ELSEVIER
Journal of Pragmatics 28 (1997) 153-187
Alors and donc in spoken French: A reanalysis Maj-Britt
Mosegaard
Hansen*
Romance Studies. University ~C,~penhagen. Njalsgade 80. 2300 Copenhagen S. Denmark Received 9 February 1995; revised version 16 November 1996
Abstract This paper presents a new analysis of the common French discourse markers alors and doric. After a methodological/theoretical introduction, in which I propose, among other things, a definition of the class of discourse markers, I go on to consider the distribution and semantics of the two particles, mainly in contemporary spoken French, but with some consideration of their diachronic origins and evolution. While a unified semantic account is given for donc, I suggest that alors, which is clearly multifunctional, may usefully be seen as a radial category.
1. Introduction In this p a p e r I p r e s e n t an a n a l y s i s o f the d i s c o u r s e m a r k e r s alors a n d done, as t h e y are u s e d in c o n t e m p o r a r y s p o k e n F r e n c h . T h e f o l l o w i n g are t y p i c a l e x a m p l e s : 1 ( I ) ... c e q u e j e c o n s t a t e et q u i m e fait rire d a n s ce q u e v i e n t d e dire C l a u d e E s t i e r c ' e s t q u e q u ' o n t dit L i o n e l J o s p i n aussi b i e n q u e M o n s i e u r P o p e r e n q u ' a dit ~ I would like to thank Peter Harder, Frederike van der Leek, Henning N¢lk, Ebbe Spang-Hanssen, and three anonymous referees for the Journal of Pragmatics for helpful comments on earlier versions of this paper. Needless to say they should not be held negatively responsible for its contents. * E-mail:
[email protected] I would like to thank Professor Mar~-Annick Morel of the University of Paris Ili for allowing me access to the corpora collected by the Dept. of French Linguistics. A few transcription conventions should be noted: , and ,, indicate short and somewhat longer pauses, respectively. & and && indicate the beginning and end of overlap (where two or more instances of overlap occur in sequence, § and §§ are also used to distinguish them). Capital letters, other than in proper names and titles, indicate intonational et~lphasis. XXX indicates a number of inaudible syllables, and? indicates rising intonation. As some examples may contain more than one instance of a particular marker, the one commented on will be in bold italics. I have taken the liberty of occasionally 'simplifying' the transcriptions by changing 'phoneticized' orthography to conform with the standard, and by leaving out a number of pauses, hesitations, and restarts, where these did not occur in the immediate vicinity of the markers being commented on. This should make the examples somewhat easier to comprehend for the untrained reader. 0378-2166/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved PH S 0 3 7 8 - 2 1 6 6 ( 9 6 ) 0 0 0 8 6 - 0
154
M.-B. Mosegaurd Hansen / ,hmrmd ~?/Pragmatics 28 (I 997) 153-187 Georges Marchais l'autre jour, c'est que il faut lancer les uns une contre-offensive repartir ~, l'offensive, or qui lance une contre-offensive, ben les gens qui ont 6t6 battus, alors je ne vois pas pourquoi Claude Estier nous reproche "~ Amouroux e t ~ moi de dresser un constat d'6chec,, qui est, constat d'6chec qui actuellement Est la motivation de leur politique (VSI: 13)
(2) ... et je consid~re cela comme 6tant pratiquement une preuve du fait que ce syst~me des nombres entiers a une richesse en quantit6 d'informations qui n'est pas 6puisable qui est infinie et donc je d6nie ~ u n individu quelconque la possibilit6 d'avoir cr66 un tel systbme (NN: ii) The study forms part of a larger research project on the distribution and semantics of a type of discourse markers which are particularly characteristic of (relatively informal) spoken French. Examples other than the items dealt with here include bon, ben, eh bien, and puis (cf. Hansen, 1995a,b; 1996a,b). Some of them, such as bon or ben, are used almost exclusively in this mode, and even those that may be found in (relatively formal) written texts, such as alors and donc, are more common, and have a wider range of functions in speech. I have therefore restricted my study of these items to their function in what one might call contemporary spoken 'standard' French, i.e. the kind which is spoken by educated Parisian speakers and which exhibits no noticeable regional or social characteristics. 2 Apart from a very few constructed examples, the analysis is based entirely on nonelicited data from taped verbal interactions of various kinds, involving native speakers only. The corpora used consist of the following: (a) Two informal face-to-face conversations, and one equally informal telephone conversation, all involving pairs of friends; (b) Four radio debates involving between two and five participants, in which the order and/or length of turns is not pre-allocated, and in which topic flow is to some extent locally managed, although a global agenda does of course exist; (c) One university lecture (clearly prepared in advance of its delivery, but not read aloud from a manuscript), containing a few questions and comments from the audience. Altogether the data total roughly five hours of talk. As I am interested in the distribution of the markers 'only' to the extent that it allows me to determine their semantics and function in the construction of discourse, no attempt has been made to control for sociological variables such as age, sex, occupation, or geographical origin of the speakers. Theoretically, my work relies mainly on the framework of Conversation Analysis (for introductions cf. Levinson, 1983: ch. 6; Heritage, 1984: ch. 8; Schiffrin, 1994: ch. 7 et passim), and by that of Relevance Theory (cf. Sperber and Wilson, 1986). Considerations of space preclude a detailed exposition of these theories, but what is important here is that, although they are very different in that Relevance Theory is -~ 1 am fully aware of the somewhat cavalier nature of this definition, but as my work d,-,es not situate itself within the field of variationist sociolinguistics, and as I do think the 'unmarked' variety referred to will be intuitively identifiable to anyone who is sufficiently familiar with French, I will not attempt a more precise circumscription of it.
