Navigation – Plan du site


Cet article étudie la structure du discours du point de vue des changements de topiques, et plus précisément en se concentrant sur l’adverbe autrement et son rôle central dans une telle opération. Ce marqueur nécessite un antécédent, le topique qu’il ferme, qu’il trouve généralement parmi les éléments saillants du discours tels que les groupes nominaux topicalisés et les cadratifs. Cependant, comme il peut y avoir plusieurs éléments de cette sorte dans le contexte gauche, d’autres procédés sont aussi employés pour s’assurer que les interlocuteurs comprennent clairement quel topique est fermé, c’est-à-dire quel élément est l’antécédent d’autrement. Ces procédés sont essentiellement l’introduction du nouveau topique, similaire à bien des égards (syntaxe, sémantique) au précédent; la répétition d’un topique d’ordre supérieur est aussi un moyen important pour clarifier la structure du discours. Dans les cas où il n’existe pas de tels marqueurs, on fait l’hypothèse que les topiques existent au moins implicitement. D’une manière générale, la structure du discours, et particulièrement le contexte gauche, doivent être fortement organisés pour permettre à autrement de fonctionner correctement. Par conséquent, l’étude d’autrement offre un point de vue intéressant sur la structure du discours prise comme un tout, y compris sur la contrainte de la Frontière Droite: non seulement autrement trouve-t-il son antécédent sur la Frontière Droite, mais qui plus est il contribue à la façonner en fermant des topiques et en modifiant la disponibilité des référents de discours.

Haut de page

Texte intégral

1. Introduction

1If we define quite bluntly discourse topics as what a discourse segment is about, then discourse topics are obviously vital to the correct understanding of any extended piece of speech. Investigating what they really are and how they emerge from discourse is an important part of discourse theory, however there is one fact that has been less often studied, namely that speakers change topics. Indeed, conversations, provided they are long enough, never address one and only one subject, but instead regularly turn to different topics or split big ones into smaller parts, which themselves are topics, albeit subordinate ones.

2This seemingly natural process requires extensive use of linguistic devices to help interlocutors keep track of what is being said. In this paper I focus on one such device, the French connective autrement, but I will show that it relies on discourse structure on a larger scale and requires that the left context be properly organized. In other terms, although the marker is crucial, it still needs discourse structure as a whole to work correctly. The latter in turn is shaped by topics and topic shifts, and autrement sheds an interesting light on its mechanisms.

3This study is based on seven oral corpora (see Isambert, 2010, for details), out of which 179 occurrences of autrement in the use investigated here have been extracted and annotated. Sections 5 and 6 emphasize the results pertaining to anaphora resolution. Here is what was annotated; not all categories are relevant for the use investigated here:

  • Type of the antecedent;
  • Position of autrement;
  • Type of the predicate modified by autrement;
  • Tense (of the proposition hosting the antecedent and of the proposition hosting autrement);
  • Use of negation in the host proposition;
  • Scope of the adverb;
  • Use of a conjunction before the adverb (and type, if any);
  • Discourse relation between the propositions conjoined by autrement;
  • Markers closing the previous topic;
  • Markers introducing the new topic.

2. Three uses of autrement

4The French adverb autrement has three main uses, which constitute three steps on a grammaticalization path (Isambert, 2010). First, it is an adverb of manner modifying a verb phrase:

[1] Nous avons tenté pendant longtemps de faire cette explication par des voies démocratiques, pacifiques. Malheureusement, ça n’a pas marché et certains Ivoiriens ont décidé de s’expliquer autrement.
  ‘We’ve been trying for a long time to give that explanation in a democratic and pacific way. Unfortunately, that didn’t work and some Ivorians have decided to explain themselves in another way.’

5In this use autrement takes an antecedent from which it derives a different “manner”. In [1] the antecedent is par des voies démocratiques, pacifiques and autrement thus means par d’autres voies, that is in a way that is neither democratic nor pacific.

6The second use, the most investigated by scholars (Inkova-Manzotti, 2002; Lamiroy, Charolles, 2005; Isambert, 2006, among others) is as a connective:

[2] Ne viole pas la règle, autrement tu seras viré.
  ‘Don’t break the rule, otherwise you’ll be fired.’

7Here autrement denotes “negative hypothesis”, i.e. it makes the assumption that a previously stated condition (generally in the form of a deontic proposition) is not fulfilled; the proposition that follows states what would ensue. This connective is often used to justify the antecedent, either proving a point or stressing the desirability of a state of affair (in [2] for instance, the interlocutor should see to it that she does not break the rule).

8Finally, autrement can also be used to shift topics of conversation, either globally, as in [3], where the proposition could follow anything (hence the ellipsis), or more specifically, as in [4], where a sub-topic (soccer) is introduced under the heading of a larger one (sports).

[3] […] Autrement, ça va?
  ‘[…] Apart from that, how are you doing?’
[4] Je joue beaucoup au tennis. Autrement, j’aime bien le foot.
  ‘I play tennis a lot. Apart from that, I like soccer.’

