Elusive connectives. A case study on the explicitness dimension of discourse coherence (1).

Fabricius-Hansen, Cathrine

Abstract

The present article is an explorative study concerned with the elusiveness of certain connectives (discourse particles), intralinguistically and across languages: the fact that one and the same connective may seem redundant in one context but indispensable in another context; and that a connective may tend to appear in one language under conditions where no explicit connective device is used in some other language. More specifically, the article deals with the German particles (adverbs) wieder 'again' and dabei, lit. 'thereby' and their glosses in English and Norwegian, which are studied across a corpus of text samples and sentence-aligned translations in German, English, and Norwegian (the Oslo Mulitlingual Corpus). The two first sections present a preliminary outline of the problem and the method to be used, along with some terminological clarification. Each of the following two main sections briefly outlines the semantics of one German particle, presents data from the corpus showing that it remarkably often has no explicit translational image in English or Norwegian, and ends by discussing characteristic cases and general tendencies to be derived from the data of comparison. The final section (Section 5) summarizes the findings and proposes provisional conclusions pointing towards bidirectional optimality theory as a fruitful theoretical background for further research in this area.

1. Introduction

Over the past two or three decades, discourse particles and connectives have been studied from different theoretical perspectives, and our insight into their "nature" has increased considerably. But one aspect (of some of them) has, as yet, received little attention: their optionality, or even apparent redundancy, related to their still not arbitrary status as a means to organize the coherence of the discourse. This is the subject of the present article.

In a relevance theoretic setting, for instance, the function of a discourse particle occurring in a sentence S is to restrict the set of implicatures allowed by S in the context of a preceding utterance or sentence 'S. The implicatures of an utterance include those contextual assumptions which the addressee has to supply in order to preserve his or her assumption that the utterance is consistent with the principle of relevance (cf. Blakemore 1992). The connective, then, makes the discourse relations (in a broad sense) between 'S and S and the information structure of the discourse more EXPLICIT by expressing overtly what might be inferred or implicated anyway; that is, what is already implicitly "there," at least potentially; and by filtering out certain possibilities, it makes the discourse more INFORMATIVE or precise, thus guiding the reader or hearer towards the interpretation intended by the author or speaker.

On the other hand, the interpretation potential of a semantically more or less underspecified connective is normally constrained in a systematic manner by the context it occurs in, including its position in S. In fact, the influence of the context may be so strong that it pushes the interpretation towards or even beyond the (fuzzy) borders of the semantic domain covered by the connective (Fabricius-Hansen and Behrens 2001).

The complicated interplay between what connectives allow and what their contexts demand has mostly been studied monolingually and very often on the basis of sequences of two sentences without a natural context (but see, e.g., Doherty 1999; Hasselgard et al. 2002; Fretheim and Johansson 2002; Saebo 2004). The present article approaches the subject from an interlingual text-oriented perspective, focusing on what I call "elusive" connectives in German, in particular wieder 'again' and dabei, lit. 'thereby.' These connective adverbs surprisingly often disappear (remain untranslated) in English translations from German and, conversely, emerge (translate nothing) in German translations from English (cf. Fabricius-Hansen 2001; Fabricius-Hansen and Behrens 2001); thus they would seem to corroborate von Stutterheim's (1997) observation that German tends to mark referential movement within the temporal domain more persistently than is the case, for example, in English. A similar but somewhat less marked tendency holds for translations between German and Norwegian.

In a recent paper, Zeevat (2003) suggests that discourse particles are markers of a relation of the content of the current sentence to the context (or another parameter of the utterance context) and can be there because of either a functional necessity (if the relation in question is unmarked, the wrong interpretation results) or a universal principle that requires the marking of the relationship. This would also require some functional explanation, but it is not the same: we would find the marker even if there is not a functional necessity (Zeevat 2003: 10).

That is, Zeevat shares with Blakemore the view that discourse particles (may) function as a device to prevent misunderstanding and guide the reader or hearer towards the "right" interpretation of the relationship between neighboring sentences or text segments. In addition, he assumes that "there is a strong functional pressure to have ways of expressing these relations," speculating that we may even have to do with a universal requirement of marking such relations. The nature of this requirement remains obscure, however.

Obviously, if connectives may disappear under translation without being compensated for by other means of expression, their use cannot be explained by functional necessity or universal requirement alone; these notions will have to be modified or relativized somewhat. Thus, studying the translational pattern of elusive connectives, in particular the conditions under which they tend to be omitted and added, and the effects of their (dis)appearance, may not only help us understand the semantics of the individual connectives involved in the study but also shed some light on general vs. language specific aspects of discourse linking. That is the primary concern of the present article. It should be understood as an explorative case study, providing empirical data that an adequate theory of discourse particles or connectives will have to take into account.

The study is based on three subcorpora of the Oslo Multilingual Corpus, (2) consisting of excerpts (of approximately 10-15,000 words each) from 20-30 original texts in English, German, and Norwegian, and their authorized translations into the two other languages. The parallel texts are sentence-aligned. Searching for again in English source or target texts, for instance, automatically gives you all sentences containing again and the corresponding sentences in the German and Norwegian target or source texts, including (at most) 25 sentences preceding and/or following the relevant sentence.

The article is organized in two main sections (Sections 3 and 4), one for each of the two German adverbs mentioned above, and a concluding summary (Section 5). Each main section offers a brief outline of the semantics of the connective, a presentation of quantitative findings, and a discussion of selected examples directed towards the following questions:

--How do wieder and dabei and their counterparts in, for example, English and Norwegian contribute to discourse interpretation or processing? How do they interact with (discourse) information structure?

--What triggers the "disappearance" or "emergence" of these adverbs in translation? Is it possible to identify discourse conditions that require the presence of one or the other?

First, however, my use of the term "connective" may need a comment.

2. Connectives and explicitness of linking

For some reason, terminological diversification and confusion is particularly abundant in the area we are concerned with here. It should therefore be stressed that I use the term connective in a rather broad sense, as a cover term for conjunctions (and/und but/aber), (discourse) particles like too/auch, even/sogar, and adverbs like therefore/deshalb, then/dann.

What they have in common--apart from being noncomplements--is that they "link" the abstract (event-like) referent described by the verb projection they command or modify to a discourse referent of the same type introduced in the relevant context. With sentence-internal conjunctions, of course, the relevant context is the immediately preceding clause. Otherwise it may comprise a sentence or a larger discourse segment to the left of but not necessarily immediately preceding the current sentence; or it may be identical to the situational context.

Some connective adverbs have a fairly transparent anaphoric or deictic origin that explains their connective capacity: they require a proper antecedent for their interpretation, that is, the link is established by anaphoric resolution. German dabei, lit. 'there-by' belongs to that category. Connectives like too/auch and again/wieder, on the other hand, can be described as nonanaphoric presupposition carriers; here the link is "picked up" by nonanaphoric presupposition justification (but see Zeevat 2003).

Following the same line of reasoning, prepositional adjuncts containing definite descriptions like for that reason/aus dem Grunde and under this condition/unter dieser Bedingung can be considered connectives, too. They differ from simple connectives like so, then, therefore by containing a (more) explicit description of the link, that is, the antecedent of the definite description, and its relation to the referent introduced by the modified clause itself; but the difference is of a gradual kind.

Widening the perspective, then, we end up with a continuum of connective devices, comparable to the "continuum of explicitness of linking" or "syndesis-asyndesis continuum" suggested by Lehmann (1988) but operating in two dimensions rather than one: with respect to the means of determining the "link" (henceforth: the antecedent), on the one hand, and the degree of explicitness or preciseness with which the relation between the two relevant discourse referents is determined, on the other hand.

Thus, in a given context, the ing-adjunct ("converb construction," cf. Haspelmath and K6nig 1995) having chopped them and the prepositional phrase after that may function as equivalent alternative expressions, receiving the same interpretation in context; cf. (1b) and (1c). The anaphoric prepositional phrase is inherently less explicit than the ing-adjunct when it comes to characterizing the event that is linked to the matrix-clause event but determines the temporal relation between the two events at least as explicitly as the ing-construction. Both are less explicit than the full temporal clause in (la) and more explicit (one way or the other) than the simple anaphoric adverb in (ld) (cf. K6nig [1995: 62]). As we shall see below (Section 4), English translations of German dabei operate over the whole range of possibilities within this area.

(1) He chopped the trees.

a. When he had chopped them, he shaped them.... (Thompson and Longacre 1985: 213)

b. Having chopped them, he shaped them.

c. After that, he shaped them.

d. Then he shaped them.

