actor, French, and very famous.3 This descriptioncorresponds to the discourse representation structure Kl given inFigure 1. The top box contains the DRS variables for the discoursereferents introduced in the discourse (x, y, e, s) or contextually given, likethe time of utterance (n, 'now'); and in the bottom part the conditionson the discourse referents are specified. There are different types ofdiscourse referents: individuals (x, y); eventualities, that is, events (e) andstates (s); times (i); etc. The symbols 0 and < denote the overlapping

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

524 C. Fabricius-Hansen

Kl:

e s x y

Frankreich(y)s: trauern- ber(y, e)s O ne:TOT-WERDEN(x)[e<n]franz sisch(x)Schauspieler(x)[m nnlich(x)]sehr-bekannt(x)

Kl':

Figure 1. DRSsfor discourse (1) and discourse ( Γ )

erTOT-WERDEN(x))e < nfranz sisch(x)Schauspieler(x)[m nnlich(x)]

s1: sehr-bekannt(x)s'< n

Frankreich(y)s: trauern-iiber(y, e)s O n

and precedence relation between (the time of) an event or a state andsome other event or time (or between times directly). The conditions inbrackets — that χ is male and that e precedes η — are not expressedseparately in (1), in a lexical item or morpheme of their own, but canbe inferred from some other condition so expressed — that χ isSchauspieler 'male actor' — or by common-sense reasoning: the dyingevent must have taken place before the mourning sets in.4 I have enrichedthe formalism in this ad-hoc manner so as not to complicate mattersunduly by having to do some sort of explicit lexical decomposition.

In the end, the DRS Kl' assigned to (Γ), as shown in Figure 1, ispractically identical to Kl because the discourse referents presupposedby the anaphor er 'he' and sein- 'his' in the second and third sentence of(Γ), respectively, will eventually be identified with the referent χ intro-duced by the indefinite noun phrase ein Schauspieler in the first sentence;and the event referent presupposed by the definite noun phrase seinenTod in the third sentence must be identified with the event referent eintroduced by the main verb sterben 'die' in the first sentence.5 Butwhereas Kl is constructed in one complex step due to the fact that (1)consists of one sentence only, (Γ) is processed sentence by sentence, asshown in Figure 2. First, a DRS ΚΓ/1 is constructed for the first sentence,then that DRS is combined with the DRS Kl'/2 to yield the extendedDRS Kiyi +2, and that again is updated with the DRS Kl'/3 for thethird sentence, resulting in the final DRS ΚΓ. The updating itself is aunification process: unlike indefinite noun phrases and finite verb phrases,anaphoric pronouns cannot introduce new discourse referents,6 so thereferents they introduce — represented in the DRSs for the individualsentences by variables with the condition '= ?' (Asher 1993) — have tobe identified with discourse referents in the DRS built up so far; and the

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 525

KWl:e: TOT-WERDEN (x)e < nfranz sisch(x)Schauspieler(x)[m nnlich(x)]

K171:

z s n

s': sehr-bekannt(z)s' < nz = ?

K171+2:

e: TOT-WERDEN (x)e < nfranz sisch(x)Schauspieler(x)[m nnlich(x)]

s1: sehr-bekannt(x)s' < n

K173:

s y z e n

Frankreich(y)s: trauern- ber(y, e')s O ne1: TOT-WERDEN(z)

Figure 2. Construal of DRSfor discourse (Γ)

conditions on those referents will consequently be unified with the condi-tions on the discourse referents they are identified with.7 Likewise, definitenoun phrases normally demand that the context contain an accessiblediscourse referent compatible with the condition given in the noun phrasethat the discourse referent introduced by the noun phrase can be identifiedwith (uniqueness and familiarity presupposition) (Asher 1993: 84).

Kl and ΚΓ (Figure 1) differ only in minor respects: the condition thatthe dying event e precedes the utterance time is introduced explicitly inΚΓ, triggered by the tense form (present perfect) of the verb introducingthe event referent e itself. In Kl, however, that condition is inferredpragmatically from the condition that y mourns over e. Κ Γ also containsone discourse referent that has no counterpart in Kl, viz. the state s'characterized as a state of the type 'sehr-bekannt(x)', which is introducedin ΚΓ/2 and localized temporally by the finite verb war 'was' in thesecond sentence of (Γ); the condition isehr-bekannt(x)' is present in Kl,too, but there it is not represented as a (possibly) temporally restrictedstate, that is, as a stage-level property, but simply as a property of theindividual x. This difference between Kl and ΚΓ seems rather unimport-ant, and it may be due to an inadequate representation of the copulaconstruction — the DRT literature has, in fact, little to say concerningthe relationship between predicative and nonpredicative adjectives.8

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

526 C. Fahr icius-Hansen

Disregarding the situationally given utterance time («), the problematics'9 and the corresponding condition 6s<n9 in ΚΓ, we conclude that Kland Κ Γ both introduce four discourse referents and state altogether eightconditions on those referents. Two of those conditions — ie<rf and'mannlich(x)' — are stated explicitly in Κ Γ but lexically entailed orpragmatically inferred in Kl. In (1), the whole amount of informationis expressed in one sentence containing ten word forms, whereas in (Γ),it is distributed over three sentences and 14 word forms. Taking the inputDRS to a discourse to be empty (cf. Kamp and Reyle 1993), let us callthe set of nonredundant conditions contained in the DRS eventuallyconstructed for the discourse together with the discourse referents intro-duced in the discourse the discourse information. Then, obviously, thesentences forming discourse (Γ) on an average contain less discourseinformation than the single sentence that makes up discourse (1); and inthis sense the informational density of (1) might be said to exceed thatof (Γ). But the average amount of discourse information per sentence isnot the only important thing. Let us imagine for a sentence S in somediscourse that just two out of six discourse referents introduced duringthe processing of S will eventually turn out to be new in S, that is, notidentifiable with some other discourse referent in S, and that just threeout of nine conditions explicitly stated in S turn out to be new in S, thatis, not derivable from some other condition expressed in S. Then onecould call the information less concentrated, less condensed in that sen-tence than in a sentence where, for example, four out of six discoursereferents and/or five out of nine conditions were new in the given sense.That is, the less unification is demanded in the processing of S and theless redundancy there is among the conditions in the DRS constructedfor S, the more concentrated is the information given in S. And of course,given two discourses showing the same average of discourse informationper sentence but differing as to the relative frequency of new discoursereferents and nonredundant conditions per sentence, the discourse withthe highest average concentration of information per sentence will bejudged more loaded with information than the other. Apart from minordifferences,9 (la)-(lb) below represent other possible discourse realiza-tions of Κ1/ΚΓ lying somewhere between (1) and (Γ) as far as theinformational density is concerned, (la) consists of one sentence like (1),but the relative clause causes more unification than is needed for theprocessing of (1); and (Ib) contains two sentences, one more than (1)but one less than (Γ).(1) a. Frankreich trauert ber den Tod eines franz sischen

Schauspielers, der sehr bekannt war.'France mourns the death of a French actor who was veryfamous.'

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 527

b. Ein französischer Schauspieler, der sehr bekannt war, ist ge-storben. Frankreich trauert über seinen Tod.

French actor who was very famous has died. France mournshis death J

1.3. So far, we have viewed informational density in a text as a functionof the average amount of discourse information per sentence and thedegree of concentration with which the information is presented in eachsingle sentence. One way to avoid repetition and redundancy is to relyon so-called accommodation, that is, to use presupposition-triggeringconstructions in contexts that do not explicitly satisfy the presuppositions,thus making it necessary to enrich the DRS under construction with the"missing" elements (discourse referents and/or conditions) so as to pre-vent the processing from breaking down.10 Definite descriptions, forinstance, require that the discourse referent they introduce and describebe identified with an "old" discourse referent compatible with the descrip-tion; that is, in certain respects they behave like anaphoric pronouns (seeabove). So if the context does not provide an appropriate discoursereferent-condition pair, it will be accommodated in the DRS constructedfor the sentence in question in order for the processing to proceed (cf.Saeb0 1993: 27). That is what happens in (2a) below, which may open adiscourse between two persons both of whom know NN. The informationthat NN has written a — or one salient -— book on tense is not stateddirectly, but presupposed by the definite description; and the listener willhave to accommodate the discourse referent together with the conditionthat it is a book on tense written by NN into his or her DRS for (2a)before processing the main predicate. (Depending on her or his knowledgeof NN's [latest] publications, s/he may of course be able to identify theaccommodated referent with a referent in the DRS representing her orhis private knowledge.) In (2b), the information that NN has written abook on tense is expressed directly in the first sentence and provides anappropriate "antecedent" for the definite description in the second sen-tence of (2b) and (2c) — as well as for the anaphoric pronoun in (2d).In the end, then, (2a)-(2c) are assigned virtually the same DRS, viz. K2in Figure 3 (abstracting away from event, state, and time referents), butthe steps leading to that DRS are different. And according to our "defini-tion," (2a) has the highest informational density of the four texts.(2) a. Das Buch, das NN über Tempus geschrieben hat, ist unlesbar.

The book that NN has written on tense is unreadable.'b. NN hat ein Buch über Tempus geschrieben. Das Buch, das NN

über Tempus geschrieben hat, ist unlesbar.'NN has written a book on tense. The book that NN haswritten on tense is unreadable.'

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

528 C. Fabricius-Hansen

K2:

y

Buch-über-Tempus(x)NN(y)geschrieben-haben(y, x)unlesbar(x)

K2e:

Figure 3. DRSsfor discourses (2a)-(2d) and (2e)

y

Buch-über-Tempus(x)NN(y)POSS(y, x)unlesbar(x)

c. ' NN hat ein Buch über Tempus geschrieben. Das Buch istunlesbar.'NN has written a book on tense. The book is unreadable.'

d. NN hat ein Buch über Tempus geschrieben. Es ist unlesbar.'NN has written a book on tense. It is unreadable.'

Factive predicates like bedauern 'regret' are particularly clear examplesof accommodation-triggering constructions. The fact presupposed by thesecond sentence in (3) — that NN has written a book on tense — neednot be connected to anything mentioned earlier in the discourse or beingpart of what to the listener knows already. It is simply entered into theDRS for that sentence, together with the asserted condition that NNregrets that fact.

(3) NN ist eine merkwürdige Person. Sie bedauert, daß sie ein Buchüber Tempus geschrieben hat.'NN is a strange person. She regrets having written a book on tense.'

1.4. Accommodation may lead to a DRS containing more informationthan is licensed overtly by the sentence being processed and the DRSbuild up so far (Saeb0 1996: 189). On the other hand, accommodation islanguage-driven, being triggered by specific syntactic constructions and/orlexical items. (2e) illustrates another mechanism leading to discourserepresentations containing information that goes beyond what is "said"in the discourse itself.(2) e. NNs Buch über Tempus ist unlesbar.

'NN's book on tense is unreadable.'If nothing is said to the contrary, one would probably take (2e) to expressthe same thing as (2a). Consequently, if a DRS is a mental representationof a discourse, (2e) under this interpretation should be assigned the sameDRS as (2a), viz. K2. But of course, (2e) in itself is rather unspecific orunderdetermined as far as the nature of the relation between NN andthe book in question is concerned; NN might for instance be the ownerrather than the author of the book (cf. Kamp and Reyle 1993: 240ff.).