9Despite many differences, the three uses share some central properties. First, autrement is always an anaphor, and thus requires an antecedent to make sense; anaphor resolution is no simple task, and it is especially complex with autrement, due to its ability to derive a representation from many kinds of contexts (Isambert, 2010). Second, autrement builds a representation from its antecedent and does not simply carry it on; in other terms, it triggers an accommodation (Lewis, 1979) which would require substantial morpho-syntactical modifications of the antecedent if such a representation were to be uttered (at least an inverted polarity, and generally the transformation of an independent clause into an if-clause, as in [2], where autrement turns don’t break the rule into if you break the rule).

10These properties allow us to reconstruct the grammaticalization of autrement as a classical transformation of an intra-sentential adverb into a discourse organizer, where semantic bleaching is counterbalanced by new, metalinguistic functions (Lamiroy, Charolles, 2004), tending to subjectification (Traugott, 1995). In this paper, I will focus on the third use, where the antecedent is a topic and autrement organizes the flow of discourse.

3. Anatomy of an example

11The following example is taken from an oral corpus (CLAPI); it is an extract from an interview between a linguist (L) and a priest (P). Numbering is set in order to stress the structure of the dialogue; barring the middle ellipsis, the example is one continuous piece of dialogue.

a L: Oui bon alors on va parler de la profession, alors, qu’est-ce que vous pouvez me dire justement, un petit peu comme ça librement, tout ce qui vous intéresse sur votre… sur votre profession là, sur la vôtre.
  L: Well, we’re going to talk about your job, so, what can you tell me, just like that, freely, everything you find interesting in your job?’
b P: Oui oh ben disons beaucoup de choses bien sûr, tout est intéressant dans la mesure où on a affaire à des gens qui sont intéressants.
  P: Well, let’s say, many things of course, everything is interesting as long as we’re dealing with interesting people.’
c [Recounts example: 3min+]
  [Recounts example: 3min+]
d On est pris un peu comme un syndicat d’initiative, distrayez-nous pendant un petit moment.
  ‘People tend to take us for a tourist office: entertain us for a little while!’
e Voilà! [Pause: 2s]
  ‘That’s it! [Pause: 2s]
f Autrement, les choses intéressantes c’est : pour tous ceux qui essaient de faire quelque chose, je vois, on essaye de, aujourd’hui de mettre en place ce qu’on appelle une sorte de concile ; on est en train de faire une espèce de concile local.
  ‘Apart from that, interesting things are: for those who try to do something, I can see, we’re trying to set up a kind of council; we’re doing a kind of local council.’
  (Corpus CLAPI)

12The topic of this extract is explicitly stated by the interviewer in [5a] as the interesting things in the interviewee’s job, repeated by the priest in [5b] and reintroduced in [5f] (les choses intéressantes). Thus this topic is clearly negotiated and acknowledged. It leads to a detailed narration (skipped over in [5c]), more than three minutes in length. The duration is important: an extended discourse segment occurs, not just a pair of propositions. [5d] is actually part of the example recounted in [5c]. I have singled it out here only because it clearly summarizes the entire narration (about people marrying in the church without real faith) by generalizing over a single case. Similarly, voilà in [5e] is an important marker that ends the discourse segment, backed up by a silence. Finally, autrement in [5f] definitely switches to a new sub-topic, the local council; most importantly, the main topic, the interesting things, is mentioned again, so that the topic switch is properly located.

13This example illustrates the use of autrement to organize discourse (Isambert, 2008a). However, discourse management does not rely on autrement exclusively, but on a set of devices instead (stating the main topic, introducing sub-topics, closing them); those markers, barring the repetition of the main topic, appear in the left context, and display a striking diversity, as we will see presently when we try to answer the question asked in the title of this paper. But first we shall take a look at a special instance of this use of autrement in order to exclude it from this study.

4. A special case: “short-sighted” autrement

14Although clearly an instance of the third use of autrement in many respects (Lamiroy, Charolles, 2005; Isambert, 2010) and not of its second use (as a connective) as stated by some authors (Inkova-Manzotti, 2002), “short-sighted” autrement, illustrated in [6] (two people talking about farmers), displays properties that do not line up very well with the notion of topic shift investigated here.

[6] A: Ils font pas de fruits.
B: Des fraises.
A: Ah si que les fraises oui, autrement ils font pas de fruits.
  A: They don’t do fruits.
B: Strawberries.
A: Oh yes only strawberries yes, apart from that they don’t do fruit.’
  (Corpus de la Parole)

15The question raised by this use, called “autrement de complémentarité” by the authors just mentioned, is whether this use is topical or not. Admittedly, there is a main topic (fruits, itself a sub-topic in the conversation), an instance of which is singled out in a sub-topic (strawberries) only to be discarded by autrement. Although this might be considered discourse structure on a small scale, the absence of other markers to organize this “micro-segment” casts doubt on the nature of such examples. Moreover, there is an additional meaning attached to autrement here: one does not just switch topics, one also excludes an element (strawberries) from an ensemble, or deems it irrelevant to a situation, thus leading in [6] to an apparent contradiction: the farmers grow strawberries without doing fruits. Autrement here is meant to accommodate such a contradiction. This implies that a specific discourse relation holds between the two propositions (they do strawberries and they don’t do fruits), namely contrast or opposition, whereas pure topic shift is meant to acknowledge and negotiate the absence of a relation between discourse segments.