3. Wieder

3.1. Semantics

It is generally recognized that wieder like again and its counterpart in Norwegian (igjen) has two major uses, a REPETITIVE use (2) where it is often interchangeable with noch einmal, erneut (once more, anew; nok en gang, pany), and a RESTITUTIVE (reversive, (3) restorative (4)) use (3) implying the restoration of a state that has held before. (5)

(2) a. The testing I had done so far told me nothing about Dr P.'s inner world. Was it possible that his visual memory and imagination were still intact?

I asked him to imagine entering one of our local squares from the north side, to walk through it, in imagination or in memory, and tell me the buildings he might pass as he walked. He listed the buildings on his right side, but none of those on his left.

I then asked him to imagine entering the square from the south. AGAIN he mentioned only those buildings that were on the right side, although these were the very buildings he had omitted before. (OS1)

b. [...] Ich bat ihn, in seiner Erinnerung oder in seiner Vorstellung einen der Platze in unserer Stadt zu uberqueren und mir die Gebaude zu beschreiben, an denen er vorbeikam.

Er zahlte die [Gebaude] auf der rechten Seite, nicht aber die zu seiner Linken auf.

Dann bat ich ihn, sich vorzustellen, er betrete den Platz von Suden her. WIEDER beschrieb er nur die Gebaude zur Rechten, eben jene, die er zuvor nicht genannt hatte.

(3) a. Und nun [...] bekam ich dieses blasse, zarte, trockene, nach nichts schmeckende Ding auf die Zunge gelegt--ich war drauf und dran, es WIEDER auszuspucken! (HEB1)

b. And then [...] that pale, fragile, dry, tasteless thing was placed on my tongue--I almost spat it out AGAIN!

In theoretical semantics, it has been debated for many years whether to account for the repetitive-restitutive dichotomy in purely structural terms, in terms of lexical ambiguity or polysemy, or by other means. (6) In the present context, we need not go into that discussion, however. Following Klein (2001: 278), a nonformalized version of the structural approach found in von Stechow (2003), I shall simply say that the adverb makes the meaning contribution "... and this not for the first time" to the sentence it occurs in. What "this" is--for example, an event, as in (2) or the result of a reversible change of state, as in (3)--depends on what the adverb, semantically speaking, has scope over in the clause it modifies, and the information structure of that clause--which, in its turn, is strongly determined by the preceding context. The important point is that wieder/again is a presupposition trigger, instructing the reader or listener to search the previous discourse for the entity that justifies its presupposition, that is, a previous instance of "this." According to the standard analysis (or analyses), wieder/again, like auch/too, is void of assertive content, that is, its meaning contribution is purely presuppositional. (7) This view is now being been challenged, at least as far as the repetitive variety is concerned (Huitink 2003; Saebo 2004). But still it can be maintained that in a context justifying its presupposition, wieder/again does not influence the truth conditions of the sentence and discourse it occurs in. (8)

Typically, the restitutive variety occurs in discourses describing a sequence of two inverse changes-of-state (or a series of mutually complementary states) involving one and the same participant(s) x, as illustrated in (3): the tasteless thing getting into and out of the mouth (i.e. being first outside, then in, and then outside the mouth). That is, the adverb associates with the predicate of the modified clause or VP. The sentence has a neutral topic-comment structure with focus accent on the predicate, which represents new information, relevant alternatives being other things that might happen to or hold of x under the given circumstances, for example, that the tasteless thing remains in the mouth or is swallowed. The adverb itself is an unaccented "reminder" of given information, viz. that x has been in the state implied or described by the predicate before.

In the repetitive case, on the other hand, we typically have to do with a sequence of times or situations instantiating one and the same abstract property, as illustrated in (2): the property that Dr. P. mentions only buildings on the right hand side in the given situation. The time or situation referents in question are topical in the given (narrative) discourse. But as topic times, they may be implicit, that is, they are not necessarily expressed in the sentences themselves but may be taken over or construed from the preceding context (von Stutterheim 1997; Dimroth forthcoming). In (2), the topic times are the two occasions where the patient is asked to mention the buildings he passes in his imagination: but neither is overtly expressed in the sentences describing the patient's response to the test. Since the comment part of the sentence containing the repetitive adverb--the material to the right of the adverb--describes a property that is already instantiated in the discourse, it is deaccented. The adverb itself, then, is the only constituent left to carry the sentence accent which has to be realized (cf. Dimroth forthcoming; Saebo 2004). Thus, the contribution of the repetitive adverb is focused: the information that the current topic is not the first discourse referent instantiating the event type described in the current sentence.

The interplay between information structure at discourse level and the repetitive-restitutive dichotomy can be seen quite nicely in (4): as it stands, this text will preferably be understood as focusing on the protagonist's changing moods (smiling-gloomy-smiling). That is, lachelte/smiled in the first sentence is interpreted inchoatively, marking the beginning of a state that ends by 'his' becoming gloomy; and wieder/again will be deaccented, receiving a restitutive-like interpretation: the protagonist returns to his friendly mood.

(4) a. Sobald er mich sah, lachelte er ein wenig und hob die Axt, und es war schrecklich, mit welchem Zorn er auf das Holz losschlug. Er wurde dann finster und sang seine Lieder. Wenn er die Axt niederlegte, lachelte er mich WIEDER an, und ich wartete auf sein Lacheln wie er auf mich, der erste Fluchtling in meinem Leben. (EC1)

b. The instant he saw me, he smiled slightly and raised the ax, and it was terrible to watch his rage as he smashed into the wood. He became gloomy then and sang his songs. When he put the ax down, he smiled at me AGAIN, and I waited for his smile just as he waited for me, he, the first refugee in my life.

But if the second sentence is left out, as in (4'), a repetitive reading will be preferred. Now, the "quaestio" rather concerns what happens---or how the protagonist acts--at different topic times; and the adverb carries the accent, focusing on the information that the same thing happens the second time.

(4') a. Sobald er mich sah, lachelte er ein wenig und hob die Axt und schlug auf das Holz los. Wenn er die Axt niederlegte, lachelte er mich WIEDER an.

b. The instant he saw me, he smiled slightly and raised the ax, smashing into the wood. When he put the ax down, he smiled at me AGAIN.

3.2. Data

From a semantic point of view, wieder, again, and igjen are lexical counterparts (with certain exceptions as far as igjen is concerned), sharing the repetitive-restitutive variation described above. (9) But their mutual correspondence (10) in translations is relatively low, especially as regards wieder and again, indicating differences of language use (cf. Konig et al. 1990).

In the present context, we are concerned with what I shall call zero ratios for wieder as compared to again and igjen, that is, the frequency with which the adverb is "omitted" (X [right arrow] Zero) in translations into or "added" (X [left arrow] Zero) in translations from the other language. Zero correspondencies are opposed to translation pairs where the adverb has an explicit counterpart in the parallel text, that is, an expression encoding the (repetitive or restitutive) meaning contributed by the adverb itself. This expression need not be the lexical counterpart of the adverb as determined above. Thus, in all three languages, there are adverbial alternatives to the repetitive use of wieder/again/igjen, representing possible alternatives to the adverb in translation: noch einmal, nochmals, abermals, erneut, von neuem in German, once more, a second time in English, and atter (en gang), nok/enda en gang, pa ny(tt), en gang til in Norwegian. Prenominal another/en annen is another possible source or target of wieder with scope over an indefinite NP (5).

(5) a. If only he can find ANOTHER director like Francois Masson, his work will come into its own again. (ABR1)

b. Wenn er bloss WIEDER einen Regisseur wie Francois Masson findet, wird er einen neuen Durchbruch haben.

c. Hvis han bare kunne finne EN ANNEN regissor som Francois Masson, da ville arbeidet hans igjen komme til sin rett.

The restitutive variety has no general competitor in any of the three languages. But depending on the context, wieder may correspond to the prefix [re.sup.-11] (6) or to back in English and tilbake in Norwegian (7); etc.

(6) a. Auch der kleine, der molekulare Burgerkrieg dauert nicht ewig. Nach der Strassenschlacht kommen die Glaser, nach der Plunderung schliessen zwei Manner mit Zangen und Kabelenden das Telefon in der verwusteten Zelle WIEDER an. (HEB1)

b. Even the small-scale, molecular civil war doesn't last for ever. After the street battle, the glazier arrives; the telephone in the vandalized kiosk is RECONNECTED by two men with pliers and connection blocks.

c. Heller ikke den lille, den molekylaere borgerkrigen varer evig. Etter gateslaget kommer glassmesteren, etter haerverket kobler to menn med tang og ledningstumper IGJEN til telefonen i den odelagte kiosken.

(7) a. Aurora, die mit dem Erreichen der Grossjahrigkeit WIEDER in das Elternhaus ubersiedelt war [...], bekam die Folgen zu spuren. (EHA1)

b. Aurora, who on reaching legal age had moved BACK into the family home [...], felt the consequences.

c. Aurora, som i og reed myndighetsalderen var flyttet TILBAKE til barndomshjemmet [...], skulle snart merke folgene.