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 529

Linguistically, (2e) gives rise to the DRS K2e (Figure 3) only (wherePOSS is the underdetermined relation expressed by the genitive case),whereas the step from K2e to K2 — the specification of the relationbetween χ and y — is motivated by conceptual and/or world knowledgerather than linguistic knowledge. But again, measured against the finalDRS K2, one might claim that the information load in (2e) is higherthan in (2a) because (2e) in the given (default) context conveys the sameinformation as (2a) but with less linguistic material.

1.5. Initially, I asked what it means for a text A to have a higherinformational density than some other text B. We have seen now that wewould probably say that the informational density is higher in A than inΒ if at least one of the following conditions holds, other things beingequal:

i. the average amount of discourse information per sentence is higherin A than in B;

ii. on an average, the individual sentences in A contain fewer redun-dant conditions and give rise to less unification of discourse referentsthan the sentences in B;

iii. more accommodation is needed to process A than to process B;iv. there are more underdetermined conditions in A than in B with

the effect that the DRS constructed for A needs to be enriched bynonlinguistic knowledge systems to a higher degree than does the DRSforB.

(i) and (ii) concern the relation between explicit information and overtlinguistic "material" in the discourse: the more information per sentenceand per word, the higher the informational density, (iii) and (iv), on theother hand, have to do with what is said "between the lines"; suchinformation may be prompted by linguistic means, viz. presuppositiontriggering expressions and the like in (iii), or it may, to readers or listenerswith a given knowledge background, be inferrable from what is actuallysaid, in (iv). One might object that since the information in question isin a certain sense not "there" in case (iv), it should not be taken intoaccount when measuring the informational density in a discourse. Onthe other hand, as long as such information will be inferred more or lessautomatically by standard default reasoning, given a standard knowledgebackground, and the writer therefore expects the reader to add it to thelinguistically motivated information in his or her final mental representa-tion of the discourse, it is somehow part of the discourse; and beingshorter, the discourse may reasonably be said to have a higher informa-tional density than a text giving rise to the same DRS by overt linguisticmeans. Be that as it may — as far as German and Norwegian expository

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

530 C. Fabricius-Hansen

prose is concerned, the parameters mentioned above may all help toexplain the impression that German texts tend to be heavier, more loadedwith information than parallel Norwegian texts. What I claim, then, isthat the author of an expository German text in the typical case will usefewer sentences and/or fewer clauses than the author of a correspondingNorwegian text to convey the same information. The German text willcontain more constructions that call for accommodation and completionby inference on the part of the reader. Consequenly one would expectGerman texts to contain more "content words" per sentence and perfinite verb, fewer anaphoric pronouns, more semantically complex words(e.g. compound nouns), etc., than Norwegian texts. Empirical investiga-tions of so-called parallel corpora consisting of parallel authentic texts(Fabricius-Hansen and Solfjeld 1994) and German texts together withtheir Norwegian translations (cf. Solfjeld, this issue) seem to confirmthese hypoheses.

2. Informational density as a problem for translation(German-Norwegian)

2.1. If it is true that Norwegian nonfictional prose on the whole has alower information load than German nonfictional prose in the senseoutlined above, then one should expect stylistically adequate Norwegiantranslations from German to differ from their source texts in correspond-ing ways (cf. Solfjeld, this issue). In translating from German intoNorwegian, one would typically have to spread the information containedin the source text somewhat, using more anaphoric elements and morelexical redundancy in order for the translation product to be judgedstylistically adequate relative to the given text type (cf. Koller 1981). (1)above is a case at hand: although a transfer lexicon for all lexical items("content words") in (1) and ( ) can be established, as shown in (4),the most direct translation o f ( l ) , as in (5), sounds rather clumsy orstilted — far more so than (1) itself. (Besides, the implicit condition thatthe actor is male, is lost [see below] because skuespiller in contrast toSchauspieler is neutral as to sex.) One would have to choose a lesscompact alternative like, for instance, (5').11

(4) Frankreich Frankrike FranceSchauspieler skuespiller actord0d Tod deathsehr bekannt svaert kjent very famousmännlich mannlig male

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Kl'

e:TOT-WERDEN(x)franz sisch(x)Schauspieler(x)[m nnlich(x)]e < n

s1: sehr-bekannt(x)s'< n

Frankreich(y)s: trauern- ber(y, e)s O n

Figure 4. DRSs for discourses (Γ) and (5')

Informational density 531

K51

e s s xe:d0(x)fransk(x)skuespiller(x)e < n

• mannlig(x)s': svaert-kjent (x)s' < η

Frankrike(y)s: s0rge-over (y, e)s:0n

trauern bersterbensein

s0rge overd0vaere

mourndiebe

(5) Frankrike s0rger over en svaert kjent skuespillers d0d.Trance mourns the death of a very famous French actor.'

(5') En fransk skuespiller er d0d. Han var svaert kjent. Frankrike s0rgerover hans d0d.Ά French actor has died. He was very famous. France mournshis death.'

(5') corresponds to (Γ) and will eventually get the DRS K5' shown inFigure 4, the Norwegian counterpart of ΚΓ. Κ Γ and K5' differ only inone respect: the maleness condition on χ — 'm nnlich(x)' — is in a wayimplicit or dependent in ΚΓ, being part of the explicit condition'Schauspieler(χ)'. But in K5' the corresponding condition 'mannlig(x)'is explicit, being introduced independently in the DRS by the anaphoricpronoun han, which like English he presupposes a male antecedent.

2.2. Now, let us consider a more complicated example, viz. the authenticGerman text extract in (6).12

(6)[1] Politisch betrachtet, nahm die Reformation in Deutschland einen

anderen Weg als die reformatorischen Bewegungen in der Schweiz,den Niederlanden und England.'Politically, the reformation took another course in Germany thanthe reformation movements in Switzerland, the Netherlands, andEngland.'

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

532 C. Fabricius-Hansen

[2] Die Option Luthers für die Bindung der reformierten Kirche an dieLandesherren'Luther's decision to tie the reformed church to the regionalsovereigns'und die Auseinandersetzung zwischen dem reformatorisch aufbegeh-renden städtischen'and the conflict between the reformatorily uprising town'Bürgertum und den sozial-rebellierenden Bauern stützten imErgebnis die feudale'burghers and the socially rebelling farmers supported in the end'Kleinstaaterei in Deutschland, die sich hier aus geographischenGründen ohnehin besser'feudal particularism in Germany, which here for geographicalreasons anyway better'erhalten konnte als in den westlichen Nachbarstaaten,'last could than in the western neighboring states.'

(60[2] Luthers beslutning om a knytte den reformerte kirken til fyrstene

og konflikten mellom de reformatorisk-opr0rske borgerne i byeneog de sosialt rebellerende b0nder stettet til syvende og sist oppunder det feudale smästateri i Tyskland, som uansett av geografiskegrunner kunne klare seg bedre der enn i de vestlige nabolandene.'Luther's decision to tie the reformed church to the regional sover-eigns and the conflict between the burghers rising in favor of thereformation and the socially rebelling farmers in the end supportedfeudal particularism in Germany which for geographical reasonsanyway was able to last longer there than in the neighboring statesin the west.'

(6') gives a Norwegian translation of the second sentence of (6) thatkeeps as close to the source text as possible. The result is stylistically asbad as its English equivalent. It seems almost impossible to translate thissentence into good Norwegian — or English, for that matter — withoutbreaking it up into more sentences and clauses. That is, one has toorganize the information contained in that single sentence in a way thatallows the reader to build up in several steps the same DRS (or ratherthe Norwegian equivalent to that DRS) as the DRS that is constructedfor (6/2) (= example 6/sentence 2) in one very complicated step. It wouldtake us too far to go through the process of constructing a DRS for(6/2) in all details. But for the sake of illustration, we shall take at leastsome preliminary steps.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 533

Obviously, the finite verb of the main clause, stützten 'supported',introduces an event referent (eO). The two coordinated definite nominaliz-ations — die Option Luthers für ... 'Luther's decision ,..' and dieAuseinandersetzung zwischen ... 'the conflict between ...' that make upthe subject of the main clause likewise introduce event referents (el, e2).But as we saw above, discourse referents connected to definite descriptionsshould be identified with discourse referents already present in the dis-course and compatible with the descriptions (uniqueness and familiaritycondition). The events in question have probably been explicitly men-tioned in some part of the preceding text. But for el at least, it is alsoconceivable that the author has not done so, trusting the intended Germanreaders to know already that Luther chose to tie the reformed church tothe regional sovereigns. In that case, el is expected to be identified notwith an event referent in the DRS for the (preceding) text itself, but withan event referent in the DRS representing the knowledge of the educatedGerman reader. To readers not sharing that knowledge, the definite articlewould be unmotivated and call for accommodation of a more absolutekind (see above). An adequate Norwegian translation might thereforeintroduce this event separately, in a sentence of its own, before referringto it as one of the factors supporting feudal particularism in Germany.13

The present participles modifying Bürgertum 'burghers' and Bauern 'farm-ers' also introduce eventualities (e3, e4), as does the verb of the relativeclause modifying Kleinstaaterei (e5). The construction rules do not permitmodifying participles and relative clauses, being embedded within nounphrases, to introduce discourse referents directly into the universe — thetop box — of the main DRS. In the present case, however, the modifiersin question are nonrestrictive and therefore give rise to existential presup-positions, such as the presupposition that the burghers fought for thereformation, etc. So e3 and e4 will eventually be introduced into theuniverse of the whole DRS by way of accommodation. But in accordancewith the "anaphoric island" property of noun phrases, eventualities intro-duced by modifying participles, even in the nonrestrictive case, are notaccessible antecedents (Asher 1993: 76) for anaphoric pronouns withinthe same or a following sentence. That is, e3 and e4 act as so-calledimplicit discourse referents, as indicated by the boldface symbols inFigures 5 and 6: referents that are existentially quantified, just as"normal" ones are, but not accessible by pronouns (cf. Kamp andRoßdeutscher 1994a: 123).14 The eventualities introduced by finite verbs(eO and e5) are explicitly located relative to the speech time n. Thetemporal location of the other eventualities is underdetermined, butnonfinite verb form and/or pragmatic reasoning will help determine itmore or less unambiguously relative to n or relative to some eventuality

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

534 C. Fabricius-Hansen

in the DRS universe. In addition to the eventualities, the discourseuniverse contains a number of discourse referents of other types, viz.singular or collective individuals (sums of individuals) introduced byproper names (Luther, Deutschland) and definite noun phrases (die refor-mierte Kirche 'the reformed church', die Landesherren 'the regionalsovereigns', das städtische Bürgertum, 'the burghers', die Bauern 'thefarmers', die feudale Kleinstaaterei 'feudal particularism', die westlichenNachbarstaaten 'the neighboring states in the west'). Simplifying, we cannow assign to (6/2) the DRS K6/2 shown in Figure 5. Here temporalconditions that are not explicitly stated but can be inferred are added inbrackets. Likewise it is inferred that the discourse referent V (the neigh-boring states in the west) is meant to be identical to or at least include —symbolized by « — the referent V (Switzerland and the Netherlandsand England) introduced in (6/1). The first sentence in (6) will get theDRS K6/1 in Figure 5, and updating K6/1 with K6/2 yields the DRS K6shown in Figure 6.