16Other examples, however, do not depart so clearly from the prototypical use of autrement as a topic shift:

[7] A: Est-ce que vous sauriez me dire comment on fait une omelette chez vous?
B: En cassant les œufs, ça c’est sûr. Mais autrement, la cuisine ce n’est pas mon rayon.
  A: Could you tell me how one makes an omelette at your place?
B: You break eggs, for sure. But apart from that, I don’t know much about cooking.’
  (Corpus de la Parole)

17[7] displays mixed properties, functioning semantically as described above, yet retaining characteristics of the topic shift, for instance the mention of the (implicit) main topic “cooking”. Its prefixing with mais must be taken into account, since mais is a hallmark of “short-sighted” autrement which does not occur easily with the topical use.

18This example is evidence that the distinction between the topical use and the “short-sighted” use of autrement is not clear-cut and that they have overlapping properties; it should be noticed that the latter use feeds on lists or sets of elements (e.g. strawberries are members of the set of fruits), and as we will see presently, lists are a major topical organization of discourse; on the other hand, as has been said above, “short-sighted” autrement could also qualify as a small-scale discourse marker (hence the name I have given it). For the rest of this paper I will focus on clear cases of topical autrement, but one should keep in mind that they are only prototypical instances in a continuum of uses from the pure connective to the pure discourse marker, a conclusion in line with the general theory of grammaticalization: the topic shift has evolved out of the connective but intermediate uses still exist (Isambert, 2010).

5. Antecedents

19In all its uses, the antecedents to autrement are quite diverse, and the topic shift is no exception (one may also consider that there is only one type of antecedents, namely discourse topics, and that they are quite diverse in their realization; the distinction is immaterial here). In this section I will review the most usual ones; unsurprisingly, all rely on discourse structure. Here are the kinds of antecedent found in corpora:

  • Topicalized noun phrases (outside lists): 12%;
  • Enumerated items: 40%;
  • Discourse segments (without explicit topic): 30%;
  • Frames: 8%;
  • Others: 10%.

20How such antecedents are resolved is investigated in the next section.

5.1. Topicalized noun phrases

21Quite often (12% of the occurrences), the antecedent to autrement is a topicalized noun phrase, where “topicalized” here is meant to denote a syntactic property, although semantics obviously follows suit. The property at stake is what Jacobs (2001) calls “informational separation”, where the topic is first introduced, and then the predication. In French, informational separation is generally marked by left dislocation (Lambrecht, 1994), as illustrated by [8], where the antecedents to autrement is les paramètres:

[8] Donc les paramètres, c’est heu c’est plus général sur le heu le logiciel dedans y a heu la promotion la les dates de stage, parce que les les dates de stages sont connues pour tous les étudiants mais elles changent tous tous les ans quoi en fait en fonction du calendrier, le heu dernier numéro du contrôle continu d’examen, T.P. de la matière enseignant, ça c’est renseigné automatiquement chaque fois qu’on fait la saisie dans le logiciel, mais on peut les modifier si on a besoin d’intervenir sur quelque chose. Voilà, donc qu’est-ce qu’y a d’autre autrement, heu gestion des mots de passe, donc on peut créer un autre utilisateur si on veut.
  ‘So parameters, that’s more general, in the software there’s the class, training dates, because training dates are known for all students but they change each year depending on schedule, the last number of the exam test, practical work for teaching subject, this one is filled automatically every time you give data to the software, but you can change them if you have to modify something. That’s it, so what else apart from that, well password management, so you can create a new user if you want.’
  (Corpus PIC)

22The phrase les paramètres is clearly topicalized: it appears on the left of its host sentence, in which it does not play a direct syntactic role. It plays the role of a local topic, detailed in the discourse that follows, and is itself a sub-topic to the more general subject of conversation (computers). Autrement targets it and switches to another topic, namely passwords; the discourse on parameters has been previously closed with voilà, and autrement itself is backed up by qu’est-ce qu’il y a d’autre, where one cannot miss the marker autre, the morphological root of autrement.

5.2. Enumerated items

23Lists are another structure where autrement occurs very frequently (up to 40% of the examples); they are not incompatible with topicalized elements seen in the previous section, quite the contrary: lists often involve topicalization, and topicalized elements often constitute implicit lists. The following example is taken from the same interview as [5] above.

[9] L: Justement quand vous avez le temps quelles émissions de télévision vous regardez ? Est-ce qu’il y en a que vous regardez habituellement ?
P: Non pratiquement, le film des fois le vendredi soir quand j’ai le temps d’en prendre un morceau ou le dimanche soir des fois, et j’aime pas quand ils sont sous-titrés. Alors c’est celles-là que je regarde le plus. Autrement les informations quand je peux avoir soit les informations euh de onze heures ça dépend à quelle chaîne ou ça dépend à quel moment on est là, et puis autrement j’ai pas une émission précise précis non.
  L: Precisely, when you have time, what kind of programs do you watch? Are there some you usually watch?
P: Not really, movies sometimes on Friday nights when I have time to catch part of them or on Sunday nights sometimes, and I don’t like subtitles. So that’s what I watch most. Apart from that the news, when I can get news at eleven, it depends on the channel or it depends on the time you’re home, and then apart from that I don’t have a particular program, no.’
  (Corpus CLAPI)