In the tables below, all such correspondencies are registered as nonzero correspondencies, including translation pairs involving lexicalized verbs and special varieties of igjen (see Note 9), for example, recognize : erkenn- ... wieder / kjenne igjen, wieder abschaffen : repeal. The main point is whether a "free" occurrence of the repetitive-restitutive adverb has an overt translation image in the corresponding (target or source) sentence, as in the examples above, or not, as in (8).

(8) a. "I should have agreed to teach summer school," Sarah said. "Something to give some shape to things." (AT1)

b. "Ich hatte diesen Ferienkurs doch ubernehmen sollen," sagte Sarah. "Damit die Dinge wenigstens WIEDER etwas Form annehmen."

c. "Jeg skulle sagt ja takk til laererjobben pa den sommerskolen," sa Sarah. "Et eller annet som kunne fatt litt orden pa tilvaerelsen."

Table 1 shows that almost 40% (39.4%) of the 249 wieder tokens in original German texts are omitted in the English target texts (wieder [right arrow] zero frequency, row 1) and, conversely, that more than one third (34.4%) of all wieder occurrences in German translations from English have no overt source (wieder [left arrow] zero frequency, row 2). In comparison, the zero correspondencies for again with respect to German are 9.9% for translations into German (again [right arrow] zero) and 5.4% for translations from German (zero [left arrow] again), that is, almost four and seven times as low, respectively; cf. Table 2.

The zero ratio is relatively high for wieder with respect to Norwegian (Table 3), too, and higher for source than for target texts (wieder [right arrow] zero 24.2%, wieder [left arrow] zero 16.8%). The zero ratio for igjen with respect to German (Table 4) does not differ significantly from the ratio for again (Table 2). (12)

The results for again compared to igjen, finally, are shown in Tables 5 and 6.

The zero ratio for again with respect to Norwegian (Table 5) corresponds fairly well to what we found with respect to German (Table 2). But again is even less elusive in Norwegian than in German translations (8.4%:9.9%) and it is added a little more often in translations from Norwegian than in translations from German (6.9%:5.4%). As for igjen with respect to English, Table 6 compared to Table 4 shows that the zero ratio igjen [right arrow] En does not differ significantly from igjen [right arrow] Ge (8.2%: 8.8%) whereas igjen emerges more often in translations from English than in translations from German (12.7% :8.8%).

Table 7 summarizes our findings as far as zero ratios are concerned.

If we--admittedly somewhat ad hoc--take a zero ratio of 10% to fall within the limit of what may be expected for one reason or the other, we can conclude that wieder does indeed disappear and emerge remarkably often in translations between German and English and quite often with respect to Norwegian, too. (13) Why should that be so, given functional equivalence between the connectives on the one hand and equivalence between source and target text(s) on the other hand? It should be added that wieder does not seem to be omitted quite arbitrarily, either. Thus, comparing English and Norwegian parallel translations, we find that 136 (= 71%) of 191 explicit translations of wieder in Norwegian have explicit counterparts in the English target texts, representing 82% of the explicit total in English (Table 8). And 43 (= 74%) of 58 Norwegian zero translations are also zero translations in English; but these represent less than half (44%) of the total number (98) of English zero correspondencies. That is, the English target texts differ from the Norwegian target texts mainly by omitting the adverb in 55 cases where Norwegian has an explicit translation: otherwise, the correlation between the two target languages is relatively high.

3.3. Discussion

Wieder/zero correspondencies involving English seem to be found predominantly in prototypical restitutive contexts where the presupposition triggered by the adverb is satisfied completely in the local context immediately to the left of the modified clause or VP; cf. (9)-(10).

(9) a. Der Mann ist zu Boden getaumelt, mit einem eigenartigen, kaum horbaren Schmerzensschrei, und dann sofort WIEDER aufgestanden, ohne dass ich ihm auch nur die Hand gereicht hatte. (PH1)

b. The man fell to the ground with a strange, almost inaudible cry of pain, then instantly stood up without my even offering him a helping hand.

(10) a. As I stared into the gash I heard a sharp noise, as of something sundering, and I shut my eyes in horror, and when I opened them I found myself somewhere else. (BO1)

b. Als ich in die Spalte starrte, horte ich ein scharfes Gerausch, als ob etwas zerriss, und schloss vor Entsetzen die Augen. Als ich sie WIEDER aufmachte, war ich irgendwo anders.

When nothing in the preceding context could prevent the reader or hearer from drawing the right inferences and thus arrive at the intended discourse interpretation anyway, wieder or its counterpart is redundant from a (discourse) semantic point of view; and that may favor its absence in English. Blatant discourse semantic redundancy does not seem to disfavor the use of wieder in German, however--perhaps because from the perspective of discourse processing it still has its merit as an overt instruction of how to embed the information expressed in the modified clause or VP into the current discourse representation; in fact, native speakers tend to find corresponding sentences or discourses without wieder less acceptable. But evidently, again is not excluded under similar conditions in English, either; cf. (3) above and (11) below. The two languages simply seem to follow different conventions or preferences in this area, with Norwegian somewhere in the middle.

(11) a. I shut my eyes and when I opened them AGAIN I saw people who walked backwards [...]. (BO1)

b. Ich schloss die Augen, und als ich sie WIEDER offnete, sah ich Leute, die ruckwarts gingen, [...].

If it is true that local redundancy favors the absence of again in English as compared to German, one would expect the indispensability of (wieder and) again to increase with the amount of processing, including accommodation and the like, demanded in order to arrive at the intended interpretation without that particle, or the amount of interpretation possibilities (indeterminacy) ruled out by its presence, that is, its informativeness.

In (12), for instance, the restitutive adverb, modifying the metaphorical VP die Kriegsflagge zu hissen/to hoist the flag, cannot be omitted: it triggers the presupposition that the resultant state, that is, the flag being up, has obtained before or in other words, that the change of state described as 'to hoist the flag' reverses a preceding change of state ('to lower the flag'). The need to justify that presupposition makes the reader interpret to hoist the flag as the reversal of the change of state described as changing their striking coloring for a monotonous dull grey (Fabricius-Hansen 2001). Without the adverb, the text becomes incomprehensible (incoherent).

(12) a. [...] ja, bei manchen [Korallenfischen] hat man den Eindruck, sie mussten die kampfauslosende Farbung ablegen, um eine friedliche Annaherung der Geschlechter uberhaupt moglich zu machen. Ganz sicher gilt letzteres fur die bunten, oft scharf schwarzweiss gezeichneten Fischchen einer Gattung von "Demoiselles", die ich mehrmals im Aquarium ablaichen sah und die zu diesem Behufe ihre kontrastreiche Farbung gegen eine einfarbig stumpfgraue vertauschen, um nach Vollzug des Laichaktes alsbald WIEDER die Kriegsflagge zu hissen. (KOL1)

b. [...] in some [coral fish], one has the impression that they are obliged to divest themselves of their fight-eliciting colors in order to make friendly contact between the sexes possible. This certainly applies to the demoiselle group; several times I saw a brilliantly black-and-white species spawning in the aquarium; for this purpose changing their striking coloring for a monotonous dull grey, only to hoist the flag AGAIN as soon as spawning was over.

Repetitive wieder/again tends to occur in a greater distance from its antecedent than the restitutive variety (Fabricius-Hansen 2001; Saebo 2004). Normally, then, it may be expected to have a higher functional load as a coherence-inducing device signaling '... and this not for the first time' (cf. Section 3.1). As witnessed in (13), the repetitive adverb may even be a necessary means to prevent the relevant clause or VP from locally inviting a contrastive interpretation that is excluded by the preceding (global) context (cf. Section 3.1): (14) omitting again/wieder would present the property of seeking Philby's interpretation as something new and, in combination with the contrastive topic now/jetzt, give rise to the implicature that Philby's interpretation was not sought on previous occasions--which is explicitly contradicted by the preceding context. That is, the connective explicitly blocks an interpretation that would otherwise be natural at the local level but lead to incoherence at the discourse level.

(13) a. When the present Soviet leader had arrived at the KGB as Chairman, Philby had already been there for years and was considered something of a star. He lectured on the Western intelligence agencies in general and on the British SIS in particular. [...].

The Chairman, a highly intelligent and cultured man, had shown a curiosity, short of fascination but above mere interest, in Britain. Many times, over those years, he had asked Philby for an interpretation or analysis of events in Britain, its personalities and likely reactions, and Philby had been happy to oblige. It was as if the KGB Chairman wanted to check what reached his desk from the in-house "Britain" experts and from those at his old office, the International Department of the Central Committee under Boris Ponomarev, against another critique. Several times he had heeded Philby's quiet advice on matters pertaining to Britain.