From K6/2 we can generate more textlike representations like (7/2a-f)and (8/2a-f). Here, anaphoric elements and definite descriptions withexplicit antecedents are set in italics. In (8), text-structuring entities (seebelow) have been added and emphasized. Apart from minor differencesconcerning the explicitness of temporal conditions, (7) and (8) will get

K6/1: K6/2:

u z V U e n

Deutschland(z)die Reformation in z (u)die Schweiz&die Niederianden&England (V)die reformatorischen Bewegungen in V (U)e: u politisch gesehen einen anderen Weg

nehmen als U

y z X y' Z u1 V eO el e2 e3 e4 e5 n

Luther(x)die reformierte Kirche(y)die Landesherren(X)el: dafür optieren, y an X zu binden[el<n]das städtische Bürgertum(y')die Bauem(Z)e2: y'sich mit Z auseinerandersetzen[e2<n]e3: reformatorisch aufbegehrcn(y')[e3<n][e30e2]e4: sozial rebellieren(Z)[e4<n][e40e2]die feudale Kleinstaaterei(u')Deutschland(z)die westlichen Nachbarstaaten(V')e5: u' sich aus geographischen Gründen

besser in z als in V' erhalten könnene5<neO: el&e2 im Ergebnis u1 in z stützeneO<n

Figure 5. DRSsfor sentences [I] and [2] of discourse (6)

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 535

K6:

y z u X y' Z U u1 V V e eO el e2 e3 e4 e5 n

Deutschland(z)die Reformation in z (u)die Schweiz&die Niederlanden&England(V)die reformatorischen Bewegungen in V (U)e: u politisch betrachtet einen anderen Weg nehmen als Ue<nLuther(x)die reformierte Kirche(y)die Landesherren(X)el: dafür optieren, y an X zu binden[el < n]das städtische Bürgertum(y')dieBauern(Z)e2: y' sich mit Z auseinerandersetzen[e2<n]e3: reformatorisch aufbegehren(y')[e3<n][e30e2]e4: sozial rebellieren(Z)[e4<n][e40e2]die feudale Kleinstaaterei(u')die westlichen Nachbarstaaten(V')[V «V]e5: u1 sich aus geographischen Gründen

besser in z als in V' erhalten könnene5<neO: e l &e2 im Ergebnis u'in z stützeneO<n

Figure 6. DRSfor discourse (6)

the same DRS as (6), namely K6. The Norwegian translations of (7)and (8) are given in (7') and (8') respectively.

[1] Politisch betrachtet nahm die Reformation in Deutschland einenanderen Weg als die reformatorischen Bewegungen in der Schweiz,den Niederlanden und England, (e)

[2a] Luther optierte dafür, die reformierte Kirche an die Landesherrenzu binden, (el)

[2b] Das städtische Bürgertum und die Bauern setzen sich miteinanderauseinander. (e2)

[2c] Das Bürgertum begehrte reformatorisch auf. (e3)[2d] Die Bauern rebellierten sozial. (e4)[2e] Luthers Option (el ) und die Auseinandersetzung zwischen Bürgertum

and Bauern (e2) stützten im Ergebnis die Kleinstaaterei inDeutschland. (eO)

[2f] Sie/ Die Kleinstaaterei konnte sich dort aus geographischen

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

536 C Fabricius-Hansen

Gründen ohnehin besser erhalten als in den westlichenNachbarstaten. (e5)

(T)[1] Politisk sett tok reformasjonen en armen retning i Tyskland enn

reformasjonsbevegelsene i Sveits, Nederland og England, (e)'Politically, the reformation took another course in Germany thanthe reformation movements in Switzerland, the Netherlands, andEngland.' (e)

[2a] Luther gikk inn for ä knytte den reformerte kirke til fyrstene. (el)'Luther decided to tie the reformed church to the regional sover-eigns.' (el)

[2b] Det var konflikt mellom borgerne i byeneog b0ndene. (e2)'There was a conflict between the burghers in town and thefarmers.' (e2)

[2c] Borgerne reiste seg til fordel for reformasjonen. (e3)The burghers rose for the reformation.' (e3)

[2d] Bondene gjorde oppr0r av sosiale grunner. (e4)'The farmers rebelled for social reasons.' (e4)

[2e] Luthers valg o g konflikt en mellom borgere og bonder st0ttet tilsyvende og sist opp under smästatfeudalismen i Tyskland. (eO)'Luther's decision and the conflict between burghers and farmersin the end supported feudal particularism in Germany.' (eO)

[2f ] Den/Feudalimen kunne uansett av geografiske grunner klare segbedre der enn i de vestlige nabostatene. (e5)'It/Feudal particularism was able to last longer there than in theneighboring states in the west for geographical reasons anyway.'(e5)

(8)[ 1 ] Politisch betrachtet nahm die Reformation in Deutschland einen

anderen Weg als die reformatorischen Bewegungen in derSchweiz, den Niederlanden und England, (e)

[2a] SO optierte Luther dafür, die reformierte Kirche an dieLandesherren zu binden (el)

[2b] UND es gab eine Auseinandersetzung zwischen dem städtischenBürgertum und den Bauern (e2):

[2c-d] Das Bürgertum begehrte reformatorisch auf (e3), WÄHRENDdie Bauern sozial rebellierten. (e4)

[2e] Dies stützte im Ergebnis die Kleinstaaterei in Deutschland. (eO)[2f] Sie/Die Kleinstaaterei konnte sich dort aus geographischen

Gründen ohnehin besser erhalten als in den westlichenNachbarstaten. (e5)

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 537

(8')[1] Politisk sett tok reformasjonen en annen retiring i Tyskland enn

reformasjonsbevegelsene i Sveits, Nederland og England, (e)'Politically, the reformation took another course in Germanythan the reformation movements in Switzerland, theNetherlands, and England.' (e)

[2a] Luther gikk inn for a knytte den reformerte kirke til fyrstene.(el)'Luther chose to tie the reformed church to the regional sover-eigns,' (el)

[2b] OG det var konflikt mellom borgerne i byeneog b0ndene (e2):'And there was a conflict between the burghers in town and thefarmers.' (e2)

[2c-d] Borgerne reiste seg til fordel for reformasjonen (e3)9 MENSbondene gjorde oppr0r av sosiale grunner. (e4)'The burghers rose for the reformation (e3) whereas the farmersrebelled for social reasons.' (e4)

[2e] Dette st0ttet til syvende og sist opp under smästatsfeudalismeni Tyskland. (eO)'This eventually supported feudal particularism in Germany.'(eO)

[2f ] Den/Smästatsfeudalismen klarte seg uansett av geografiske grun-ner bedre i Tyskland/ifer enn i nabolandene i vest. (e5)'It/Feudal particularism managed better there than in the neigh-boring states in the west for geographical reasons anyway.' (e5)

As far as I can judge, the German original (6) is a fairly good representa-tive of German expository prose. (8'), on the other hand, can pass asquite normal Norwegian expository prose. And (8') is a stylistically moreadequate translation of (6) than (60. Now, what distinguishes (6) fromthe German version (8) that functioned as the immediate source of (8')?And what has happened on the way from (6) to (8') in addition to thedirect translation of words and phrases?

(6) consists of two sentences. The second sentence has a rather complexsyntactic structure, containing two coordinated subject noun phrases (dieOption ... and die Auseinandersetzung ...), one subordinate clause (therelative clause modifying the object of stütz ten, viz. die Kleinstaaterei...),and two clauselike participle phrases functioning as premodifyingparts of noun phrases postmodifying the second subject nounAuseinandersetzung. The sentence contains (at least) six different piecesof information in the form of conditions on eventualities (eO-e5). Theeventuality eO described in the main clause is introduced independently

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

538 C. Fabricius-Hansen

and directly as new information by way of the construction rules. el~e4occur in syntactically subordinate positions and are propagated to theuniverse of the main DRS by more indirect means, el and e2, as part ofdefinite descriptions, are clearly not (presented as) new at all but ratheraccommodated; and e3, e4 are declared as nonaccessible implicit referent(cf. above). The information (e5) contained in the nonrestrictive relativeclause may be asserted rather than presupposed, but at any rate, it isintroduced in a subordinate noun-modifying position like the attributivepresent participles giving rise to e3 and e4, thereby being marked assecondary. We conclude that the main clause informationally has aspecific position in (6/2): the eventuality eO it introduces and describesstands out as the main, or basic, or focused information unit in thesentence. The information given in (the main clause of) that sentenceobviously should be understood as a specification of or elaboration onthe information presented in the first sentence, viz. that the reformationpolitically took another course in Germany than in its neighbors in thewest. Thus our text has a very simple and transparent structure at textlevel: it consists of a sentence introducing a topic and a second sentenceelaborating on that topic. At the syntactic-semantic level, however, thesecond sentence shows a complex hierarchical structure that packstogether a lot of information, presenting some as main and some as moresecondary, but on the whole without specifying any discourse relationsbetween the individual pieces of information. The adverb ohnehin'anyway' signals that there is no (primary) causal connection betweensome fact mentioned in the context (the main clause) and the eventualitydescribed by the sentence containing the adverb (the relative clause) (cf.Eisenberg and König 1984). Describing an eventuality that holds priorto or independently of the events introduced in the main clause, therelative clause may be said to contain background information.

(7/2) represents an intermediate step between the DRS K6/2 and thetext version (8/2) underlying the Norwegian translation (872). Here, theinformation given in (6/2) is distributed over a sequence of six syntacti-cally very simple sentences, each introducing and describing one of thesix eventualities eO-e5. In other words, the hierarchical ordering charac-terizing (6/2) has been replaced by a nonhierarchical, linear orderingprinciple at the syntactic level. This has the effect that the implicitdiscourse referents of K6/2 are made explicit; and the sentence corre-sponding to the main clause predicate of (6/2), that is, the sentencecontaining the main information (eO), must follow the sentences introduc-ing the eventualities el-e4 referred to in the subject constituent of themain clause of (6/2): el (Luther's decision) and e2 (the conflict betweenburghers and farmers) occur in the first argument position of the predicate

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 539

(stützten) and therefore must be available in the discourse universe asaccessible antecedents for anaphoric pronouns or definite descriptions inthe sentence introducing and describing eO; compare [2e] in (7). e3 ande4 jointly specify the conflict between burghers and farmers (e2) andtherefore somehow have to be presented in connection with e2, that is,before eO.

(7) is not particularly convincing as a text in its own right. It seemssomewhat incoherent or unstructured: apart from ohnehin in (7/2f) thereare no discourse particles or other overt signals of the text relationsbetween the individual sentences. And from their order and their contentalone one apparently does not get enough clues to build a coherentstructure for the whole text. The main problem is perhaps the sequence(7/2b-d). But the relation between (7/1) and the subsequent sentencesis also rather obscure, very much in contrast to the transparent topic-elaboration structure found in (6).

In (8), discourse relations are made more explicit. e3 and e4 aredescribed as contrasts specifying the conflict (e2) mentioned in the preced-ing sentence, compare (8/2b-d); and el and e2 are coordinated in(8/2a-c). The clumsy coordinated subject phrase in (7/2e) has beenreplaced by the abstract anaphor dies 'this/that'. That of course intro-duces a certain ambiguity in that there is more than one accessible andproper antecedent for the anaphor (see below), whereas the subject in(7/2e) is referentially unambiguous. But explicitly coordinating (8/2a)and (8/2b) helps in picking up the corresponding abstract entity as theintended antecedent.