24In this example, the list as a structural device is clearly elicited by the interviewer’s question about TV programs; the interviewee is thus led to review kinds of programs in turn, and first mentions movies and then news, articulating the two with autrement. List items are generally short in content, and do not require additional markers (like voilà) to delimit sub-topics: the structure of the list is ensured by syntactic relations between the items and the overarching topic; namely le film and les informations answer the question quelles émissions de télévision vous regardez? (what programs do you watch?) and are objects to an implicit predicate je regarde (I watch…) derived from the question. In both cases, the list item starts as sentence fragments (an object and a subordinate clause headed by quand) and not as a full sentence. Thus, the relation of each item to the list as a whole is preserved, and autrement is simply used as a minimal articulation. The last sentence of the example is also headed by autrement, which is halfway between a short-sighted use (the speaker has just mentioned particular programs, thus stating that he does not watch anything particular is a possible contradiction) and a topic shift use (considering other kinds of program, the speaker asserts that he does not watch anything).

5.3. Frames

25Frames are major devices in discourse structure. They often consist in adverbial or prepositional phrases setting the spatial and/or temporal circumstances of what is to come (they are generally topicalized) and delimitating thematic units (Charolles, 2003, 2005). Like topicalized noun phrases, they are easy targets for autrement, thanks to their visibility and role.

[10] À vingt ans mon dieu c’était la guerre, la guerre de Quatorze, c’était bien terrible tout ça. Je me rappelle que j’étais j’avais vingt-deux ans j’étais à Paris et la Bertha nous tombait sur la tête vous savez la euh comment je veux dire ce que les z- les Allemands nous envoyaient vous avez pas jamais entendu parler de la guerre de de quator- si mais la Bertha c’est Paris. Alors on allait dans les caves naturellement j’étais chez mes patrons, ils étaient bien gentils du reste et fallait c’était la guerre fallait aller courir au charbon rue de la gare du Nord moi j’étais rue Richelieu presque au Palais Royal alors fallait courir chercher des kilos de charbon à la gare du Nord sous son bras. On était gelés gelés gelés ça c’est ma vie à Paris vous savez c’était les bombardements c’était tout ça. Oui ça là-dessus j’ai regretté bien des choses parce que je me plaisais à Paris j’avais des patrons très gentils tout ça c’était des Juifs pourtant mais enfin ils étaient très bons pour moi question de ça autrement avant de avant cette guerre j’étais domestique dans les campagnes euh enfin dans mon au tous alentour de mon pays quoi.
  ‘When I was twenty my god it was war, the 1914-1918 war, that was terrible. I remember I was twenty-two, I was in Paris and the Bertha was falling on our heads, you know I mean, what the Germans sent us, you’ve never heard about 1914 war, yes but Bertha, it was Paris. So we went to the cellar, naturally, I was at my bosses’, they were very nice by the way, and it was war, we had to run after coal, rue de la Gare du Nord, I was rue Richelieu almost Palais Royal so we had to run after kilos of coal at the Gare du Nord under your arm. We were freezing, that’s my life in Paris, you know it was bombing and all. Yes on this point I’ve missed many things because I liked it in Paris, I had very nice bosses, they were Jews nonetheless but still they were very kind to me. Apart from that before that war I was a servant in the country, I mean around my land.’
  (Corpus de la Parole)

26Here the antecedent is à vingt ans and it heads the discourse that follows up to autrement, which is used to switch to another temporal setting (before the war), and thus introduces a new discourse segment with its own thematic unity. The length of the first segment and the absence of any closing device is remarkable: the discourse relies only on frames articulated by autrement to be segmentally structured, which shows how powerful such markers can be.

5.4. Discourse segments

27Discourse segments are organized sequences of propositions, linked by discourse relations (Asher, Lascarides, 2003) and elaborating on a common topic (Asher, 2004a). They frequently constitute antecedents of anaphora (Consten, Knees & Schwarz-Friesel, 2007), and autrement is no exception (30% of the examples). Note that the antecedents reviewed in the previous sections also involve discourse segments, but explicit markers (the antecedents proper) could be targeted by autrement; here no such devices are used. This is illustrated by the following example:

[11] Et puis après j’ai travaillé dans un bureau dans lequel se trouvait mon mari, c’était pas très pas très autorisé parce que il était plus, il était gradé malgré tout alors euh y avait des jalousies y avait des, vous savez ce qui se passe dans dans les bureaux c’est souvent des des jalousies et des méchancetés, j’en ai j’en ai encaissé pas mal, ouais alors je avec ça j’étais pas mal souffrante et j’avais j’avais hâte vraiment de lâcher. Ben autrement c’était pas désagréable de de remplir ces fiches de de s’occuper du personnel.
  ‘And then after that I worked in an office where my husband was, it was not really allowed because he was somewhat highly ranked anyway, so there was envy, you know how it goes in offices, there often is envy and meanness, I took a lot, besides I was unwell and I was really eager to quit. Well apart from that it wasn’t bad to fill those cards and be in charge of employees.’
  (Corpus de la Parole)

28Here the antecedent is the entire discourse in bold. The main topic is the speaker’s job at her husband’s office, and a sub-topic is how hard it was, illustrated by envy and the speaker being sick. She then turns to another aspect of the job, articulating both parts with autrement.