It had been some time since Philby had seen the new czar of all the Russias face to face. That was when he had attended a reception to mark the Chairman's departure from the KGB back to the Central Committee, apparently as a Secretary, in fact to prepare for his predecessor's coming death and to mastermind his own advancement.

And now he was seeking Philby's interpretation AGAIN. (FF1)

b. [...] Er hatte damals Philby oft um eine Deutung oder Analyse von Ereignissen in England gebeten, von wahrscheinlichen Reaktionen seiner fuhrenden Politiker, [...].

Es war schon funf Jahre her, dass Philby den Zaren aller Russen von Angesicht zu Angesicht gesehen hatte. Das war im Mai 1982 gewesen bei einem Empfang anlasslich der Ruckkehr des KGB-Chefs zum Zentralkomitee, angeblich als Sekretar, in Wahrheit abet zur Sicherung seines eigenen Aufstiegs nach Breschnews bevorstehendem Tod.

Undjetzt suchte er WIEDER Philbys Rat.

The repetitive adverb explicitly signals that the event type described by the clause containing the adverb is already instantiated in the discourse universe, as a condition on another discourse referent than the referent it is ascribed to in the current clause. Thus, it has a sort of anti-anaphoric function corresponding to, for example, another in NPs; and from the perspective of discourse coherence, it can be as obligatory as such a device. (15)

4. Dabei

4.1. Semantics

The connective dabei--which occurs in topic position (clause-initially) as in (15) and in the so-called middle-field as in (14)--consists of an overtly anaphoric component da 'there' combined with the preposition bei 'with, at' (etymologically = En. by). Mostly, it demands an abstract antecedent as understood by Asher (1993): an eventuality, a proposition, or a fact; cf. (14)-(15). (16)

(14) a. Vor einigen Jahren war ich, wahrend einer Turnstunde, vom Reck gesturzt; Sowade hatte mich aufgefangen und sich den kleinen Finger DABEI gebrochen, mir selbst war nichts passiert. (JUB1)

b. A few years before, I had fallen from the crossbar during P. E. Sowade had caught me and broken his little finger IN THE PROCESS. I was not hurt.

(15) a. Mein Problem ist es oft, nicht fragen zu konnen. DABEI bestehe ich fast nur aus Fragen. (PH1)

b. Inability to ask questions is often my problem. AND YET I'm made up almost entirely of questions.

As noted by Konig (1995: 62), the meaning of the preposition bei itself is very unspecific and strongly dependent on contextual enrichment. (17) Used as a spatial preposition, it situates the referent of its external argument in a region surrounding the referent of its internal argument. As a constituent of the connective dabei, bei normally relates entities of a more abstract type; so the region or domain it assigns to the internal argument (the referent of da-)--and which the referent of the external argument is situated in will have to be of a correspondingly abstract kind, for example, a "larger" eventuality or a set of (relevant) facts, depending on the nature of the arguments.

When the antecedent referent is an eventuality, dabei anchors the external eventuality argument, that is, the referent introduced by the modified verb projection, temporally and spatially in the antecedent referent; that is, it does not introduce a new reference or topic time or a new independent eventuality, but rather adds to the description of the eventuality introduced by the antecedent, specifying, for example, an "accompanying circumstance" or the like; cf. (14). In a context like this, dabei establishes a coherence relation between the relevant discourse segments. When the referents related by bei are propositions (including facts) or maybe illocutionary acts rather than eventualities, as in (15), we have to do with a rhetorical relation rather than a coherence relation, as defined by Asher (1993):

Some discourse relations segment a discourse on the basis of the rhetorical function of particular propositions in relation to propositions already established in the structure. These relations are called rhetorical relations. [...] Other discourse relations segment the discourse on the basis of relations between the eventualities introduced within the constituents. These relations are called coherence relations (Asher 1993: 264).

Unlike rhetorical relations, which are paradigmatically about constituents and only sometimes directly contribute to the truth conditions of these constituents, coherence relations [...] typically directly contribute to the truth conditional content of the constituents themselves (Asher 1993: 265).

As a rhetorical discourse marker, dabei is preferably found in clauseinitial position.

4.2. Data

Dabei differs from wieder in one important sense: it has no lexical counterpart in English or Norwegian, that is, neither language has a lexical connective with a similarly unspecific meaning. This can be seen from the fact that its translation images in English cover the whole range from simple adversative or concessive connectives over anaphoric prepositional phrases and converb constructions to subordinate temporal clauses (cf. Section 1), and similarly for Norwegian; cf. Table 9.

Since dabei may occur together with clause-initial connectives, for example, the coordinate conjunction und 'and,' and such connectives in their turn may be omitted or added under translation (Altenberg 1999), there will be borderline cases that do not render themselves easily to the explicit vs. zero categorization. I have chosen not to register the coordinate conjunct and/og alone as an explicit counterpart of dabei. Apart from and/og, English or Norwegian connectives that have no other counterpart in the German parallel text are categorized as explicit counterparts of dabei, if it makes sense from a semantic-pragmatic point of view.

Altenberg (1999) has suggested that not having a "natural" lexical equivalent could be one of the factors favoring omission of a connective under translation. Consequently, one would expect dabei to exhibit a higher zero ratio than wieder. This is confirmed by my data: as can be seen from Tables 10 and 11, dabei is even more elusive than wieder. Thus, the zero ratio for source text dabei amounts to more than 50% in English translation and little less than 40% in Norwegian target texts (Tables 10 and 11, row 1); some examples are given in (16)-(18).

(16) a. [...] ein schwarzaugiger, braunhautiger Halbwuchsiger kam in Begleitung eines ihm ahnlichen Kindes zur Tur herein und tauschte an der Theke eine grosse leere Weinflasche gegen eine volle um; DABEI stellte er das Kind als seinen Onkel vor. (PH1)

b. A black-eyed, brown-skinned adolescent came in with a child who looked like him, and went to the bar, where he exchanged a large empty wine bottle for a full one. He introduced the child as his uncle [...].

c. Det bor en del mennesker fra sydligere egner her ogsa: en sortoyet, morkhudet fremslenging kom inn i folge med en guttunge som var ganske lik ham; borte ved disken fikk de en stor flaske vin i bytte for tomflasken; den storste gutten fortalte at den minste var hans onkel.

'The bigger boy told that the smaller one was his uncle.'

(17) a. Sind Sie schon einmal im Wald [...] ausgerutscht und haben DABEI durch die Laubschicht am Boden in einen vermoderten Baumstrunk gegriffen? (PH1) 'Have you ever lost your footing in the woods [... ] and have dabei grabbed through the underbush into a rotting tree trunk?'

b. Did you ever lose your footing in the woods [...] and reach through the underbrush to grab a rotting tree trunk?

c. Har De noen gang vaert pa skogstur [...] og glidd og tatt Dem for og fatt tak i en ratten trelegg under lovet pa bakken?

'Have you ever been on a picknick [...] and lost your footing and reached out and got hold of a rotting tree trunk under the leaves on the ground?'

(18) a. Sie las dann, bequem auf ein Sofa ausgestreckt, ein oder zwei Stunden, bis es an der Zeit war, das Mittagsmahl einzunehmen. DABEI vermied sie die weit verbreitete Gewohnheit werdender Mutter, sich vom lebhaften Appetit verleiten zu lassen und fur zwei zu essen. (EHA1)

b. Then she read, stretched out comfortably on a sofa for one or two hours until it was time to eat lunch, IGNORING the common custom o fan expectant mother following her appetite and eating for two.

Once again, there is a fairly high correlation between the two target languages as far as the choice between explicit and zero translations is concerned: 38 of 52 (= 73%) explicit translations in English correspond to explicit translations in Norwegian where they represent 57% of the explicit translations; that is, the English target texts differ from the Norwegian target texts most conspicuously by preferring a zero solution in almost half of the cases where Norwegian chooses an explicit translation of dabei; cf. Table 12.

As shown in Tables 10 and 11, row 2, target text dabei exhibits a higher zero ratio than source text dabei: more than 70% and 50% with respect to English and Norwegian source texts, respectively. It should, however, be noted that 25 (22.7%) of the 110 target text clauses or VPs containing dabei translate English ing-adjuncts; and all of them fall under the zero category, thus representing 31.6% of that category. (18) In fact, clause or VP coordination with dabei in the second conjunct, as in (19), seems to be a kind of standard translation alternative for free ing-adjuncts describing an "accompanying circumstance."