As compared to (7), the number of sentences has been reduced by onein (8) because the sequence (7/2c-d) has been replaced by the complexsentence (8/2c-d). But still the syntactic structure is extremely simple.The main difference between (7) and (8) lies in the text-structuringdevices — particles, conjunctions, colon — used in (8), in combinationwith the abstract anaphor.

(8') is a direct translation of (8). Thus, as a translation of the Germanoriginal (6) it is based on a specific — but natural — interpretation ofthe discourse relation between (6/1) and (6/2) and between subparts ofthe information contained in (6/2). It goes without saying thattranslations that are less explicit in this respect are possible, too. Onesuch example is (9'), which is based on (9). Here, as in (6/2), theinformation that the burghers fought for the reformation (e3) and thatthe farmers rebelled for social reasons (e4) is syntactically subordinate(in relative clauses) to the information that there was a conflict betweenburghers and farmers (e2); and they are not directly or explicitly markedas entering into a discourse relation with each other or with the main

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

540 C. Fabricius-Hansen

clause. On the other hand, the changed order of the sentence introducingel and the sequence describing e2 together with e3 and e4 has reducedthe ambiguity created by the anaphor in the sentence describing eO.(9)/(9') consists of one sentence less than (8)/(8'), but one of the senten-ces — (9/2a)/(9'/2a) — is syntactically more complex than any of thesentences in (8)/(8').

(9)[1] Politisch betrachtet nahm die Reformation in Deutschland einen

anderen Weg als die reformatorischen Bewegungen in der Schweiz,den Niederlanden und England, (e)

[2a] Es gab eine Auseinandersetzung zwischen dem städtischenBürgertum, das reformatorisch aufbegehrte, und den Bauern, diesozial rebellierten. (e2 [e3, e4])

[2b] UND Luther optierte (SEINERSEITS) dafür, die reformierteKirche an die Landesherren zu binden, (el)

[2c] Dies stützte im Ergebnis die Kleinstaaterei in Deutschland. (eO)[2d] Sie/Die Kleinstaaterei konnte sich aus geographischen Gründen in

Deutschland/öfori ohnehin besser erhalten als in den westlichenNachbarstaaten. (e5)

(9')[1] Politisk sett tok reformasjonen en annen retning i Tyskland enn

reformasjonsbevegelsene i Sveits, Nederland og England, (e)'Politically, the reformation took another course in Germany thanthe reformation movements in Switzerland, the Netherlands, andEngland.' (e)

[2b] Det var konflikt mellom borgerne i byene, som reiste seg til fordelfor reformasjonen, og b0ndene, som gjorde oppr0r av sosiale grun-ner. (e2, [e3, e4])There was a conflict between the burghers in the towns who rosefor the reformation and the farmers who rebelled for socialreasons.' (e2, [e3, e4])

[2c] OG Luther gikk (PA SIN SIDE) inn for ä knytte den reformertekirke il fyrstene. (el)'And Luther, on the other hand, opted for tying the reformedchurch to the regional sovereigns.' (el)

[2d] Dette st0ttet til syvende og sist opp under smästatsfeudalismen iTyskland. (eO)This in the end supported feudal particularism in Germany.' (eO)

[2a] Den/Smästatsfeudalismen klarte seg av geografiske grunner uansettbedre i Tyskland/der enn i nabolandene i vest. (e5)

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 541

'It/Feudal particularism managed better in Germany/there thanin the neighboring states in the west for geographical reasonsanyway.' (e5)

Let us summarize: (6), (7), (8), and (9) are all possible discourse realiza-tions of the DRS K6 shown in Figure 6. The second sentence in (6) hasa rather complex syntactic structure and a correspondingly high informa-tional density, introducing and describing six different eventualities. Therelation between the two sentences, however, is quite transparent in thatthe second sentence obviously elaborates on or specifies the informationgiven in the first sentence. (7) represents the other extreme: each sentence,being extremely simple syntactically, introduces and describes one eventu-ality; but as a discourse, (7) more or less falls apart because the relationsbetween the sentences as parts of a text are not sufficiently clear. (8) and(9) come close to the one-eventuality-per-sentence or one-new-informa-tion-unit-per-sentence type of discourse represented in (7), but in contrastto (7) they have a transparent text structure, signalled by discourseparticles, conjunctions, and the like. (8') and (9'), which correspond to(8) and (9) respectively, are undoubtedly stylistically more adequatetranslations of (6) than (6').

2.3. The transition from (6) to the (paraphrases underlying the various)translations is based on two main principles. The first principle I shallcall information splitting: different conditions on one and the same dis-course referent that are expressed within one syntactic constituent in theoriginal complex sentence are distributed over sentences that are syntacti-cally independent of each other, converting implicit or accommodateddiscourse referents into explicitly introduced referents. This step involvesthe use of definite descriptions, anaphoric pronouns, and similar meansto ensure that the conditions in question are in the end assigned to thesame discourse referent, that is, that the scattered information is in theend properly unified. And that again constrains the order of the indepen-dent sentences: the intended antecedents of anaphoric expressions andthe like must of course be introduced in sentences that precede thesentences containing the anaphoric expressions, etc., and they must beintroduced in a form and in an order that permit each anaphoric expres-sion or definite description to pick up the intended antecedent accordingto the general principles of anaphoric resolution.

Information splitting seems natural when the source text contains adefinite description that has no antecedent in the text itself, thus causingaccommodation of a discourse referent (see above). In such a case, onemay have to introduce the discourse referent in a separate sentence

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

542 C. Fabricius-Hansen

preceding the sentence that expresses the condition given in the source-text sentence containing the definite description. In particular, this holdsif the discourse referent in question is an eventuality (or more generally:an abstract object), that is, if we have a definite nominalization like dieOption Luthers ... and die Auseinandersetzung ... in our example above.The discourse referent in that case can be introduced explicitly in asentence with the corresponding verb before being referred to with adefinite description or an abstract anaphoric expression.15

The transition/translation from (6) to (7)/(7') is based on informationsplitting alone. But as we have seen, that is not enough. Something moreis needed, which goes beyond the representation given by DRSs. Thusthe second principle governing the transition/translation from (6) to(8)/(8') or (9)/(9') says that relations holding between the sentences ofthe source text as parts of a coherent text should be respected, that is,the paraphrase or translation must be true to the so-called discoursestructure (Asher 1993) of the (source) text. Discourse structure, however,is not represented in a DRS of the "classical" type discussed so far. Forinstance, the DRS K6 for (6) as a whole treats the information given inthe two sentences as equal although the (main clause of) (6/2) is obviouslyunderstood as specifying the information presented in the first sentence,viz. that the reformation politically took another course in Germany thanelsewhere.

Thus we conclude that a DRS may be an adequate representation ofthe information common to a whole family of texts that vary with respectto syntactic complexity at different levels (sentence, clause, noun phrase)and consequently with respect to informational density too. And providedthat DRSs can be "transferred" from one language to another,16 thetransfer of the DRS for a given text in the source language may be takento represent a family of possible translations of that text as far as theinformational content alone is concerned. But DRSs cannot do justiceto those aspects of texts that make them more than a mere bundle ofinformation. And these aspects are as important for the adequacy oftranslations as the informational content — sometimes maybe even moreimportant, depending on the genre.

3. Syntactic structure, segmented discourse structure, and translation

3.1. In his book, Reference to Abstract Objects in Discourse, Asher(1993) argues that the notion of simple semantic updating found in theoriginal DRT framework (cf. section 1) is inadequate if the task is tobuild a theory of discourse structure and discourse segmentation. As we

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 543

have just seen, one needs "an additional level of discourse interpretationover and above the semantic construction rules of DRS construction,"which Asher calls segmented DRSs or SDRSs (Asher 1993: 257). Thebasic constituents of SDRSs are the DRSs constructed for the individualsentences in the discourse in that "the full stop punctuation marks thebasic complete unit of communication — a thought — which is communi-cated with a particular purpose or intention on the part of the author"(Asher 1993: 270f.). So, as a default, each new sentence introduces a newbasic information unit and a new constituent in the discourse althoughin some cases "the default may be overridden and clauses or largerstretches of text may furnish basic discourse constituents" (Asher 1993:270f.).17 Basic discourse constituents entering into discourse relationswith each other may form complex discourse units (SDRSs), which intheir turn may enter into discourse relations with each other or withbasic constituents. The result may be a rather complicated hierarchicaldiscourse structure. Asher distinguishes between different types of dis-course relations, one type being so-called rhetorical relations like elabora-tion, continuation, explanation, generalization, and contrast that saysomething about the discourse itself rather than about the world (orwhatever the discourse is about). Discourse relations may be expressedovertly by discourse-structuring particles, conjunctions, and the like.Often, however, they are not expressed at all but have to be inferredfrom the order in which the basic units are presented in the discourse(i.e. sentence order), semantic knowledge, world knowledge, etc. Thus, adiscourse containing few discourse-structuring elements may be opento different plausible interpretations as far its discourse structure isconcerned.

According to the view outlined here, interpreting a discourse involvesat least two different tasks: on the one hand, one has to construct arepresentation of the information given in the discourse (a DRS), usingthe information contained in each sentence — supplemented by informa-tion from nonlinguistic knowledge systems — to update the informationgiven so far. On the other hand, one has to assign a segmented discoursestructure (SDRS) to the text, deciding for each new sentence how thecorresponding DRS should be attached to the discourse structure builtup so far and, if necessary, revising that structure in order to make roomfor the new sentence. Here, too, nonlinguistic knowledge may play a role.

Asher (1993: 256ff.) develops a whole theory of discourse structure,including some basic principles of discourse construction and rules forattachment, which I cannot discuss here.18 But I shall suggest a SDRSanalysis for our text example (6) and its paraphrase/translation (8)/(8').The analysis is meant as a necessary supplement to the DRS analysis

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

544 C. FabriciuS'Hansen

presented in the preceding section and will prepare the ground for somepreliminary generalizations concerning the second principle (discourse-structure fidelity) governing the transition/translation from (6) to (8)/(8')or (9)1(9).