M.-B. Mosegaard Hanson / Journal of Pragmatics 28 (1997) 153-187
155
interested in utterance (and discourse) interpretation as a cognitive process, while Conversation Analysis sees it as an interactional one, both frameworks crucially provide for the essentially local and process-oriented management of the understanding and structuring of discourse. Although one may roughly distinguish between discourse that is primarily locally managed, i.e. on a turn-by-turn basis, and where interlocutors are principally oriented towards the process of its construction rather than the finished product, and discourse where orientation to a final global product is of importance, this is of course not a matter of either-or. The distinction is one that admits of degrees, depending on discourse type, with spontaneous conversation between intimates at one end of the continuum, and, for instance, written expository text at the other. Thus, it is clear that the corpora briefly described above fall at different points along this continuum. Nevertheless, where spoken language is concerned (barring the limiting case of a speaker reading aloud from a pre-prepared manuscript), discourse production is by definition an on-line, incremental process, involving the transformation of a non-linguistic hierarchically structured mental representation into linear linguistic expression. Speakers are thus under continuous cognitive pressure, and, depending on the turn-taking conditions that prevail, they will moreover be under interactional pressure to manage their contributions so as not to lose the floor prematurely. On top of this, speakers will continually need to negotiate actions, meanings and relevancies with their interlocutors, and extralinguistic phenomena may intrude and change existing conditions in subtle and not-so-subtle ways. All these factors contribute to the need for discourse markers as indicators of coherence in discourse. Moreover, it is at least conceivable that the use of markers of the type I am studying may reduce planning time for s?eakers, as compared to more explicit connective devices: in virtue of their multifunctionality and relative open-endedness of interpretation, markers like alors, bon, (et) puis etc. are to some extent comparable to 'all-purpose' nouns such as truc or machin (roughly: thingey or whatchamacallit), which speakers often use when more precise terms are not immediately accessible. This is in no way meant to imply that these items are devoid of coded content, or that there are no constraints on their possible uses, but merely that their range of applicability is typically quite broad.
2. Discourse markers Before proceeding to the anal,isis of alors and donc, I would like to discuss and attempt to define the notion of discourse markers as such. This is not an entirely straightforward matter, to the extent that these items do not appear to constitute a natural kind. 2.1. The problem The main difficulty lies in the fact that the category of markers is fairly clearly a functional-pragmatic, and not a formal, morphosyntactic one (cf. Lamiroy and Swig-
156
M.-B. Mosegaard Hansen / Journal ~)/~Pragmatics 28 (1997) 153-187
gers, 1991 : 123). For one thing, items which may be used as markers may originate in quite different distributional classes, where they often have formally identical counterparts that are not used as markers, and which do contribute to propositional content (as markers do not, according to the definition outlined below): e.g. adverbs (alors), adjectives (bon), imperative verb-forms (tiens), sentence fragments (si tu veux) etc. Secondly, even when used as markers, these items do not all have similar formal properties. While the items treated here and in Hansen (1995a,b; 1996a,b) have in common that they prototypically introduce the discourse segments they mark, and seem to remain external to the structure of the clause, inasmuch as they tend to constitute independent tone units, some of them are to some extent free to appear clauseinternally or -finally, and may occasionally, as with bon in (3) and done in (4), or invariably, as with all occurrences of