(19) a. He raises his hands in mock defence (beautiful, long fingers, she immediately notices, WONDERING whether he is conscious of it too). (ABR1)

b. In gespieltem Entsetzen hebt er die Hande (wunderschone, lange Finger, wie sie sofort bemerkt und sich DABEI uberlegt, ob er sich dessert gleichfalls bewusst ist).

'... as she immediately notices and dabei wonders whether ...'

c. Han lofter hendene til forsvar pa fleip (vakre, lange fingre, ser hun straks, MENS hun undres pa om han ogsa er klar over det).

'... she immediately notices, while she wonders whether ...'

Norwegian--like German--has nothing corresponding to the English ing-construction. This structural difference between English and Norwegian may be partly responsible for the fact that zero correspondencies for dabei are found with considerably lower frequency in translations from Norwegian than in translations from English (50% vs. 70%).

4.3. Discussion

Generalizing our findings in Section 3.3, we expect that

(a) dabei will get an explicit translation in English and Norwegian target texts when the discourse relation expressed by dabei that is, the coherence relation accompanying circumstance (in a broad sense) or a related rhetorical relation cannot otherwise be established with reasonable certainty and efficiency in the given target context; and, conversely, that

(b) dabei tends to occur in translations from English or Norwegian when the relevant discourse relation is explicitly encoded or can be inferred in the source text but cannot otherwise be established with reasonable certainty and efficiency in the given German target context.

On the other hand, we also expect that

(c) dabei may be quite redundant in the source text, and that

(d) dabei sometimes is added in target texts without being necessary in order to establish the intended interpretation.

On the whole, my data seem to confirm these hypotheses. It should, however, be added that clause-initial dabei may also be motivated from a clause-internal processing perspective, viz. as a device to attain a "balanced information distribution" by placing the subject in clause-medial position (Doherty 2002, 2003).

In what follows, we shall look more closely at some examples that illustrate the points made above.

(20) and (21) are examples of type (a): if dabei or its translation image are left out in (20), one might understand the second sentence as an explanation for the fact established in the first sentence, that is, one might infer the discourse relation Explanation rather than link the second sentence to the first at the eventuality level, as describing an accompanying circumstance; and (21) would seem incoherent without the connective or its translation.

(20) a. Dem werdenden Macho wurden Mutproben und Schaukampfe abverlangt. DABEI musste ein strikter Ehrenkodex eingehalten werden. (HME1)

b. They were subjected to tests of courage and had to demonstrate their fighting skills. IN ALL THIS a strict code of honor was enforced.

c. Av den kommende macho ble det krevd prover pa mot og oppvisningskamper. HER matte en strikt oereskodeks overholdes.

'Here a strict code of honor had to be kept.'

(21) a. Mein Problem ist es oft, nicht fragen zu konnen. DABEI bestehe ich fast nur aus Fragen. (PH1)

b. Inability to ask questions is often my problem. AND YET I'm made up almost entirely of questions.

c. Deter et problem for meg, dette at jeg ofte ikke er istand til a stille sporsmal. OG SA jeg, som i grunnen ikke har stort annet enn sporsmal i meg!

'It is a problem for me, the fact (lit. 'this') that I often am not able to ask questions. And then I, who in reality has not much else than questions in me.'

In (22), the presence of dabei ensures that the state described in the past tense relative clause is not temporally anchored in the immediately preceding main clause but rather in the independent sentence to the left of that. In the English target text the same effect is conveyed by other means, viz. the past perfect of the relative clause whereas the Norwegian target text uses both devices: past perfect combined with the anaphoric temporal connective da 'then, at that time' as an explicit counterpart of dabei.

(22) a. Die Mutter war in ihrer Jugend ofters auf einem Schlitten nach Rumanien hinubergefahren, sie zeigte mir die warmen Pelze, in die sie DABEI eingepackt war. (EC1)

b. In her youth, Mother had often ridden a sleigh all the way over to Rumania, she showed me the warm furs she HAD BEEN bundled in.

c. I sin ungdom hadde mor ofte kjort over til Romania reed slede, hun viste meg de varme pelsene hun hadde voert pakket inn i DA.

Example (23) (= 17), on the other hand, illustrates case (c): the German sentence would probably get the same (preferred) interpretation even without dabei; and both target texts leave the connective untranslated. (19) But as with redundant restitutive wieder, German native speakers would probably find the version without the connective less felicitous (cf. comments to [9] and [10], Section 3.3).

(23) a. Sind Sie schon einmal im Wald [...] ausgerutscht und haben DABEI durch die Laubschicht am Boden in einen vermoderten Baumstrunk gegriffen? (PH1)

'Have you ever lost your footing in the woods [...] and have dabei grabbed through the underbush into a rotting tree trunk?'

b. Did you ever lose your footing in the woods [...] and reach through the underbrush to grab a rotting tree trunk?

As far as target text dabei is concerned, case (b), that is, nonoptionality of target text dabei, can be illustrated by (24)-(25): dabei is necessary in order to establish the relation of simultaneity expressed by the phrasal connective and the preposition with in the source texts.

(24) a. In eating the plants we combine the carbohydrates with oxygen dissolved in our blood because of our penchant for breathing air, and so extract the energy that makes us go. IN THE PROCESS we exhale carbon dioxide, which the plants then recycle to make more carbohydrates. (CSA1)

b. Das heisst, wir essen die Pflanzen, um die Kohlenhydrate mit Sauerstoff zu verbrennen, der aufgrund unserer Veranlagung, Luft zu atmen, in unserem Blur gelost ist, und gewinnen so die erforderliche Betriebsenergie.

DABEI atmen wir Kohlendioxid aus, das die Pflanzen wiederum in einem Recycling-Prozess in Kohlenhydrate umwandeln. (CSA1TD)

(25) a. "Certainly," said Jack, WITH a small incline of the head. (ST1)

b. "Aber sicher," sagte Jack und neigte DABEI leicht den Kopf. '... said Jack und inclined dabei his head.'

(26) is another example of dabei blocking a sequential interpretation which might otherwise be natural. Here, the connective has no explicit counterpart in the source text; but it is triggered by the continuative "auxiliary" keep which has a comparable anchoring effect.

(26) a. The boy took the money in his left hand, scooping it up, shoveling it into a canvas bag strapped round his hips. He KEPT the gun, the to), gun, trained on Sharon Fraser. (RR1)

b. Der junge Typ nahm das Geld mit der Linken, schob es zu einem Haufchen zusammen und verstaute es in einer Leinentasche, die er um die Taille hangen hatte.

DABEI zielte er mit der Waffe, dem Spielzeugrevolver, auf Sharon Fraser.

'Dabei pointed he with the weapon ... at Sharon Fraser."

In (27), similarly, omitting dabei in the German translation would preferably be understood as describing an event sequence whereas the source text is more open to a nonsequential interpretation.

(27) a. "I'd like to hear more," she says quietly, leaning over to stub out her cigarette in the heavy ceramic ashtray on the floor. Her arm touches his knee; she quickly glances in his direction, but he doesn't seem to have noticed. (ABR1)

b. "Ich wurde gem davon horen," sagt sie ruhig und beugt sich vor, um ihre Zigarette in dem schweren Keramikaschenbecher auf dem Boden auszudrucken.

DABEI streift sie mit dem Arm sein Knie; sie blickt rasch zu ihm hin, doch scheint er es nicht bemerkt zu haben.

(28), finally, shows that dabei may also be quite redundantly added in translations; cf. (d) above.

(28) a. Righteously, mercilessly, he weeded out the passive voice. The effort of typing made the corners of his mouth turn down, so that no one could have guessed how much he was enjoying himself (AT1)

b. Gerecht und unbarmherzig merzte er das Passivum aus. Die Anstrengung des Tippens zog ihm die Mundwinkel herab, so dass kein Mensch vermutet hatte, wie gut er sich DABEI unterhielt.

As a marker of coherence at the lower, truth-conditional level (see Section 4.1), a main function of dabei is to prevent the narrative from "moving forward": the eventuality introduced in the modified clause or VP is presented as overlapping or at least abutting the antecedent eventuality rather than as completely following it. Thus, it blocks a sequential reading ("perfective viewpoint" according to Smith [1997]) in syndetic and asyndetic paratactical constructions where such a reading would otherwise be preferred due to the inherent aspectual properties ("aktionsart") of the modified VPs: (19) and (24) (25). In such contexts, the typical English target and, in particular, source construction seems to be an ingadjunct--not very surprisingly in view of its progressive aspect and normal discourse functions (Behrens 1998; Kortmann 1995). In fact, VP-coordination with dabei may be a kind of standard translational option expressing accompanying circumstance; cf. (29)-(32) in addition to the examples mentioned above. (20)

Norwegian has no progressive and no converb construction corresponding to the free ing-adjunct. But it also lacks a lexical equivalent of German dabei. As a consequence of that, we find a series of different Norwegian target constructions corresponding to ing-adjuncts in English and dabei-coordination in German: coordination with an explicit connective (29), subordinate clauses expressing simultaneity (30) (= [19]), free participial adjuncts (31), and prepositional adjuncts with reed 'with' + eventive noun; cf. (32).