3.2. As mentioned above, the discourse structure of (6) is fairlytransparent: the text consists of just two sentences, and the discourserelation that holds between these sentences is obviously elaboration. Thatmeans that the first sentence discourse dominates (Asher 1993: 267) thesecond. Thus we get the segmented DRS shown in Figure 7. Here thevertical arrow symbolizes discourse dominance and shadow letters denote

K6/1:

K6/2

u z V U e n

Deutschland(z)die Reformation in z (u)die Schweiz&die Niederlanden&England(V)die reformatorischen Bewegungen in V (U)e: u politisch betrachtet einen anderen Weg

nehmen als Ue < n

ty' z X Y Z u' V eO <el e2 < e3 e4» <e5> n

Luther(x)die reformierte Kirche(y)die Landesherren(X)el: dafür optieren, y an X zu binden

das städtische Bürgertum(y')die Bauern(Z)e2: y' sich mit Z auseinerandersetzen[c2<n]]e3: reformatorisch aufbegehren(y')[e3<n][e30e2]e4: sozial rebellieren(Z)[e4<e3][e40e2]die feudale Kleinstaaterei(u')Deutschland(z)die westlichen Nachbarstaaten(V')[V1 »V]e5: u1 sich aus geographischen Gründen besser

in z als in V erhalten könnene5<neO: el&e2 im Ergebnis u1 in z stürzeneO<n

Elaboration(K6/l,K6/2)

Figure 7. SDKS for discourse (6)

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 545

SRDSs. As was noted in section 2.2, it is in fact the main clause of acomplex sentence that is responsible for establishing discourse relationsbetween the sentence in question and other constituents of the discourse.The syntactic dominance structure by which the eventuality referents(apart from the main clause eventuality) have been propagated into thediscourse universe of K6/2 (see section 2.2) is marked by angled bracketsin Figure 7, the implicit referents being in bold face. The "irrelevanceparticle" ohnehin in the relative clause and the rhetorical relation (back-ground or something like that) holding between relative and main clausehave been ignored in the representation given in Figure 7.19

Now let us turn to (8). Figure 8 shows the DRSs contributed by theindividual sentences and clauses of (8) in the order in which they occurin (8). Anaphoric resolution has taken place as far as possible (seebelow), but the individual DRSs have not been unified into one big DRS,identical to K6 (Figure 6) apart from minor differences concerning theexplicitness of temporal conditions. It is evident that the amount ofdiscourse information contained in each individual DRS decreases towardthe end of the text — except for the very last sentence (8/2f), whichcorresponds to the relative clause in (6). In each of the first three sentencesthere are two or more new discourse referents and at least four conditions.The two following clauses (8/2c-d), however, each introduce and describeone (new) discourse referent only, viz. the eventuality referent connectedwith the verb. In K8/2e and K8/2f there is one new referent in additionto the eventuality introduced by the verb, and three explicit conditions,that is, again, less discourse information than in K8/l-K8/2b. The dis-course referent denoted by the second new variable w in K8/2e is intro-duced by the anaphor dies 'this/that' and therefore should be identifiedwith a referent already introduced. But being an abstract-entity anaphor,dies may need a segmented DRS rather than a simple DRS to get anappropriate antecedent (cf. Asher 1993: 258). And in fact, although othersolutions are possible too, it seems most natural to identify w with thecollection of events mentioned in the preceding sentences or with the factthat these events took place. In that case the corresponding DRSs,according to Asher (1993: 258), have to form one complex constituentin the segmented discourse structure assigned to the whole text. Theother — less plausible — candidates are (the eventuality introduced in)the clause immediately preceding the anaphor, that is, (8/2d), and the(collection of events described in) the complex sentence (8c-d).

We shall now try to assign a SDRS to (8) under the pragmaticallyquite reasonable anaphoric resolution suggested above, using "K8/1,"etc., as abbreviations for the corresponding DRSs in Figure 8 and shadowletters (given in italics in the text) for complex SDRSs.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

546 C. Fabricius-Hansen

K8/l[e]e u z V U nDeutschland(z)die Reformation in z (u)die Schweiz&die Niederlanden&England(V)die reformatorischen Bewegungen in V (U)e: u politisch betrachtet einen anderen Weg

nehmen als Ue < n

K8/2a[el]el y XLuther(x)die reformierte Kirche(y)die Landesherren(X)el: dafür optieren, y an X zu bindenel <n

K8/2b[e2]e2 y1 Zdas städtische Bürgertum(y')die Bauern(Z)e2: y sich mit Z auseinerandersetzene2<n

K8/2c[e3]e3e3: reformatorisch aufbegehren(y')e3<n [e30e2]

K8/2d[e4]e4e4: sozial rebellieren(Z)e4<n [e40e2]

K8/2e[eO]eO u' wdie feudale Kleinstaaterei(u')eO: W im Ergebnis u' in z stützeneO<nw = ?

K8/2f[e5]e5 Vdie westiichen Nachbarstaaten(V')[V «V]e5: u' sich aus geographischen Gründen besser

in z als in V erhalten könnene5<n

Figure 8. Construal of DRS for discourse (8)

Obviously, K2a, which introduces el (Luther's decision), must be takento open and K8/2b (introducing e2, the conflict between burghers andfarmers) to continue an expansion of K8/1. K8/2b in its turn is expandedby the clauses (8/2c-d), linked by the adversative subjunction während'whereas', which describe the behavior of burghers and farmers (e3 ande4), respectively. Grammatically speaking, (8/2c-d) is of course one

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 547

complex sentence — one Satzgefüge. But an adversative während clauseis not semantically integrated into its main clause. Abstracting from thecontext, (8/2c-d) is in fact equivalent to a sequence of two independentsentences where the second sentence contains an adversative discourseparticle like hingegen. Thus, the relation denoted by the subjunction is adiscourse relation (contrast) rather than a relation between eventualitiesdescribed in the DRSs. Apparently, then, the kind of clause linkage wefind here is a borderline case between subordination proper and juxtaposi-tion of independent sentences, and an exception to the default principlethat complete sentences furnish the basic discourse constituents (seesection 3.1): in the present case, the basic constituents between whichthe discourse relation contrast is taken to hold are (DRSs for) clauses,not sentences.20 In addition to expressing a contrast to K8/2c, K8/2dalso continues the expansion on K8/2b (cf. Asher 1993: 289ff.).

As mentioned above, K8/2a and K8/2b (together with the expansionof the latter) must form one complex constituent (Kf in the SDRS K8)(shown as shadow letters in the Figures) in order to provide an appro-priate entity for the variable w in K8/2e to be identified with.21 Therelevant discourse relation between what is described in K as a wholeon the one hand and K8/2e — which corresponds to the main clause of(6/2) — on the other hand could be a causal relation, as indicated bythe adverbial im Ergebnis 'in the end': what the political parties did hadthe causal effect that feudal particularism in Germany was supported.But according to Asher (1993: 281), "cause" is a so-called nonstructuralrelation and therefore cannot be linked directly to the complex constituentK'. Following the construction rule (ii) for updating nonstructural rela-tions (Asher 1993: 304), one has to construct a topic K' that closes offthe constituent K' and attaches to that topic, that is, one has to build aconstituent (K) within which the elaboration of K' in K' as well as thecause relation between K' and K8/2e is declared. The topic common toK8/2a and K8/2b might be an eventuality characterized as "politischesVerhalten der relevanten Akteure" [political reactions from the relevantparts] or something like that. In the end, then, the implicit topic DRSK' or the eventuality in question may be taken as the antecedent for theabstract anaphor dies in K8/2e (cf. the definition of availability given byAsher 1993: 313).

We now come to K8/2f, which corresponds to the relative clause of(6/2). The question is how it is related to the complex SDRS built up sofar, and in particular whether it is discourse-dominated by K8/1. I sug-gested above that K8/2f gives some supplementary background informa-tion on the discourse referent u' (die Kleinstaatrei) introduced in K8/2ein order to make it (more) understandable or plausible that the events

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

548 C. Fabricius-Hansen

K8/l[e]ι[pc u,K'[e*]

t K8/2e[eO] K8/2f[e5])litisches Verhalten der relevanten Akteure]

K8/2a[el] K82/b[e2]

4K8/2c[e3] K82/d[e4]Contrast(K8/2c, K8/2d)Continuation(K8/2c, K8/2d)

Continuation(K8/2a, K8/2b) Elaboration(K8/2b, IfC* )

Elaboration(K',iC ) Cause(K', K8/2e) Background(K8/2e, K8/2f)

Elaboration(K8/l, K)

Possibility A

Figure 9. TTiree possible SDRSsfor discourse (8)

described in Kr could have the consequences for feudal particularism thatare described in K8/2e. Thus K8/2e and K8/2f would seem to be linkedby the relation background (K8/2e, K8/2f). Being attached — by rule(i) of "updating using nonstructural relations" (Asher 1993: 276, 304) —to K8/2e within K, K8/2f in any case is not directly discourse-dominatedby K8/1. This analysis gives us the SDRS shown in possibility A ofFigure 9. It says that the proposition that there was some feudaleKleinstaaterei (u1) in Germany, which had survived for geographicalreasons and which was in turn supported by the way the political partiesacted, is ONE of the propositions that are implications of the propositionthat the course Reformation took in Germany was different from thecourse it took in the other countries.

It is also possible to interpret K8/2e as linked to the complex constitu-ent K' of possibility A of Figure 9 via continuation. According to therules (iii) for topic-based updating, we then construct a topic discoursedominating K1 as well as K8/2e, such that K8/2e continues the topic onefeels started in K1. "Die Folgen der Politik der Reformation" [consequences

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 549

K8/l[e]

K*[die Folgen der Politik der Reformation]

K

K'

K8/2a[el] K82/b[e2]

K"

K8/2c[e3] K82/d[e4]Contrast(K8/2c, K8/2d)Continuation(K8/2c, K8/2d)

K8/2e[eO] K8/2f[e5J

Continuation(K8/2a, K8/2b) Elaboration(K8/2bJK ' ' )

Continuation(K ', K8/2e) Background(K8/2e, K8/2f)

Elaboration(K*,K )

Elaboration(K8/l,K*)

Possibility B

Figure 9. (Com.)

of Reformation politics] could be a plausible candidate for such a topic.Provided that K8/2f is attached to K8/2e in the way suggested above,this analysis leads to the SDKS shown in possibility B of Figure 9. TheSDRSs presented in possibilities A and B of Figure 9 are not the onlydiscourse structures one could assign to (8). One might, for instance,claim that the existence of the Kleinstaaterei in Germany at the relevanttime is background to more or less everything expressed up to K8/2e,that is, that K8/2f should be attached higher up in the structure. In thatcase, we must construct a common topic for the complex constituentcontaining K8/1 and everything directly or indirectly discourse-dominatedby K8/1 on the one hand and K8/2f on the other hand — say some-thing like "Reformation und Kleinstaaterei in Deutschland und Westeuropa"[Reformation and particularism in Germany and Western Europe]. IfK8/2e, in its turn, is attached via the cause relation, as assumed inpossibility A, this analysis gives us the SDRS shown in possibility C ofFigure 9; it is a trivial task to construct a corresponding variant of thecontinuation alternative presented in possibility B of Figure 9.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

550 C. Fabricius-Hansen

K8

K*[Reformation und Kleinstaaterei in Deutschland und Westeuropa] K8/2f[e5]

K8/l[e]

K'[e*][politisches Verhalten der relevanten Akteure]

K8/2e[eO]

K1

K8/2a[el] K82/b[e2]

r1

K8/2c[e3] K82/d[e4]Contrast(K8/2c, K8/2d)Continuation(K8/2c, K8/2d)

Continuation(K8/2a, K8/2b) Elaboration(K8/2b, K *

Elaboration(K', κ ·) Cause(K', K8/2e)

Elaboration(K8/l,l

Elaboration(K*, K* ) Background( K*, K8/2f)

Possibility C

Figure 9. (Cont.)