puis, be intonationally integrated: (3) ah bon ah ben je sais pas qa parce que bon nous on a ies notions qu'on a entendues cet 4t6, c'est tout (VE: 12) (4) A. mais Estier c'est peut-&re de la mauvaise tactique 61ectorale, car deux listes, une du RPR et une de I ' U D F peuvent faire, 53 ou 54%, 55% tandis qu'une liste unique, qui va susciter h c6t6: des listes de faible importance qui peut faire que 48 ou 49 qa serait doric une mauvaise tactique & 61ectorale && B. & vous ~tes pour && deux listes alors (VS2: 16) Other markers, such as hein and quoi, which are not treated in my study (but see H61ker (1988) for an in-depth analysis of quoi), only, or at the very least with overwhelming frequency, occur at the end of the segments they mark:
(5) A. ... ii se passe quelque chose de trbs curieux actuellement avec ce livre c'est que les gens de i'opposition le trouvent tr6s bon et les gens de la majorit6 aussi B. cherchez l'erreur A. qa explique qu'elle en vende 60.000 en moyenne par semaine ce qui est:: une performance hein (MP: 5) (6) A. mais comment elles 6taient habill6es ces soeurs B. avec des esp6ces de couvertures et le visage tout blanc et tout atroce quoi (CV: 7) Others again, such as si tu veux or voile, seem to have no clear preferences as to position, although it may be that different positions are responsible for subtle changes in meaning or function. It is thus not obvious whether markers may be said to constitute a paradigm in the traditional sense. There are at least three arguments in favor of postulating some sort of paradigm. First, markers which occur in similar positions in the utterance may often be substituted for one another, although not without at least some change in
M.-B. Mosegaard Hansen / Journal of Pragmatics 28 (1997) 153-187
157
meaning, tone, or implications of the utterance. Thus, if (7) and (5) are grammatical, then so are (7') and (5'):
(7) A. il est tout petit il est & il est B. & je vois pas bon c'est pas grave (CV: 6) (7') A. il est tout petit il est & il est B. & je vois pas enfin c'est pas grave
(5) A .... il se passe quelque chose de tr~s curieux actuellement avec ce livre c'est que les gens de l'opposition le trouvent tr6s bon et les gens de la majorit6 aussi B. cherchez l'erreur A. qa explique qu'elle en vende 60.000 en moyenne par semaine ce qui est:: une performance hein (MP:5) (5') A. ... il se passe quelque chose de tr6s curieux actuellement avec ce livre c'est que les gens de l'opposition le trouvent tr~s bon et les gens de la majorit6 aussi B. cherchez l'erreur A. qa explique qu'elle en vende 60.000 en moyenne par semaine ce qui est:: une performance quoi Second, just as in the case of modal particles (see for instance Foolen, 1993: 170), markers which occur in similar positions frequently cluster in these positions, and when they do, there are at least some restrictions on their order of occurrence, but at the same time, such clusters can always be reduced by one or more markers, without loss of grammaticality (although again not without some change of pragmatic import): (8) ... et elle, bon eux ils disaienl le plateau de Guiseh, puis alors bon donc on a fait cette visite ...(VE: 20) (8') *... et elle, bon eux ils disaient le plateau de Guiseh, doric bon alors puis on a fait cette visite (8') ... et elle, bon eux ils disaient le plateau de Guiseh, bon doric on a fait cette visite (8") ... et elle, bon eux ils disaient le plateau de Guiseh, puis alors on a fait cette visite Third, these items seem to haw~ at least roughly comparable functions in discourse (to be outlined below). However, if we do take this a:; indicating the existence of a paradigm of sorts, it is unlikely that we are dealing with a paradigm in the traditional structuralist sense, where the meanings of individual members is assumed to essentially emerge from the relations of contrast that they sustain with other m e m b e r s ? In many cases, the The arguments presented here are adopted from Bybee (1988), who argues against the structuralist notion of paradigms even in cases of more highly grammaticalized items such as nominal cases or tenseaspect distinctions.