(29) a. Ich habe dieselbe Ruckenhaltung, sagte ich mir. Unbeweglich verglich ich meinen Rucken mit dem Rucken meines Grossvaters und ich dachte DABEI an eine ganz bestimmte Fotografie, die nur ein Jahr vor dem Tod meines Grossvaters gemacht worden ist. (TBE1)

b. I have the same posture, ! told myself. Without moving I compared my own back with my grandfather's, THINKING of a particular photograph that had been taken only a year before his death.

c. Jeg har den samme ryggholdningen, sa jeg til meg selv. Ubevegelig sammenlignet jeg ryggen min med ryggen til min bestefar og jeg tenkte DA pa et ganske bestemt fotografi som var tatt bare et ar for min bestefars dod.

'... and I thought then/at that time of ...'

(30) a. He raises his hands in mock defence (beautiful, long fingers, she immediately notices, WONDERING whether he is conscious of it too). (ABR1)

b. In gespieltem Entsetzen hebt er die Hande (wunderschone, lange Finger, wie sie sofort bemerkt und sich DABEI uberlegt, ob er sich dessen gleichfalls bewusst ist).

'... as she immediately notices and dabei wonders whether ...'

c. Han lofter hendene til forsvar pa fleip (vakre, lange fingre, ser hun straks, MENS hun undres pa om han ogsd er klar over det).

'... she immediately notices, while she wonders whether ...'

(31) a. I went through a compound, came out at the housefront, and found him there, waiting.

He pursued me, RAVING in grotesque languages. (BO1)

b. Ich ging durch einen Compound, kam auf der Vorderseite der Hauser wieder heraus, und dort wartete er schon auf mich.

Er verfolgte mich UND phantasierte DABEI in grotesken Sprachen.

c. Jeg gikk gjennom en compound, kom ut pa forsiden av huset og sa ham sta der og vente.

Han forfulgte meg, BABLENDE pa groteske sprak.

'He pursued me, rave.PART.PRES, in grotesque languages.'

(32) a. She had even been thinking as she raced home that if Bert turned out to be one of the men that Jasper attached himself to, as had happened before, like a younger brother, SHOWING a hungry need that made her heart ache for him, then he wouldn't be off on his adventures. (DL2)

b. [...], dass Bert zu den Mannern gehorte, die Jasper wie ein jungerer Bruder verehrte [...] und DABEI soviel Hunger und Verlangen erkennen liess, dass es Alice seinetwegen schwer urns Herz wurde.

'... that Bert belonged to the men that Jasper admired like a younger brother und dabei showed so much hunger and need that ...'

c. [... ] at Bert var en mann av den typen Jasper kunne slutte seg til, som en slags lillebror og MED en hunger etter a bli godtatt som fikk hjertet hennes til a verke av medlidenhet.

'... that Bert was a man of the type Jasper might attach himself to, like a kind of younger brother and with a hunger for being accepted that...'

Quite often, however, the Norwegian target text has a simple coordinate structure without any connective in the second conjunct, inviting a sequential reading that is ruled out in the English and German parallel texts (33).

(33) a. He smiled slyly, NODDING. (WBI)

b. Er lachelte verstohlen und nickte DABEI.

c. Han smilte litt lurt og nikket.

'He smiled slyly and nodded.'

For the reasons mentioned above, one might suspect that Norwegian paratactic structures are more often undetermined between a sequential and a nonsequential reading than is the case in German; or, alternatively, that nonparatactic structures are preferred under conditions where a paratactic dabei-construction is natural in German. Conversely, since German has at its disposal a connective that explicitly blocks a sequential reading, its absence may have the effect of pushing the interpretation in the opposite direction, inviting a sequential reading when possible ("pragmatic strengthening," cf. Levinson 2000). This is a subject for further research, however.

5. Conclusion and outlook

As mentioned in the introduction, the present article does not pretend to be more than an explorative empirical study. There is, of course, a lot more to be done, empirically and theoretically; in fact, I have barely scratched the surface of the subject matter. But I hope to have made the following points.

As far as METHODOLOGY is concerned, corpus-based, multilingual translation comparison may function as an eye-opener in the study of connectives in general and elusive connectives in particular (cf. also Altenberg 1999, 2002; Fretheim and Johansson 2002; Saebo 2004).

As for EMPIRICAL RESULTS, our study has shown that one and the same anaphoric or presupposition triggering connective may be more or less indispensable vs. redundant, depending on whether or how easily the particular variety of discourse coherence correlated with its use in the given context may be inferred or implicated anyway. Thus, although the connective may not be strictly necessary in order to arrive at the interpretation triggered by its presence, it may be preferable because it eliminates ambiguity, reduces underterminacy, or prevents incoherence--or garden path effects--by blocking an interpretation that would otherwise be natural at the local (sentence) level but lead to incoherence at the (global) level of discourse interpretation.

We have also seen that language-specific properties in the domains of, for example, aspectual system, word order, and focus assignment may make a connective (more) indispensable in one language under conditions where its counterpart is (more) redundant in the other language.

Undoubtedly, these circumstances, in addition to the fact that dabei has no lexical counterpart in English, go a long way to explain the results obtained for dabei with respect to English (see Section 4.2). However, my findings indicate that this is not the whole story, that is, that structural contrasts and lexical asymmetry cannot alone account for the elusiveness of dabei and, in particular, wieder with respect to English and Norwegian. In addition, we may have to do with different stylistic preferences or different weighting of relevant constraints, including preferences or constraints of a prosodic nature that have not been taken into account here. Thus, simplifying very much, English and German would seem to adhere to the strategies described in (i) and (ii), respectively, as far as coherence in the temporal domain and the domain of eventualities is concerned; and perhaps (ii) has less weight in Norwegian than in German. (21)

(i) If the informational effect of using the connective is rather low, then don't use it. ("Be brief!")

(ii) If using the connective is more informative than not using it, then use it! ("Be precise!")

These two strategies can be subsumed under what has come to be known as the R- or I-principle (relevance, informativeness) and the Q-principle (quantity), respectively (Horn 1984; Blutner 2000):

(iii) R-/I-principle: Say no more than you must (given Q)! (iv) Q-principle: Say as much as you can (given R/I)!

The R-/I-principle is based on speaker's economy. It selects the most coherent or strongest interpretation compatible with what is a minimum of linguistic material, that is, in our case zero as opposed to an explicit connective. The Q-principle, on the other hand, "acts as a blocking mechanism and blocks all the outputs [interpretation possibilities] that can be derived more economically from an alternative linguistic input" (Blutner 2000: 198), the alternative to a connective being no connective (i.e. zero). That is, the Q-principle represents heater's economy--or speaker's need to convey his or her message fully: using a connective restricts the set of interpretations to be considered, thus making it easier for the hearer to arrive at the intended interpretation.

Summarizing, we conclude that a connective is redundant from the perspective of speaker's economy if the following conditions are fulfilled:

a. The interpretation [tau] assigned to the sentence containing the connective in the given context is a possible interpretation of the zero alternative in the given context as well.

b. The hearer will most likely assign r to the zero alternative anyhow, given the R/I-principle.

Whether or to what extent a connective is in fact redundant in a given context depends on the semantics of the connective itself, on the one hand, and properties of the intra- and extrasentential context it occurs in, on the other hand. As we have seen, ease of presupposition justification and anaphoric resolution play an important role in this respect: redundancy is negatively correlated with the amount of contextual enrichment (accommodation, bridging) needed in order to link the connective to the preceding context in an appropriate way (see Section 3.3).

But even when fairly redundant in view of the R-/I-principle, the connective will be well motivated in view of the Q-principle: by explicitly pointing backward, linking the relevant sentence to the preceding context, it will guide the hearer directly, without detour, towards the most coherent interpretation of the discourse processed thus far.

Our conclusion seems to be accordance with the relevance theoretic view on discourse markers--included connectives--referred to in the Introduction (Section 1). At the same time, however, the discussion above suggests that the bidirectional optimality theoretic setting outlined by Blutner (2000), by explicitly splitting the governing principle into two opposite principles that both conversational partners have to take into account, may represent a still more promising THEORETICAL FRAMEWORK for further research in this area.