29When the antecedent is a discourse segment, it is not necessarily headed by an explicit discourse topic. In [11], the speaker does not introduce the part in bold with something like here’s the unpleasant part. Instead, topicality emerges from discourse structure itself; although this point is debated in discourse theory, discourse cohesion cannot be readily accounted for without (Asher, 2004a, 2004b). Although it is impossible to settle anything definitely here, it should be remarked that the ability for discourse segments to function as antecedents to autrement is an argument in favour of “emerging topics”, since autrement takes topics as antecedents elsewhere, as shown in the previous sections; if we did not assume such a topic to exist in [11], this would imply that autrement works differently here, which is not the case in other respects. The segment must have some kind of unity to work properly with autrement, and this unity must be thematic (Isambert, 2010; see also the references by Asher against referential chains as a sufficient device to ensure discourse cohesion, and Isambert, 2008b, for experimental data).

6. Delimitating topics

30So far I have said that the antecedent to autrement is a (possibly implicit) topical element in the left context. However, this poses a problem: there might be (and there generally are) several active topics in a given context. I have mentioned topics and sub-topics, for instance. The question is: how does autrement target the right topic? In other words, how are wrong anaphor resolutions avoided? Given the schematic structure below, how does one know what topic is closed?

Topic ... Sub-topic ... Sub-sub-topic ... autrement ... new ((sub-)sub-)topic?

31One obvious answer is that the discourse following autrement is at the same level as the antecedent. While this answer is sensible, it does not suffice, for two main reasons: first, because “being at the same level” is not a clear linguistic property, although it makes intuitive sense; second because most examples do contain explicit markers that help speakers follow the right track.

6.1. Closing devices

32First, let us see how topics are closed in the first place. Example [5] above is replete with markers indicating that the current topic is to be closed: first a kind of summary statement in [5d] brings the narration to a conclusion; it is followed by voilà (itself prefixed with a short, perhaps insignificant pause) and then a two-second silence. Voilà also occurs in [8] to close the current sub-topic, and [11] contains the less obvious ben just before autrement.

33In more than 20% of the utterances, autrement is preceded by such a marker (besides voilà: oui, d’accord, ok). Also, autrement frequently occurs at the beginning of a turn (25% of the utterances), either after a short (one-word) interruption by the interlocutor, or precisely as a mean for the interlocutor to interrupt the speaker.

34Those markers, although they appear quite frequently in such positions, are not unambiguous: they might appear elsewhere. They must be important clues nonetheless, otherwise they would not occur with such regularity. They are also strictly speaking unnecessary: in several examples they do not occur. But what seems redundant in retrospect makes more sense at production time, especially in oral speech: discourse structure is an “emergent” property of speech and signalling it redundantly probably ensures it is understood correctly.

35Those closing devices do not automatically ease the discovery of the right antecedent for autrement, and as we will see presently other strategies are used. However such devices are sometimes more explicit and prepare the ground for a new (sub-)topic level with the one they close. This happens in [8], where qu’est-ce qu’il y a d’autre (what else) occurs between voilà and autrement. The interrogative form indicates that more is to come, whereas autre links the upcoming topic to the previous one. Here, autrement seems redundant, especially after the morphologically related autre, but even if we choose to ignore the remark on redundancy above, autrement seems to become a reflex when shifting topic: it isn’t mandatory yet, and will perhaps never be, but it does occur with higher frequency than what might be expected from sheer superficial “economy”, including contexts where the topic shift is made clear by other means.

6.2. Syntactic and semantic similarity

36Let us now turn to the question of finding the right antecedent. Obviously autrement alone does not suffice, and here we will see how our intuition about topics at the same level makes linguistic sense. The more frequent strategy (40% of the examples) is to state the new topic immediately after autrement; from the point of view of discourse structure, this new topic is on a par with the previous one because it has most, if not all, the syntactic and semantic properties of the antecedent to autrement. If the antecedent is a topicalized noun phrase, then the new topic is one too, as in [8] above, where the antecedent to autrement is the topicalized les paramètres and the new topic, gestion des mots de passe, is also topicalized. In [9], le film (antecedent) and les informations (new topic) are both sentence fragments working as answers to the question and as objects to the implicit predicate I watch. Finally, in [10] the antecedent is the framing prepositional phrase à vingt ans and the new topic is avant cette guerre, a framing prepositional phrase too.

37As might be expected, semantics plays an important role too. For instance, in [9], film (movie) and informations (news) are both hyponyms of émissions de télévision (TV programs); thus, they are lexically equal, one more clue to their being on the same discursive level. In other terms, the range of possible antecedents to autrement is narrowed to those similar to the new topic, and in all cases there is only one such element available in the preceding discourse. In [10], it is crucial that à vingt ans and avant cette guerre are both temporal frames, otherwise the latter would not easily replace the former (Charolles, Vigier, 2005).