Obviously, the original text (6) and the paraphrase/translation (8)/(8')are not strictly equivalent. In particular, the abstract anaphor dies'this/that' in (8/2f) is ambiguous in a way that the coordinated structurefunctioning as subject in the main clause of (6/2) is not. But (8) underthe discourse segmentations K8 shown in Figure 9 undoubtedly representsplausible interpretations of (6). Thus (8'), being better adjusted to thestylistic norms holding for modern Norwegian expository prose, maycount as a reasonably adequate translation of (6). This means that thereduced syntactic complexity that characterizes the translation from (6)to (8') has been compensated, so to speak, by an increased complexityas far as the more or less implicit discourse structure is concerned. Thesyntactic structure of the complex second sentence (6/2) is in fact almostmirrored by the SDKS K8 in possibility A of Figure 9 (and likewise for

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 551

K8/l[e]Elaboration

K8/2f[e5]CaUSe K8/2e[eO] Background[politisches Verhallen]

Elaboration L

(8/1) (8/2a) (8/2b) (8 (8/2d) (8/2e) (8/2f)

S(6/1) (6/2)

Figure 10. Mapping between syntactic structure of discourse (6) and SRDSfor discourse (8)

the interpretation in possibility B). This is shown in Figure 10. On thehorizontal line the sequence of sentences in (8) is symbolized. Below theline one finds — besides a symbol for the first sentence of (6) — a rathersuperficial graphic representation of the syntactic structure of (6/2), andabove the line a so-called constituent graph (Asher 1993: 289) for theSDKS K8 in possibility A of Figure 9. Again, for ease of processing Ihave added symbols for the eventualities introduced by the respectiveconstituents. We see that the syntactic structure of (6/2) is in a waymapped onto the SDKS structure, and vice versa: the participle phrases

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

552 C Fabricius-Hansen

(PtP), which are the most embedded syntactic constituents, correspondto the deepest embedded discourse constituents K8/2c and K8/2d. The(head of the) NP[e2] node dominating the participle phrases correspondsto the discourse constituent K8/2b d-dominating the SDRS K" that K8/2cand K8/2d are constituents of. The two coordinated NPs dominated bythe subject NP in the syntactic tree correspond to the two constituentsof the SDRS K1 in the constituent graph. The relative clause RC[e5]alone seems to break with the principle that the syntactic structure ismirrored by the discourse structure, keeping DRSs corresponding tosyntactic sisters together as constituents of the same complex SDRS andrespecting c-command relations. In the constituent graph, the discourseconstituent K8/2f corresponding to the relative clause is a sister of thediscourse constituent K8/2e corresponding to the verb of the main clause,whereas the relative clause is c-commanded by the verb in the syntactichierarchy.22 That is, the information contained in the relative clause ishigher in the discourse structure assigned to (8) than is warranted by thesyntactic structure of (6). It has, in a way, been hierarchically pro-moted — and as far as I can judge, that is in fact the feeling one haswhen comparing (6) and (8): that the last sentence of (8) has more"weight" than the corresponding sentence-final relative clause in (6).That again might explain the difficulties (8f) creates for the constructionof a SDRS for (8), as witnessed by the discussion above.

If K8/2f is indeed meant to be background information and if it isalso to be realized as an independent sentence, it would perhaps be morenatural — at least for a Norwegian text — to move it from its text-finalrelative-clause position to a position to the left of the information itserves a background for, as shown in (10)/(10').23 That would breakwith the principle that otherwise holds as far as the order of constituentsis concerned: that the basic constituents of the SDRS occur in the sameorder as the syntactic constituents their content is based on — as far asis compatible with the anaphoric aspects of information splitting, ofcourse (but see [9]/[9']). And it calls for other revisions, too, in order tobalance the change of order. But on the other hand, the result conformsstill better to the tendency we observed in Figure 8, that the amount ofnew information per sentence decreases toward the end of the text (sec-tion). In contrast to (8/2e) (cf. Figure 8), the sentence in (10) correspond-ing to the main-clause verb of (6/2), namely (10/2f), introduces one newreferent only, viz. the eventuality in question; and every sentence contain-ing information that is relevant for the processing and deeper understand-ing of that sentence precedes it — all premises precede the conclusion,so to speak.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 553

(10)[1] Politisch betrachtet nahm die Reformation in Deutschland einen

anderen Weg als die reformatorischen Bewegungen in der Schweiz,den Niederlanden und England, (e)

[2a] In Deutschland konnte die feudale Kleinstaaterei sich aus geo-graphischen Gründen besser erhalten als in den westlichenNachbarstaaten. (e5)

[2b] Luther optierte NUN dafür, die reformierte Kirche an dieLandesherren zu binden, (el)

[2c] UND es gab eine Auseinandersetzung zwischen dem städtischenBürgertum und den Bauern (e2),

[2d] INDEM das Bürgertum reformatorisch aufbegehrte (e3),[2e] WÄHREND die Bauern sozial rebellierten. (e4)[2f] Dies stützte im Ergebnis die Kleinstaaterei in Deutschland

ZUSÄTZLICH. (eO)(10')[1] Politisk sett tok reformasjonen en annen retning i Tyskland enn

reformasjonsbevegelsene i Sveits, Nederland og England, (e)'Politically, the reformation took another course in Germanythan the reformation movements in Switzerland, theNetherlands, and England.' (e)

[2a] I Tyskland hadde smästatsfeudalismen hadde av geografiskegrunner klart seg bedre en i nabolandene i vest. (e5)In Germany, feudal particularism had for geographical reasonsmanaged betterothan in the neighboring states in the west.' (e5)

[2b] Luther gikk NÄ inn for ä knytte den reformerte kirke til fyrs-tene. (el)'Luther now chose to tie the reformed church to the regionalsovereigns.' (el)

[2c] OG det var konflikt mellom borgerne i byene og b0ndene (e2):'And there was a conflict between the burghers in town and thefarmers.' (e2):

[2d-e] Borgerne reiste seg til fordel for reformasjonen (e3), MENSbondene gjorde oppr0r av sosiale grunner (e4).'The burghers rose for the reformation (e3) whereas the farmersrebelled for social reasons.' (e4)

[2f ] Dette stottet til syvende og sist YTTERLIGERE opp undersmästatsfeudalismen i Tyskland. (eO)'That eventually was a further support to feudal particularismin Germany.' (eO)

The SDRS one would assign to (10) would correspond to possibility A

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

554 C. Fabricius-Hansen

in Figure 9, except that K10/2a (corresponding to K8/2f) precedes thesub-SRDS K' (the SDKS containing K10/2b-e) and the discourse relationholding between the three constituents of the elaboration SDKS K wouldprobably be continuation rather than cause and background, respectively.But I shall not go into further details as far as the SDKS constructionfor (10)/(10') and other paraphrase/translations of (6) is concerned.

3.3. I mentioned above (section 2.3) that the second main principlegoverning the transition from (6) to the less compact paraphrases ortranslations presented in (8)/(8') and (9)/(9') — and in (10) — is thatthe target text should be coherent and maximally true to the discoursestructure exhibited by the source text. That is, it should be possible forthe recipient to assign to the target text a segmented discourse structurethat is in accordance with the discourse structure it would be natural toassign to the source text. But syntactically complex sentences may nothave — or may not need to be assigned — a segmented discourse structurebeyond what is expressed overtly by conjunctions, subjunctions, discourseparticles, and the like. Converting such a sentence into a sequence ofindependent sentences as part of a whole discourse therefore involves acertain amount of interpretative work, on the one hand, and the use ofdiscourse structuring devices — including sentence order — to make surethat the target text will or can get the "right" discourse interpretation,on the other hand. The example discussed above shows that as far as thetop level of the SDKS is concerned, that may be an impossible task; themore simple independent sentences per complex sentence in the sourcetext, the greater the risk that its top-level discourse structure will not berecoverable with reasonable certainty in the paraphrase or translation.The example also suggests that the paraphrase or translation shouldinvite the construction of an SDKS that mirrors the syntactic structureof the source text to a certain extent, as may be seen in Figure 10.

i. In the SDKS, information corresponding to the main clause (verb)of the sentence is not d-dominated by information coming from constitu-ents that are c-commanded by the main-clause verb (phrase) in thesyntactic structure.

ii. Sententialized bits of information extracted from syntactic sistersare constituents of one and the same sub-SDRS of the SDKS thatdiscourse-dominates the sententialized information extracted from thedominating syntactic node; compare <K8/2c, K8/2d> and <K8/2a,K8/2b> in Figure 10.

iii. Independently sententialized coordinated structures are linked bythe discourse relations continuation, parallel, or contrast', compare<K8/2a, K8/2b>.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 555

iv. The order of the basic constituents of the target SDKS correspondsessentially to the order of the syntactic constituents they are derivedfrom. But deviations may occur depending on anaphoric relations andthe discourse relations stipulated between the constituents; thus, sententi-alized information interpreted as background may perhaps be pushed tothe left in the target text, as shown in (10), and constituents linked by(nonnarrative) continuation, parallel, or contrast may change their internalorder, as shown in (9)/(9').

Such relations, of course, hold irrespective of whether the syntacticallysimpler text is a paraphrase of the more complex one, as in the examplediscussed so far, or the other way around. Thus, when a typicalNorwegian text is to be translated into a typical German text, one willhave to collect information on discourse referents that is distributed overtwo or more sentences (clauses) into one syntactically more complexsentence (clause) in a way that somehow mirrors the segmented DRSassigned to the text in question. Consequently, if discourse relations arenot expressed overtly, the text must be interpreted in this respect beforebeing translated or changed, in order to decide which information shouldbe syntactically downgraded and how.241 cannot discuss the correspond-ing translation difficulties and strategies at full length here. But I shallpresent and briefly comment on a translation example taken fromFabricius-Hansen (1986), where it is discussed more in detail.

The Norwegian original (11) is the fifth paragraph in a review of thebook The Nazi Question by Pierre Aycoberry. The review, written by aNorwegian professor of social sciences, was published in the newspaperDagbladet on June 27, 1983. The preceding paragraph contains a shortdescription of the overall aim of the book. A German sentence-by-sentence translation is given in (1 ).25

( )[1] Trivelig lesning er dette ikke.

This is not pleasant reading.'[2] MEN boka bar en rekke fortrinn.

'But the book has many merits.'[3] Aycoberry kaster sitt nett vidt.

'Aycoberry casts his net widely.'[4] Her er ikke BARE forskeres dom.

'Here is not only researchers' verdict.'[5] Han tar OGSÄ med samtidige observat0rers vurderinger.

'He also includes judgments from contemporary witnesses.'[6] Ofte er samtidige skildringer like fülle av innsikter om nazismen

som etterkrigstidens kj01ige studier.Often contemporary descriptions are as full of insights as balancedpostwar studies.'

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

556 C. Fabricius-Hansen

[7] MEN de kan ogsä vaere fülle av forvirring.'But they can also be full of confusion.'

(11')[1] Angenehme Lektüre ist dies nicht.[2] ABER das Buch hat eine Reihe von Vorzügen.[3] Aycoberry wirft ein weites Netz aus.[4] Hier findet man nicht NUR die Urteile von Forschern.[5] Er berücksichtigt AUCH die Aussagen von Zeitgenossen.[6] Oft zeigen die zeitgenössischen Darstellungen ebenso viel an

Einsichten wie kühle Forschungsarbeiten der Nachkriegszeit.[7] ANDERERSEITS sind sie oft voller Mißverständnisse.

When it comes to the SDRS construction for (11)/(1 ), the discourse-structuring particles and adverbs allow us to pair [1] and [2], [4] and[5], and [5] and [6], respectively; and the sub-SDRS containing the pair<[4], [5]> should obviously be understood as an elaboration of [3],specifying what is meant by 'casts his net widely'. [3] in its turn mightbe taken as the beginning of an elaboration of [2], specifying at least onemerit, viz. that Aycoberry casts his net widely. But then, what about theother merits? Somehow the expected continuation is missing, unless weare willing to accept the content of the sentence pair <[5], [6]> as such —which is rather unlikely: <[5], [6]> comment, rather, on the new referentintroduced somewhat implicitly in [4], that is, the contemporary descrip-tions. So let us imagine that the merit topic is in fact continued in afollowing paragraph, or that the text is simply ill-formed. In the end thatgives us an SDRS like Figure II.26 Stylistically, (11) may be ratherextreme even for a Norwegian text. But the German version in (1 )seems really bad. The (relatively free) translation given in (12) undoubt-edly conforms better to the stylistic norms holding for literary reviews asfar as the informational density is concerned.