Table 1. Wieder translated into ([right arrow]) and from ([left arrow])
English *

 Explicit [not equal to] Zero
 again again

1. [right arrow] En 98 (39.4%) 53 (21.2%) 98 (39.4%)
2. [left arrow] En 205 (33.8%) 193 (31.8%) 208 (34.4%)

 Total

1. [right arrow] En 249 (100%)
2. [left arrow] En 606 (100%)

* Excluding immer wieder 'over and over again,' hin und wieder 'once
and again'

Table 2. Again translated into ([right arrow]) and from ([left arrow])
German

 Explicit [not equal to] Zero
 again again

1. [right arrow] Ge 205 (59.4%) 106 (30.7%) 34 (9.9%)
2. [left arrow] Ge 98 (66.7%) 41 (27.9%) 8 (5.4%)

 Total

1. [right arrow] Ge 345 (100%)
2. [left arrow] Ge 147 (100%)

Table 3. Wieder translated into ([right arrow]) and from ([left arrow])
Norwegian

 Explicit [not equal to] Zero
 again again

1. [right arrow] No 156 (67.2%) 44 (16.6%) 64 (24.2%)
2. [left arrow] No 314 (64.3%) 92 (18.9%) 82 (16.8%)

 Total

1. [right arrow] Ge 264 (100%)
2. [left arrow] Ge 488 (100%)

Table 4. Igjen translated into ([right arrow]) and from ([left arrow])
German

 Explicit [not equal to] Zero
 wieder wieder

1. [right arrow] Ge 315 (63.4%) 138 (27.8%) 44 (8.8%)
2. [left arrow] Ge 174 (60.0%) 95 (32.2%) 26 (8.8%)

 Total

1. [right arrow] Ge 497 (100%)
2. [left arrow] Ge 295 (100%)

Table 5. Again translated into ([right arrow]) and from ([left arrow])
Norwegian

 Explicit [not equal to] Zero
 igjen igjen

1. [right arrow] No 225 (65.2%) 91 (26.4%) 29 (8.4%)
2. [left arrow] No 300 (76.3%) 66 (16.8%) 27 (6.9%)

 Total

1. [right arrow] No 345 (100%)
2. [left arrow] No 393 (100%)

Table 6. Igjen translated into ([right arrow]) and from ([left arrow])
English

 Explicit [not equal to] Zero
 again again

1. [right arrow] En 301 (60.6%) 154 (31.2%) 42 (8.2%)
2. [left arrow] En 246 (42%) 253 (53.6%) 86 (12.7%)

 Total

1. [right arrow] En 497 (100%)
2. [left arrow] En 716 (100%)

Table 7. Zero ratios for wieder, again, and igjen

Source-text occurrence Zero ratio (%)

igjen [right arrow] En 8.2
again [right arrow] No 8.4
igjen [right arrow] Ge 8.8
again [right arrow] Ge 9.9
wieder [right arrow] No 24.2
wieder [right arrow] En 39.4

Target-text occurrence Zero ratio (%)

again [left arrow] Ge 5.4
again [left arrow] No 6.9
igjen [left arrow] Ge 8.8
igjen [left arrow] En 12.7
wieder [left arrow] No 16.8
wieder [left arrow] En 34.4

Table 8. Wieder [right arrow] English and Norwegian

 [right arrow]
Source text wieder English

 Explicit Zero

[right arrow] Norwegian Explicit 136 55
 Zero 15 43
English Total 151 98

Source text wieder Norwegian
 Total

[right arrow] Norwegian Explicit 191
 Zero 58
English Total 249

Table 9. English and Norwegian translation images of dabei

Category Examples

Simple clause-initial (and) yet, but, however men 'but',
connective / thus allikevel 'however'
conjunction moreover selv om, skjont
([not equal to] 'although'
and/og)

Spatial connective here her 'here'
adverb (clause
initial or
noninitial)
Temporal connective na 'now', da 'then'
adverb (clause samtidig 'at the same
initial or time'
noninitial)
Instrumental dermed, derved 'with
connective adverb that' derigjennom
 'through that'
PP containing an in (all) this, at i denne forbindelse
abstract anaphor or this activity, 'in this connection'
definite description in this situation, under denne processen
 in the process 'under this process'
 i den anledning 'on
 this occasion'
Prepositional (in/by) in doing so, in so [No structural equi-
ing-construction doing, by doing so valent to ing-cons-
with anaphoric tructions in
predicate Norwegian]
Prepositional (in/by) in hurrying to the end
ing-construction with of the bridge
"full" predicate
Temporal clause (as/ as I did so, when he mens sa skjer 'while
when/while) with did it, while he is at so happens'
anaphoric predicate it
Temporal clause (as/ and as I breathed mens unggutten la ut
while) with "full" while the boy was 'while the boy was
predicate talking talking'

Table 10. Dabei translated into ([right arrow]) and from ([let arrow])
English

 Explicit Zero Total

1. [right arrow] En 52 (47.67%) 57 (52..%) 109 (100%)
2. [left arrow] En 31 (28.2%) 79 (71.8%) 110 (100%)

Table 11. Dabei translated into ([right arrow]) and from ([let arrow])
Norwegian

 Explicit Zero Total

1. [right arrow] No 75 (62.5%) 45 (37.5%) 120 (100%)
2. [left arrow] No 19 (45.2%) 23 (54.8%) 42 (100%)

Table 12. Dabei translated into English and Norwegian

 [right arrow]
Source text dabei English

 Explicit Zero

[right arrow] Vorwegian Explicit 38 29
 Zero 14 27
English Total 52 56

Source text wieder Norwegian
 Total

[right arrow] Vorwegian Explicit 67
 Zero 41
English Total 108

Notes

(1.) This article is based partly on joint work with Bergljot Behrens (see Fabricius-Hansen and Behrens 2001). In addition, I have profited from discussions with Kjell Johan Saebo and Torgrim Solstad. I also thank two anonymous referees for useful recommendations. Correspondence address: Gennanistisk institutt, Postboks 1004, Blindern, N-0315 Oslo 3, Norway. E-mail: c.f.hansen@german.uio.no.

(2.) See [http://www.hf.uio.no/german/sprik/english] and Johansson (1998) for details concerning corpus design.

(3.) Konig et al. (1990).

(4.) Fretheim (i.p.).

(5.) Unless otherwise indicated, the examples below are taken from the OMC (see Section 1). In each case, the original version is given under (a). The source text is identified by the abbreviation used in the OMC.

(6.) See Fabricius-Hansen (2001), Klein (2001), Jager and Blutner (2003), Pittner (2003), and von Stechow (2003) for recent discussions: cf. also Bierwisch (2003).

(7.) Or procedural in relevance theoretic terms; cf. Fretheim (i.p.).

(8.) Huitink (2003) assigns the precedence condition on the presupposed and asserted instantiations of "this" to the assertion rather than the presupposition; and Saebo (2004) argues that additive particles, when associated with a topic, add the presupposed alternative to the topic of the assertion, creating an "aggregate topic" (see also Reis and Rosengren 1997). In either case, however, the information bit that is added to the assertion is already in the context, given an anaphoric account of presuppositions. Thus it seems redundant from a truth-functional point of view. But being part of the assertion, as opposed to something that merely has to be justified in the preceding context, the added bit of information can influence the conversational implicatures to be drawn from the sentence in the given context (Saebo 2004)--as suggested by Blakemore (cf. Section 1).

(9.) See Konig et al. (1990) and Fretheim (i.p.), in addition to the titles mentioned in Note 6. As pointed out by Fretheim (i.p., Sections 7 and 8), there are two other uses of igjen, with the meaning of 'shut' and 'left (over)' (German ubrig), that are neither repetitive nor restitutive. These variants seem compatible with the more general semantic picture of wieder/again presented in Fabricius-Hansen (2001).

(10.) Altenberg (1999: 254) defines the mutual correspondence (MC) of items in two languages as the frequency with which they are translated into each other, to be calculated by means of the formula (([A.sub.t] + [B.sub.t]) x 100) : ([A.sub.s] + [B.sub.s]), where [A.sub.t], [B.sub.t] and [A.sub.s], [B.sub.s] are the compared items in the target and source texts, respectively. In the trilingual corpora of the OMC (see Section 1), the MC is only approximately 50% for wieder compared to again, and around 62% for the two other pairs (wieder: ig]en and again:igjen).

(11.) The prefix re-, however, is not exclusively constitutive (Fretheim i.p.).

(12.) The Norwegian (source- and target-text) data include igjen-variants that do not fit into the restitutive-repetitive pattern (see Note 9). They have all been registered as having nonzero counterparts. If they are excluded from the data, zero ratios involving igjen will go up somewhat since the registered number of zero correspondencies will represent a higher percentage of the total number of igjen tokens.

(13.) The zero ratio for igjen [left arrow] English is higher than 10% but in this case the result is not corroborated by igjen [right arrow] English (8.2%), indicating that "translationese" may have played a role here.