38Thus, for a new topic to be at the same level as the antecedent of autrement is not only intuitively satisfying, it is also backed up by data. Interestingly, whereas we started by asking how autrement could target the right antecedent, we now might ask the opposite question: given that topics match each other so clearly, what is the need for autrement? In other words, since a new topic obviously replaces the previous one, why does one need an additional topic shift marker at all? Interestingly, redundancy does not explain it all: in some of the examples above, removing autrement does not just decrease redundancy, it also decreases acceptability (in [5] and [9], for instance), unless other means are used (strong closing devices as described in the previous section); even the frames in [10], which are generally thought as strong discourse organizers at least in written text (Charolles, 2005), do not make discourse structure clear enough to dispense with autrement.

39In still other cases, relations between topics are actually shaped by autrement. It is the case for instance in [11], where the discourse without autrement would be contradictory, the speaker asserting that her workplace was at once pleasant and unpleasant. In the following (literary) example, a similar contradiction would ensue without autrement:

[12] Juste au bord des eaux qui baissent chaque jour, une teinte verte persiste aux branches ; autrement, n’importe où l’on regarde, c’est, dirait-on, la rouille de l’arrière-automne, ou les grisailles de l’hiver.
  ‘Near to the dwindling waters, a green hue remains to the branches; elsewhere [autrement], wherever one looks, it looks like the rust of after-autumn, or winter’s greyness.’
  (Pierre Loti, L’Inde sans les Anglais, 1903)

40The interesting point here is that while juste au bord des eaux qui baissent chaque jour and n’importe où l’on regarde are both frames and thus could stand on a par, the latter would be understood as included in the former (i.e. wherever one looks falls under the scope of near the dwindling waters by default) but for the presence of autrement which prevents such an interpretation: the speaker has switched topics, and near the dwindling waters is excluded from the new one, despite its generality.

41Thus, in cases where autrement is required as well as in those where it seems redundant, there is a constant trade-off between what is contributed by relations between topics and what by autrement. There is a strong tendency for both sets of clues to occur together, and expectations stemming from studies of written text or artificial examples should be revised when it comes to oral speech: redundancy is in the eye of the beholder and with such frequency and motivation it is hardly redundancy anymore.

6.3. Repetition of the overarching topic

42Yet another way for autrement to find its antecedent without ambiguity is to mention the topic hierarchically superior to the one being shifted (i.e. the antecedent itself; this accounts for 40% of the examples). This is exemplified in [5f], where les choses intéressantes follows autrement immediately, that noun phrase repeating the explicit topic of the current conversation turn as stated by the interviewer under the form tout ce qui vous intéresse. The following example is clearer still:

[13] A: Alors est-ce que est-ce qu’on peut parler de votre région ? Qu’est-ce que ça évoque pour vous ?
B: Oh ben la région euh y a j’ai toute ma famille qu’habite là alors euh… c’est je trouve ça bien quoi. Et euh on habite euh tout euh, j’ai mes grand-mères et tout ça qui habitent dans le même village à part euh j’ai un oncle qui a qui est parti habiter ailleurs à cause de son travail, alors euh parce que ici il peut pas faire le travail qu’il fait alors euh voilà. Euh autrement dans la région y a c’est surtout l’agriculture.
  A: So can we talk about your region? What does it mean to you?
B: Oh well, the region, all my family lives here so… I think it’s nice. And we live, well, my grandmothers and all, they live in the same village, except an uncle, he left to live elsewhere because of his job, because here he can’t do the job he’s doing, so that’s it. Apart from that in the region it’s mostly farming.’
  (Corpus de la Parole)

43The interviewer explicitly asks the interviewee about a specific topic (the interviewee’s region), which the latter then splits into sub-topics (his family and the region’s economy), articulated by autrement and introduced by a predication on the main topic itself. Accounting for this strategy can be done along the following lines: the speaker shifts topics with autrement, which implies that autrement finds an antecedent; the question, once again, is: how is this antecedent unambiguously selected? Let us suppose that topics are hierarchically ordered (i.e. a topic is either the main topic or a sub-topic to another one); since the speaker reinstates the topic la région after autrement, obviously that is not the topic he chooses to close; he cannot close a topic hierarchically higher either, otherwise la région would be automatically closed too (Isambert, 2008a, and next section). The only remaining candidate is a sub-topic to la région, in this case the speaker’s family.

44If we indulge in some speculation at this point, it seems to be the case that repeating an overarching topic after autrement closes all subtopics. In [13], the speaker talks about his family (a sub-topic to la région) and then more precisely about his uncle (a sub-topic to the family topic); we know that autrement shifts a topic, we know it must be a sub-topic to la région, but there is no a priori reason that it be the immediately subordinate one (the speaker’s family) instead of one lower still (his uncle). Nonetheless, both in [5] and [13], and in other examples as well, the repetition of a topic implies that all sub-topics are closed, and it never happens that the speaker closes some sub-topic and leaves other open. It is as if the overarching topic were not reinstated only for convenience, leaving the possibility that sub-topics are still active: instead, such a move means that the speaker goes on building discourse structure from the point indicated by the reinstated topic. This implies some strong assumptions about discourse structure that I investigate in the next section.