(12)[1] OBSCHON keine angenehme Lektüre [Kll/l], ist Aycoberrys

Untersuchung ein wichtiges Buch [Kl 1/2].'Although no pleasant reading ([Kll/l]), Aycoberry's study is animportant book [Kll/2].'

[2] Der Verfasser holt weit aus [Kl 1/3];'The author covers very much' [Kl 1/3];

[3] er bringt NICHT NUR die Standpunkte der Forschung [Kl 1/4],SONDERN AUCH zeitgenössische Beurteilungen des Geschehens[Kl 1/5], die ZWAR häufig voller Mißverständsnisse stecken[Kl 1/7], oft ABER AUCH mindestens ebenso viel Einsicht in den

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 557

Kil

[Bewertung des Buches]

Kl l/l Kl 1/2

iKl 1/3 K? ...

Kl 1/4 Kl 1/5 K1

[zeitgenössische Schilderungen]

Kl 1/6 Kl 1/7Continuation(Kl 1/6, Kl 1/7)Contrast(K96, K97)

Continuation(Kll/4, Kl 1/5)Comment(Kl 1/5, K1) Elaboration(K', K * M )

Elaboration(K9/3,K")

Contrast(Kll/l,Kll/2) Continuation(Kll/l,Kll/2) Elaboratio(Kll/2,

Elaboration(K, K )

Figure 11. Possible SDRSfor discourse (11)

Nationalsozialismus bezeugen wie ausgewogene Forschungsarbeitender Nachkriegszeit [Kl 1/6].'he discusses NOT ONLY researches' views [Kl 1/4], but also con-temporary interpretations of the events [Kll/5], which are oftenfull of misunderstanding [Kll/7], BUT often ALSO demonstrateat least as much insight into nazism as balanced postwar studies[Kll/6].'

Here the seven sentences in the source text have been reduced to three.The first sentence (12/1) contains the contrasting information given inKl 1/1 and the topic DRS Kl 1/2 of the sub-DRS K' of Figure 11, down-grading Kll/1 rather than Kll/2 to a free adjunct because Kll/2 func-tions as an explicit topic for (or is explained by) the remaining text. Thesecond sentence in (12) corresponds to the topic DRS Kll/3 that is partof the elaboration of Kl 1/2. The third sentence (12/3) gives the informa-

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

558 C. Fabricius-Hansen

tion contained in the sub-SDRS K" elaborating on Kll/3 in Figure 11.It consists of a main clause corresponding to the contrasting DRSs Kl 1/4and Kl 1/5, and a relative clause corresponding to the most deeply embed-ded commentary sub-SDRS K"'. (12/3) is separated from (12/2) by asemicolon rather than a full stop, reflecting the fact that it is part of theelaboration of Kll/2. Figure 12 shows in a very simplified manner howthe SDRS Kll is mapped onto the syntactic constituents of (12).

4. Summary and conclusion

In the present paper I have been concerned primarily with theoreticaland practical aspects of translating between languages that followdifferent norms with respect to informational density, at least for certainnonfictional text types. The two language (norm) types are representedby German and Norwegian respectively:

German nonfictional prose in its typical form demonstrates a ratherhigh informational density, packing much information into each sentenceand/or clause by way of a complex syntactic structure at different levelsand relying heavily on accommodation and knowledge-based inferenceas a means of enriching the information expressed by overt linguisticmaterial. One might say that German nonfictional prose shows a predomi-nantly "vertical" structure.

Norwegian nonfictional prose typically has a more "horizontal," linearstructure, distributing the discourse information over a sequence of syn-tactically rather simple sentences with the effect that in the most extremecases, each sentence mentions and describes one eventuality only.Consequently, successful processing demands less accommodation andinference — but more unification — than in the German case; the resultis a very straightforward incremental kind of processing.

We have seen that discourse representation structures, as they weredefined originally (cf. Kamp and Reyle 1993), can adequately representthe informational content of discourses. And given a high degree oflexical (notional) mapping between source and target language, DRSscan be useful — or at least thought of— as a transfer level in translationas far as the informational content is concerned: the DRS assigned tothe source text is transferred — or "translated" — into a target-languageDRS from which target-language texts can be generated. If necessary, itmay be enriched by information from other knowledge systems so as toensure that the target-language recipient will in the end build up a mentalrepresentation (DRS) for the target text that is equivalent to the DRSthat the source-language recipient assigns to the source text.27 The targetDRS will represent a family of informationally equivalent possible

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 559

L ElaborationK

^%Contrast

Kfl/1 Continuation „Kll/2Elaboration?

Elaboration

Elaboration

K"'

ontra>Contin.

Kll/6 Kll/7

(11/1) (11/2) (11/3) (11/4) (11/5) (11/6) (11/7)

_ Explanation „

(12/1) (12/2)

Elaboration

Figure 12. Mapping between (syntactic) structure of discourse (12) and SRDS for dis-course (11)

translations of the source text — just as the source DRS represents notonly the source text itself but a family of informationally equivalentparaphrases of that text.

DRSs, however, are restricted to a representation of the informationalcontent. Other important aspects, such as what Asher (1993) calls thesegmented discourse structure of texts, on the one hand, and the illocutio-nary structure (cf. Brandt 1990,1994; Brandt and Rosengren 1992; Brandtet al. 1992), on the other, are not handled by the theory. One reason isthat the DRS, as it stands, represents the RESULT of processing only —information pure, so to speak, made up of discourse referents and condi-tions on those referents. It does not reflect the way these entities entered

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

560 C. Fabricius-Hansen

the DRS, whether directly via construction rules or indirectly by way ofaccommodation or inference. It says nothing about the relations betweenthe individual pieces of information, for example that one is an elabora-tion of another, or the purpose the speaker may have in giving theinformation, etc. But a translation has to respect discourse properties ofthis kind too, otherwise it will not count as a translation, or at least notas an adequate one. And as a tool for translation theory and practice,the DRT framework has to be extended — or supplemented —accordingly.

In this paper I have concentrated on the type of discourse propertiesthat Asher's framework of segmented discourse structure representationis meant to account for, leaving, for example, illocution in the strictersense aside. When it comes to discourse structure, it turns out that thesyntactic complexity characteristic of the German expository style iscompensated to a certain degree by a transparent discourse structure.This is because sentences — or more precisely: their DRSs — are in thedefault case the basic constituents of segmented DRSs: as far as the toplevel is concerned, the SDRS of a text with few but syntactically complexsentences will in the default case be simpler and more transparent —easier to construct — than the SDRS for a text distributing the sameinformation over a long sequence of independent and syntactically simplesentences, that is, a text in the typical Norwegian style.

I have tried to show that translation (or paraphrasing) from "verti-cally" structured texts with high informational density into "horizontally"structured texts of the one-eventuality-per-sentence type is a complicatedtask. It seems to be governed by two main principles, which I haveproposed calling (i) the principle of information splitting and (ii) theprinciple of discourse-structure fidelity.

i. According to the information-splitting principle, different condi-tions on one and the same discourse referent that are expressed withinone syntactic constituent in the source text may be distributed overdifferent sentences or clauses; in this process, implicit discourse referentsand accommodated referents or conditions may become explicit. Thatleads to an increased use of definite descriptions, anaphoric pronouns,and the like, and therefore to increased redundancy. And since theremust be appropriate and accessible antecedents for anaphoric elements,etc., when they are first used, information splitting strongly influencesthe order of the relevant sentences or clauses.

ii. The principle of discourse-structure fidelity says that the transla-tion should make it possible for the target-language recipient to assignto the target text a segmented discourse structure that is in accordancewith the discourse structure it would be natural to assign to the source

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 561

text. With information splitting at work at the same time, this meansthat the discourse relations holding between (the main clause of) thesentences in the source text should be intact after the number of sentenceshas increased. And it means in fact that the hierarchical syntacticstructures of the source text should somehow be mapped onto SDKSstructures; that is, the syntactic hierarchies are to a certain degree (re)in-terpreted as SRDS structures. The linguistic means of expressing dis-course relations between independent sentences are discourse particlesand the like. Consequently, one would expect an increased use of suchmeans in the target text (but see note 24).

Our discussion of (6) and its various paraphrases and translations hasshown that the two principles are in conflict with each other: the moreinformation splitting is done, the more difficult will it be to reconstructthe SDKS of the original text, at least at the top level. What one gainsin ease of processing, as far as the individual sentences and the construc-tion of a purely informational DRS (cf. section 2) for the whole text isconcerned, may have its cost when it comes to the construction of aSDKS for the more horizontally structured paraphrase or translation.The ideal translation, then, will have to find a reasonable balance betweenthe conflicting forces; that is, it must maximize information splitting —without loss of discourse information (see section 1.2), of course — asmuch as necessary relative to the given stylistic conventions but minimizethe SDKS ambiguities and changes that process may give rise to. Whereexactly the right balance is may to a certain degree depend on text genreand function: for argumentative texts, for instance, it will often be soimportant to keep the argumentative structure transparent that a certaindeviance from the norms concerning syntactic simplicity will be tolerated.But for narrative texts, it may be the other way around. At any rate, thequestion of the right balance normally cannot — and should not — besettled for individual sentences or sentence pairs once and for all: informa-tional density, discourse structure, and style are properties of texts.Consequently, the balance should be made with respect to the text as awhole: information splitting that seems quite adequate locally may turnout to be less adequate globally, that is, with a view to the overalldiscourse structure and style of the translation.

Translation from "horizontally" structured into "vertically" structuredtexts of course calls for the inverse operations: information collectingand mapping of SDKS onto syntactic structure. The main difficulty liesin assigning a SDKS to the source text and determining which conditionon a given discourse referent is to be syntactically downgraded, and howthat should be done.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

562 C. Fabricius-Hansen

Overall discourse structure and stylistic convention are aspects oftranslation that in my opinion do not quite get the attention they deservein Monika Doherty's extremely relevant work on translation and parame-ters of style (cf. for instance Doherty 1992, 1995). On the other hand, Ihave not even touched upon the subject that is her main concern: theconsequences that language-specific parameters like left vs. right peripher-icity have for the way information is structured within sentences — andtherefore should also have for translation.

It remains to be said that the conclusions I have drawn above arebased on relatively few examples. But it seems obvious that the relationsbetween "vertically" and "horizontally" structured texts (translations orparaphrases) that are informationally equivalent are an interesting subjectfor further research within the (S)DRT framework.

Received 28 June 1995 University of OsloRevised version received26 October 1995

Notes

1. I have benefitted very much from discussions in the Computer Linguists Group in Osloand Arbeitsgruppe Strukturelle Grammatik, Berlin, in particular with ManfredBierwisch, Bergljot B. Brynildsen, Kjell Johan Saeb0, and Henk Zeevat. I also thankMonika Doherty for critical and helpful comments, and Bergljot B. Brynildsen andStig Johansson for correcting my English. Finally, I want to thank an anonymousreviewer who wrote some detailed and very clarifying comments on my (S)DRTanalyses, which led — among other things — to a partial revision of the SRDS assignedto (8)/(8') hi the first version. Correspondence address: Germanistisk institutt,Postboks 1004, Blindem, N-0315 Oslo 3, Norway.