(14.) Cf. Dimroth (forthcoming), Saebo (2004).

(15.) In my data, there are 21 cases of target text wieder corresponding to another in the source text; cf. (5).

(16.) See Krause (2002: 35f.) for exceptions.

(17.) See also Fabricius-Hansen (1999).

(18.) Interestingly, free ing-adjuncts represent only 10%, of the zero correspondencies in English target texts; cf. (18).

(19.) (16) may seem a similar case: but different from the source text, the target texts are both open for a sequential interpretation of the described eventualities.

(20.) As mentioned in Section 4.2, I have registered ing-adjuncts as zero correspondencies unless they contain some lexical or phrasal translational image of dabei. If we subtract the 25 ing-sources from the 79 zero correspondencies in Table 10, we are left with 54 (= 49%), that is, less than in Norwegian in translations from Norwegian (53.4%, cf. Table 11).

(21.) In the end, of course, this means that social conventions differ somewhat between the three language communities.

References

Altenberg, Bengt (1999). Adverbial connectors in English and Swedish. In Out of Corpora. Studies in Honour of Stig Johansson, H. Hasselgard and S. Oksefjell (eds.), 249-268. Amsterdam: Rodopi.

--(2002). Concessive connectors in English and Swedish discourse. In Information Structure in a Cross-linguistic Perspective, H. Hasselgard et al. (eds.), 21-44. Amsterdam: Rodopi. Asher, Nicholas (1993). Reference to Abstract Object in Discourse. Dordrecht: Kluwer.

Behrens, Bergljot (1998). Contrastive discourse: an interlingual approach to the interpretation and translation of free ING-participial adjuncts. Unpublished doctoral dissertation, University of Oslo.

Bierwisch, Manfred (2003). Heads, complements, adjuncts: projection and saturation. In Modifying Adjuncts, E. Lang et al. (eds.), 113-159. Berlin and New York: Mouton de Gruyter.

Blakemore, Diane (1992). Understanding Utterances. Oxford: Blackwell.

Blutner, Reinhard (2000). Some Aspects of Optimality in Natural Language Interpretation. Journal of Semantics 17, 189-216.

Dimroth, Christine (forthcoming). Fokuspartikeln und Informationsgliederung im Deutschen. Studien zur deutschen Grammatik. Tubingen: Stauffenburg.

Doherty, Monika (ed.) (1999). Sprachspezifische Aspekte der Informationsverteilung. Berlin: Akademie-Verlag.

--(2002). Language Processing in Discourse. London: Routledge.

--(2003). Topikalisierungsstrategien aus der Perspektive diskursadaquater Ubersetzungen. Linguistische Berichte 194, 183-212.

Fabricius-Hansen, Cathrine (1999). Bei dieser Gelegenheit--on this occasion--ved denne anledningen. German bei--a puzzle in translational perspective. In Out of Corpora. Studies in Honour of Stig Johansson, H. Hasselgard and S. Oksefjell (eds.), 231-248. Amsterdam: Rodopi.

--(2001). Wi(e)der and Again(st). In Atuliatur Vox Sapientiae. A Festschrift for Arnim von Stechow, Studia grammatica 52, C. Fery and W. Sternefeld (eds.), 101-130. Berlin: Akademie-Verlag.

--; and Behrens, Bergljot (2001). Elaboration and related discourse relations viewed from an interlingual perspective. In Proceedings from the Workshop on Text Structure, University of Austin, October 2000, C. Smith (ed.). Also as SPRIKreport Nr. 13 [http://www. hf.uio.no/german/sprik/publikasjoner.shtml].

Fretheim, Thorstein (i.p.). A relevance-theoretic account of the way we use and understand the English temporal adverb again and its Norwegian counterpart igjen. Languages in Contrast.

--; and Johansson, Stig (2002). The semantics and pragmatics of the Norwegian concessive marker likevel: Evidence from the English-Norwegian Parallel Corpus. In From the COLT's Mouth ... and Others'. Language Corpora Studies in Honor of Anna-Brita Stenstrom, L. E. Breivik and A. Hasselgren (eds.), 81-101. Amsterdam: Rodopi.

Haspelmath, Martin; and Konig, Ekkehard (eds.) (1995). Converbs in Cross-Linguistic Perspective. Berlin and New York: Mouton de Gruyter.

Hasselgard, Hilde; Johansson, Stig; Behrens, Bergljot; and Fabricius-Hansen, Cathrine (2002). Information Structure in a Cross-linguistic Perspective. Amsterdam: Rodopi.

Horn, Lawrence R. (1984). Toward a new taxonomy for pragmatic inference. Q-based and R-based implicature. In Meaning, Form, and Use in Context. Linguistic Applications, D. Schiffrin (ed.), 11-42. Washington, DC: Georgetown University Press.

Huitink, Janneke (2003). We cannot have pizza again. Paper presented at the ESSLLI 2003. The meaning and implementation of discourse particles, http://www.ling.uni-potsdam.de/~stede/ESSLLIpapers/huitnik-slides.pdf

Jager, Gerhard; and Blutner, Reinhard (2003). Competition and interpretation: the German adverb wieder ('again'). In Modifying Adjuncts, E. Lang, C. Maienborn, and C. Fabricius-Hansen (eds.), 393 416. Berlin and New York: Mouton de Gruyter.

Johansson, Stig (1998). On the role of corpora in cross-linguistic research. In Corpora and Cross-linguistic Research' Theory, Method, and Case Studies, S. Johansson and S. Oksefjell (eds.), 3-24. Amsterdam: Rodopi.

Klein, Wolfgang (2001). Time and again. In Audiatur Vox Sapientiae. A Festschrift for Arnim von Stechow, Studia grammatica 52, C. Fery and W. Sternefeld (eds.), 267-286. Berlin: Akademie-Verlag.

Kortmann, Bernd (1995). Adverbial participial clauses in English. In Converbs in CrossLinguistic Perspective, M. Haspelmath and E. Konig (eds.), 189-238. Berlin and New York: Mouton de Gruyter.

Konig, Ekkehard (1995). The meaning of converb constructions. In Converbs in Cross-Linguistic Perspective, M. Haspelmath and E. Konig (eds.), 57-96. Berlin and New York: Mouton de Gruyter.

--; Stark, Detlef; and Requardt, Susanne (1990). Adverbien und Partikeln. Ein deutschenglisches Worterbuch. Heidelberg: Julius Groos.

Krause, Maxi (2002). BEI preposition, element constitutif de particules diverse et particule verbale. Cahier du CRISCO (Centre de Recherches Interlangues sur la Signification en Contexte UMR 6170) no 10 Section 3. Universite de Caen.

Lehmann, Christian (1988). Towards a typology of clause linkage. In Clause Combining in Grammar and Discourse, Typological Studies in Language 18, J. Haiman and S. A. Thompson (eds.), 181-226. Amsterdam: John Benjamins.

Levinson, Stephen C. (2000). Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press.

Pittner, Karin (2003). Process, eventuality, and wieder/again. In Modifying Adjuncts, E. Lang et al. (eds.), 365-391. Berlin and New York: Mouton de Gruyter.

Reis, Marga; and Rosengren, Inger (1997). A modular approach to the grammar of additive particles: the case of German auth. Journal of Semantics 14, 237-.309.

Smith, Carlota S. (1997). The Parameter of Aspect. 2nd ed. Dordrecht: Kluwer.

Saebo, Kjell Johan (1997). Handout from paper presented at the 19th Jahrestagung der Deutschen Gesellschaft fur Sprachwissenschaft, AG 12. University of Oslo.

(2004). Conversational contrast and conventional parallel: topic implicatures and additive presuppositions. Journal of Semantics (Special Issue on Contrast in Discourse) 21(2), 199-217.

von Stechow, Arnim (2003). How are results represented and modified? Remarks on Jager and Blutner's anti-decomposition. In Modifying Adjuncts, E. Lang et al. (eds.), 417-451. Berlin and New York: Mouton de Gruyter.

von Stutterheim, Christiane (1997). Einige Prinzipien des Textaulbaus. Empirische Untersuchungen zur Produktion mundlicher Texte. Tubingen: Niemeyer.

Thompson, Sandra A.; and Longacre, Robert E. (1985). Adverbial clauses. In Language Typology and Syntactic Description. Vol. 2." Complex Constructions, T. Shopen (ed.), 171-234. Cambridge: Cambridge University Press.

Zeevat, Henk (2003). Particles: presuposition triggers or context markers. Draft. [http:// cf.hum.uva.nl/computerlinguistiek/henk/henkPUB.html].

Oslo University

Received 27 March 2003

Revised version received

22 September 2003