7. What it means to close a topic

45In what precedes I have made some implicit assumptions in order to carry on the analysis of topic shifts with autrement. One is that the antecedents to autrement are available for anaphor resolution. Another assumption is that closing a topic forbids any further reference to what was introduced under its heading. Both assumptions address the much-discussed issue of the Right Frontier in discourse structure.

7.1. The Right Frontier

46The notion of Right Frontier (Polanyi, 1985; Asher, 2005; Isambert, 2008b) is meant to account for the (un)availability of discourse referents at a given point in discourse. It is assumed that referents on the Right Frontier are available as antecedents to anaphora, whereas others are not. What exactly is on the Right Frontier is a difficult matter: the last proposition, of course, and any points higher in the structure, i.e. propositions to which the last one is related more or less directly by a subordinating relation as theorized in SDRT [Segmented Discourse Representation Theory] (Asher, Vieu, 2005). In contrast, coordinating discourse relations makes the point where the relation attaches unavailable. Things are not so straightforward, however, and some referents seem to be able to percolate upward and remain available even though the proposition where they occur has become unavailable, at least when those referents have sufficient importance to be considered topical (for instance, main characters in a narrative).

47The availability of the antecedents to autrement implies that they belong to the Right Frontier. Since the proposition where they first appeared is often far away in the left context, this in turn implies that they have some special place in the discourse structure. Such distinction has often been considered a hallmark of elements such as topicalized phrases and frames, whose scope has been shown to exceed the sentence where they belong (Charolles, 2003; Combettes, 2005). It is thus no surprise that the antecedents to autrement frequently turn up to be topicalized phrases and frames.

48The importance of macropropositions should not be ignored either (Guindon, Kintsch, 1984). Such propositions play the role of summaries of what is to come, and in our example are generally found in a question by an interviewer (e.g. [5], [13]). They remain longer in memory and shape the discourse that follows. In terms of discourse relations, macropropositions are elaborated by the propositions they introduce, and such elaboration is considered a prime example of a subordinating discourse relation (Asher, Vieu, 2005). Consequently, macropropositions remain available when discourse proceeds, i.e. they stay on the Right Frontier. It is again no surprise that autrement relies on such propositions, or their important part (e.g. your region in can we talk about your region?), when anchoring the topic shift.

49Thus, autrement does not bring new insights on the Right Frontier, but it does bring further evidence to its reality. Furthermore, when autrement seems to rely on invisible clues like emergent topics in [11], one is justified in thinking that such topics indeed exist in discourse structure, because there is no good reason why autrement should behave differently in those examples.

7.2. Closed topics and the Right Frontier

50Since topics are like headings over discourse segments, i.e. they are elaborated by discourse segments, if they are closed and replaced with a new one, they should not be available anymore for anaphora resolution, nor should the discourse segments that depend on them. That is, they leave the Right Frontier and the referents they contain cannot be antecedents anymore. Is this the case?

51Assessing such an assumption with corpora is no simple task, and psycholinguistic experiments have a word to say. Without such data at hand (Isambert [2008b] presents an experiment about the accessibility of topics, but not about autrement precisely), one can nonetheless gather some clues in real utterances. For one thing, there is negative evidence: it never happens that a speaker refers to an entity introduced in a discourse segment headed by a topic that has been closed by autrement. I.e. one never finds an example like [14] below, a modified version of [13] where the proposition containing autrement makes reference to an entity introduced under the previous topic (the uncle’s speaker).

[14] Oh ben la région euh y a j’ai toute ma famille qu’habite là alors euh… c’est je trouve ça bien quoi. Et euh on habite euh tout euh, j’ai mes grand-mères et tout ça qui habitent dans le même village à part euh j’ai un oncle qui a qui est parti habiter ailleurs à cause de son travail, alors euh parce que ici il peut pas faire le travail qu’il fait alors euh voilà. Euh autrement dans la région #il a tous ses amis.
  ‘Oh well, the region, all my family lives here so… I think it’s nice. And we live, well, my grandmothers and all, they live in the same village, except an uncle, he left to live elsewhere because of his job, because here he can’t do the job he’s doing, so that’s it. Apart from that in the region he has all his friends.’

52One never finds such examples, but does it mean they are unacceptable (as noted here)? It seems that the only way to make sense of [14] is that autrement does not shift the “family” sub-topic but one deeper yet, namely the speaker’s uncle’s job. This is in contradiction with the overarching topic (la région) being repeated, but then the hearer might still accommodate the prepositional phrase dans la region as a detached element without topical force, hence not a repetition of the topic per se, despite the referential identity. Such considerations might be taken into account to make sense of the [14], but it requires quite a lot of processing, and this may explain why the example does not sound good.

53Other examples may be clearer, and it seems for instance impossible to retrieve anything about movies in [9] once the speaker has turned to news. The list structure of this example, where topics are managed thanks to strong syntactic-semantic means, probably plays some role: the sounder the structure, the harder it is to violate it. Thus, autrement seems to be a strong structural device, but as we have said above, it works all the better with other markers.