2. See Kamp and Reyle (1993) for a thorough introduction and Asher (1993) for interes-ting extensions of DRT; cf. also Guenthner (1987).

3. Maybe the object that is mourned over is the fact that the dying event has taken placeor the resultant state — the actor's being dead — rather than the event itself.Complications like this will, however, be disregarded here (see Asher 1993).

4. Schauspielerin is the derived noun denoting the concept of a female actor.5. Simplifying somewhat, I have assigned to the abstract noun Tod 'death' and the verb

sterben 'die' the same content, symbolized as TOT-WERDEN 'become-dead'.6. Definite descriptions der Tod eines ... Schauspielers in (l) may be exceptions to this

principle, due to the fact that the head noun denotes a functional concept (cf. Asher1993: 84; Löbner 1985). Here, at least, it has been treated like an indefinite description,introducing e in Kl (Figure 1).

7. For the purpose of demonstration I take anaphoric pronouns to be processed bystipulation of identity rather than by absorption, i.e. replacing the pronoun by thediscourse referent selected as an antecedent (cf. Kamp and Reyle 1993:269).

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 563

8. Cf. Kamp and Reyle (1993: 267ff.). See also the discussion of the somewhat nebulousstate category in Parsons (1990: 186ff.).

9. For example as to the fact or event interpretation of the object argument of trauernüber.

10. For presuppositions and accomodation within DRT see Van Der Sandt (1992), Kampand Roßdeutscher (1992), Sab0 (1993).

11. (5') in fact represents the other extreme, (i) might be a more plausible choice:(i) En svaert kjent fransk skuespiller er d0d. Frankrike s0rger over hans d0d.

very famous French author has died. France mourns his death.'12. From Gössmann (1978).13. Similarly, the definite noun phrase die feudale Kleinstaaterei in Deutschland presupposes

an "antecedent" in the universe of the text or the reader's knowledge; and again, if ithas not been introduced in the preceding text — e.g. in a sentence like Deutschland warzur Zeit der Reformation durch feudale Kleinstaaterei gekennzeichnet 'At the time of thereformation, Germany was characterized by feudal particularism' — there might begood reasons to do so in a Norwegian translation — this piece of information is notnecessarily part of the common knowledge of the intended Norwegian readers.

14. Nonrestrictive relative clauses, however, are accessible to a certain degree (Brandt1990,1994). That is, they do not behave like subordinate noun-modifying constituentsin all respects. Correspondingly, they are treated like independent sentences byRoßdeutscher (1994: 250). But even if assertional rather than presuppositional innature, the information conveyed by a nonrestrictive relative clause normally doesseem to be downgraded somehow — to carry less informational weight than informa-tion expressed in main clauses or independent sentences (cf. Brandt 1994; Lehmann1984).

15. It may in fact be necessary to use the splitting technique even in the case of indefinitenominalizations and definite nominalizations with explicit antecedents in order toconform as far as possible with the one-eventuality-per-sentence principle.

16. This, of course, is a totally unrealistic simplification. But German and Norwegian areclose enough lexically to make it an interesting fiction, at least.

17. Compare Brandt (1990,1994), Brandt et al. (1992).18. Asher (1993), of course, is just one among several more or less formalized approaches

to discourse structure that it would take us too far to discuss here; cf. for instanceMann et al. (1994), Polanyi (1988).

19. See Asher (1993: 292f.) for the treatment of discourse particles within the SDRSframework.

20. That may be true of the nonrestrictive relative clause in (6/2), too; cf, Brandt (1990,1994).

21. In fact, it is not quite clear to me what kind of abstract entity the variable w should betaken to represent, a collective eventuality formed by eventuality summation from eland e2 in K8/2a and K8/2b, or a more factlike entity based on the SDRS K itself (seeAsher 1993: 57ff. for a typology of abstract objects; Kamp and Reyle 1993: 305ff. for asummation of discourse referents). The first solution seems to be more in accordancewith the DRS for the original text K6/2 (Figure 5), where eO is described as "el&e2 u'in z stürzen." But then again, that description may be wrong — nominalizations likedie Option Luthers do not necessarily denote eventualities, they may also denote factsand the like, depending on the context.

22. Provided, of course, that relative clauses are indeed constituents of noun (or deter-miner) phrases. Under a different syntactic analysis — if e.g. nonrestrictive relativeclauses were adjoined at a higher level — the asymmetry noted above might disappear.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

564 C. Fabricius-Hansen

23. That would be an even more natural strategy if the discourse referent u' ('dieKleinstaaterei') that fills the second argument position in the description of eO had tobe introduced as a new referent in the text itself; cf. as an alternative to (7/2a) thesequence Es gab in Deutschland (z) eine feudale Kleinstaaterei (u'). Sie (u') konnte sichdort (in z) besser erhalten als in den westlichen Nachbarstaaten (V).

24. Strangely enough, discourse particles and similar structuring devices seem not to beused as frequently in Norwegian as in German nonfictional prose. That leaves moreroom for SDKS interpretation and increases the impression of incoherence thatGerman readers get when reading Norwegian (cf. Nickelsen 1983).

25. The German translations of (11) were done by Monika Nickelsen, a bilingual nativespeaker of German and a former colleague of mine; cf. also Nickelsen (1983) for asimilar translation example.

26. Replacing ... hat eine Reihe von Vorzügen in (1 /2) by ... ist ein wichtiges Buch hasmade the continuation problem discussed above disappear; the following sentencesequence now functions rather as an explanation.

27. Of course it won't work that easily for "real" texts. But it is an interesting andimportant task to find out how far one can get and what kind of obstacles prevent thesimple transfer strategy from being successful. Strand (1993) presents a critical analysisalong these lines; cf. also Dyvik (1994).

References

Asher, Nicholas (1993). Reference to Abstract Objects in Discourse. Dordrecht: Kluwer.Brandt, Margareta (1990). Weiterführende Nebensätze. Zu ihrer Syntax. Semantik und

Pragmatik. Lunder germanistische Forschungen 57. Stockholm: Almquist and Wiksell.—(1993). Zur Grammatik und Pragmatik von Partizipialattributen. In Satz und Illokution,

vol. 2, Inger Rosengren (ed.), 193-230. Linguistische Arbeiten 279. Tübingen: Niemeyer.—(1994). Subordination und Parenthese als Mittel der Informationsstrukturierung in

Texten. Sprache und Pragmatik 32, 1-38.—; and Rosengren, Inger (1992). Zur Illokutionsstruktur in Texten. LiLi, Zeitschrift für

Literaturwissenschaft und Linguistik 86, 9—51.—; et al. (1992). Satztyp, Satzmodus und Illokution. In Satz und Illokution, vol. l, Inger

Rosengren (ed.), 1-90. Linguistische Arbeiten 278. Tübingen: Niemeyer.Doherty, Monika (1992). Informationelle Holzwege. Ein Problem der Über-

setzungswissenschaft. LiLi. Zeitschrift für Literaturwissenschaft und Linguistik 84, 30-49.—(1995). Prinzipien und Parameter als Grundlagen einer allgemeinen Theorie der ver-

gleichenden Stilistik. In Stilfragen, Gerhard Stickel (ed.), 181-197. Berlin and New York:de Gruyter.

Dyvik, Helge (1994). Semantic representations and translation. SMART Working Paper(draft). Bergen.

Eisenberg, Peter; and König, Ekkehard (1984). Zur Pragmatik von Konzessivsätzen. InPragmatik in der Grammatik, Gerhard Stickel (ed.), 313-332. Düsseldorf: Schwann.

Fabricius-Hansen, Cathrine (1986). Zum Satzbau des Deutschen aus norwegischer Sicht.Zielsprache Deutsch 17,18-23.

—(1995). Kontrastive Stilistik am Beispiel Deutsch-Norwegisch. Ein Parallelkorpusprojekt.Jahrbuch Deutsch als Fremdsprache 21, 137-148.

—; and Solfjeld, Kare (1994). Deutsche und norwegische Sachprosa im Vergleich. EinArbeitsbericht. Arbeitsberichte des germanistischen Instituts der Universität Oslo 6. Oslo:University of Oslo.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Informational density 565

Gössmann, Wilhelm (l978). Deutsche Kulturgeschichte im Grundriß. Munich: Hueber.Guenthner, Franz (1987). Linguistic meaning in discourse representation theory. Synthese

73, 569-598.Kamp, Hans; and Reyle, Uwe (1993). From Discourse to Logic. Introduction to

Modeltheoretic Semantics of Natural Language. Formal Logic and Discourse Repre-sentation Theory. Dordrecht: Kluwer.

-; and Roßdeutscher, Antje (1992). Remarks on Lexical Structure, DRS Construction, andLexically Driven Inferences. Arbeitspapier des SFB 340 21. Stuttgart: SFB.

—; and Roßdeutscher, Antje (1994a). Remarks on lexical structure and DRS-construction.Theoretical Linguistics 20, 97-164.

—; and Roßdeutscher, Antje (1994b). DRS-construction and lexically driven inference.Theoretical Linguistics 20, 165-235.

Klein, Wolfgang; and von Stutterheim, Christiane (1992). Textstruktur und referentielleBewegung. LiLL Zeitschrift für Literaturwissenschaft und Linguistik 22, 67-92.

Koller, Werner (1981). Textgattungen und Übersetzungsäquivalenz. In KontrastiveLinguistik und Übersetzungwissenschaft, Wolfgang Kühlwein, Gerhard Thome, andWolfram Wills (eds.). Munich: Hueber.

Lehmann, Christian (1984). Der Relativsatz. Tübingen: Narr.Löbner, Sebastian (1985): Definites. Journal of Semantics 4, 279-326.Mann, William C; Mathiessen, Christian M. I. M. ; and Thompson, Sandra A. (1994).

Rhetorical structure theory and text analysis. In Discourse Description, William C. Mannand Sandra A. Thompson (eds.), 39-78. Amsterdam: Benjamins.

Nickeisen, Monika (1983). Norsk — r0dt, hvitt og blätt, tysk — kronglete og grätt? Spräkog spräkundervisning 14, H. 4, 13-20.

Parsons, Terence (1990). Events in the Semantics of English. A Study in Subatomic Semantics.Cambridge, MA: MIT Press.

Polanyi, Livia (1988). A formal model of the structure of discourse. Journal of Pragmatics12, 601-638.

Roßdeutscher, Antje (1994). Fat child meets DRT. A semantic representation of the openinglines of Kaschnitz' "Das dicke Kind." Theoretical Linguistics 20, 237-305.

Saeb0, Kjell Johan (1996). Anaphoric presuppositions and zero anaphora. Linguistics andPhilosophy 19,187-209.

Solfjeld, Kare (1988). Sprachwechsel und Stilwechsel. Übersetzung deutscher Sachprosa insNorwegische. In Teaching Translation. Papers Read at a Symposium at StockholmUniversity 6-7 March 1987, Gunnar Magnusson, and Sture Wahlen (eds.), 50-66.Stockholm: University of Stockholm.

Strand, Kjetil (1993). Dynamisk semantikk og maskinoversettelse. Unpublished manu-script, Oslo.

Van Der Sandt, Rob (1992). Presupposition projection and anaphora resolution. Journalof Semantics 9, 333-378.

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM

Brought to you by | University of Oslo in NorwayAuthenticated | 129.240.0.83

Download Date | 11/30/13 4:52 PM