8. Conclusion

54While autrement is not the only marker used to manage the topical structure of discourse (others include similar devices such as sinon, à part ça…), and while it often comes accompanied by other markers, it remains crucial in an operation quite simple, yet vital to discourse: changing the matter of conversation. In doing so it is a valuable window on discourse structure, since it exhibits mechanisms that have been more often investigated with regard to discourse referents than to the apparently more fleeting discourse topics.

55An interesting fact, not addressed in this paper, is that this use of autrement is almost non-existent in written text, which is all the more striking as it is extremely frequent in oral discourse. The reason for this is that autrement is still perceived as colloquial and thus outside the written norm, a fact that probably betrays its youth: the grammaticalization that has created it from the uses mentioned at the beginning of this paper must be quite recent.

56Another factor might be that discourse structure in written texts is often made as clear as possible by other means, framing expressions being for instance used more systematically, not to mention titles and paragraphs. Thus the apparent redundancy displayed by oral utterances is unnecessary, and generally considered as “bad grammar”. In oral speech, the situation is quite different; for one thing, the hearer cannot reprocess a discourse segment by reading it again. Means similar to those in writing are used (pauses, for instance), but they are transient, and often ambiguous or ambiguously used; autrement, on the other hand, is clear, because it has no other function but to shift topics.

Haut de page


Asher, N. 2004a. Discourse Topic. Theoretical Linguistics 30: 163-202.

Asher, N. 2004b. Troubles with Topics: Comments on Kehler, Oberlander, Stede and Zeevat. Theoretical Linguistics 30: 255-262.

Asher, N. 2005. Troubles on the Right Frontier. In Papers of the Symposium on the Exploration and Modelling of Meaning (SEM-05). Biarritz. Available online:

Asher, N. & Lascarides, A. 2003. Logics of Conversation. Cambridge: Cambridge University Press.

Asher, N. & Vieu, L. 2005. Subordinating and Coordinating Discourse Relations. Lingua 115: 591-610.

Charolles, M. 2003. De la topicalité des adverbiaux détachés en tête de phrase. Travaux de linguistique 47: 11–51.

Charolles, M. 2005. Framing Adverbials and their Role in Discourse Cohesion: From Connection to Forward Labelling. In Papers of the Symposium on the Exploration and Modelling of Meaning (SEM-05). Biarritz. Available online:

Charolles, M. & Vigier, D. 2005. Les adverbiaux cadratifs en position préverbale : portée cadrative et organisation des discours. Langue française 148: 9-30.

Combettes, B. 2005. Les constructions détachées comme cadres de discours. Langue française 148: 31-44.

Consten, M., Knees, M. & Schwarz-Friesel, M. 2007. The Function of Complex Anaphors in Text. In M. Schwarz-Friesel, M. Consten & M. Knees (eds), Anaphors in Text. Amsterdam: John Benjamins: 81-102.

Guindon, R. & Kintsch, W. 1984. Priming Macropropositions: Evidence for the Primacy of Macropropositions in the Memory for Text. Journal of Verbal Learning and Verbal Behavior 23: 508-518.

Inkova-Manzotti, O. 2002. Les connecteurs accommodants : le cas de autrement. Cahiers de linguistique française 24: 109-141.

Isambert, P. 2006. Un connecteur accommodant : autrement. Master’s thesis. Université de la Sorbonne Nouvelle. Paris.

Isambert, P. 2008a. Topicality and Discourse Structure: Evidence from the French Marker autrement. In 30. Jahrestagung der DGfS, Workshop on Topicality. Bamberg.

Isambert, P. 2008b. Making Discourse Structure Realistic. In Utterance Interpretation and Cognitive Models II. Brussels.

Isambert, P. 2010. Discours et grammaticalisation : étude de l’adverbe autrement. PhD thesis. Université de la Sorbonne Nouvelle. Paris.

Jacobs, J. 2001. The dimensions of topic-comment. Linguistics 39: 641-681.

Lambrecht, K. 1994. Information Structure and Sentence Form. Topic, Focus and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press.

Lamiroy, B. & Charolles, M. 2004. Des adverbes aux connecteurs : simplement, seulement, malheureusement, heureusement. Travaux de linguistique 49: 57-79.

Lamiroy, B. & Charolles, M. 2005. Utilisation de corpus pour l’évaluation d’hypothèses linguistiques : étude de autrement. In A. Condamines (ed.), Sémantique et corpus. Paris: Hermès: 109-147.

Lewis, D. 1979. Scorekeeping in a Language Game. Journal of Philosophical Logic 8: 339-359.

Polanyi, L. 1985. A Theory of Discourse Structure and Discourse Coherence. In Proceedings of the 21st Meeting of the Chicago Linguistics Society. Chicago: Linguistics Department, University of Chicago.

Traugott, E.C. 1995. Subjectification in Grammaticalization. In S. Wright & D. Stein (eds), Subjectivity and Subjectivisation. Cambridge: Cambridge University Press: 37-54.

Haut de page

Pour citer cet article

Référence électronique

Paul Isambert, « What’s on the left? », Discours [En ligne], 8 | 2011, mis en ligne le 11 juillet 2011, consulté le 26 juin 2017. URL : ; DOI : 10.4000/discours.8423

Haut de page


Paul Isambert

Université de la Sorbonne Nouvelle

Haut de page