158
M.-B. Mose~aard Hansen / Journal ~] Pra~matics 28 (1997) 153-187
meanings and functions of one marker will overlap with those of another, so that it will at least in some utterances be of relatively little consequence which particular item is chosen, while in other contexts, one of the two will clearly be more felicitous than the other. Witness the following examples: (9) I1 pleut. Donc/alors je ne vais pas me promener (10) C'est le cousin de ma femme, et donc/?alors mon cousin par alliance (11) Tu sais tout, alors/?donc donne-moi le tierc6 In other words, markers do have inherent meanings. This is further supported by the fact that, as noted above, a number of markers originate in items or expressions which indisputably are inherently meaningful, and one may hypothesize that the nature of these lexicai sources does to some extent determine the uses of the markers that are derived from them. Finally, there seems to be no reason to suppose that items classified as discourse markers here and elsewhere in the literature evolved simultaneously, as a paradigm, or that the evolution of one marker was essentially dependent on that of others. Consequently, there is little reason to expect their evolution to be oriented towards the maximal use of oppositions. Hence, one may be suspicious of models, such as that of Schiffrin (1987), which suggest that markers may be distinguished essentially according to the planes of talk on which they are said to function (see Redeker, 1991) for a critique of Schiffrin's empirical claims). 2.2. Previous definitions of discourse markers To my knowledge, there are three relatively theory-independent definitions in the literature, namely those of Schiffrin (1987), Fraser (1990), and Redeker (1991). 4 Schiffrin ( 1 9 8 7 : 3 I) defines markers operationally as "sequentially dependent elements which bracket units of talk", the size of these units being variable and not necessarily corresponding to single, or even complete, utterances. Fraser (1990: 386-387), fairly uncontroversially, sees sentences as being able to encode on the one hand, content (= propositional) meaning, and on the other hand, pragmatic meaning. The latter is expressed by three major classes of markers, which he terms 'basic', ~commentary', and 'parallel' markers, respectively. Discourse markers are then defined functionally as "one type of commentary pragmatic marker ... signal[ing] how the speaker intends the basic message that follows to relate to prior discourse". 4 Other definitions do exist, which, however, seem more strongly dependent on the theoretical frameworks in which they are proposed. One which is quite explicit is that of the Geneva School; others are those of Ducrot and Anscombre's Argumentation Theory and of Relevance Theory, respectively. The latter two are not definitions of discourse markers as such, but comprise items of a variety of grammatical categories, which, however, are seen as having comparable semantic or pragmatic functions.
M.-B. Mosegaard Hansen /,lournal ~Pragmatics 28 (1997) 153-187
159
Finally, Redeker (1991: 116~.) defines what she calls 'discourse operators' as "word[s] or phrase[s] ... that [are] uttered with the primary function of bringing to the listener's attention a particular kind of linkage of the upcoming utterance with the immediate discourse context". All three definitions imply, in my view correctly, that discourse markers are basically connectives, but they also make additional assumptions, which upon more careful consideration turn out to be not entirely accurate. For Fraser (1990: 385-386), content, or propositional, meaning is equivalent to the 'basic message', which in the context of his definition of markers seems to imply that these items must always take an entire proposition in their scope. This would leave the puis in (12) unaccounted for: (12) ... on est all6s ~t Abou Simbel parce que y a les deux temples-la: euh:, de: comment Rams~s euh:, Rams~s II, oui et puis sa femme qui sont c6te a c6te quoi (VE: 31) Both Fraser and Redeker assume that markers must necessarily precede their host units (they speak of "the message that follows" and "the upcoming utterance"). If we accept this, we have to exclude the done in (13) from the class of markers: (13) A . . . . cela dit on s'est jamais pench6 dans sa continuit6 en France sauf Franqois Mauriac qui alors lui avait fait un livre tout a fait hagiographique pour reprendre le mot de tout ~. l'heure qui avait valu un pamphlet tr6s r6ussi d'ailleurs de Jacques Laurent qui s'appellait Mauriac sous De Gaulle B. pour lequel il a 6t6 condamn6 d'ailleurs A. pour lequel il a 6t6 & condamn6 && B. & il a 6t6 condamn6 && pour offense au § Chef de l'Etat §§ A. § et §§ c'est pas la meilleure page de l'oeuvre de Mauriac doric (MP: 11) Redeker's assumption that markers introduce 'utterances', rather than merely 'messages' (note that I do not wish the term 'message' here to be understood as equivalent to 'proposition'), would seem to exclude uses such as (14), where alors introduces the apodosis of a hypothetical construction, the whole of which must surely be considered a single utterance: 5 (14) ... je crois que si l'on commence 5_ partir du principe, que y a pas d'unit6 politique possible au Liban, alors, on on est parti pour on ne sait quoi ... (VS 1 : 29)
I should mention that Gisela Redeker (p.c.) would in fact consider (14) as comprising two utterances because of the intonational break before "llors. However, in the paper quoted here, she defines utterances as "'intonationally and structurally bounced, usually clausal units" (1991: 1168). One might legitimately question whether a si-clause without its matrix clause may be said to be structurally bounded, especially since, if we take the intonational criterion into account as well, this particular si-clause ends up belonging to the same utterance as the matrix clause of its own (according to Redeker, independently uttered) matrix clause, namely .je trois.
160
M.-B. Mosegaard Hansen / Journal t~f Pragmatics 28 (1997) 153-187
Schiffrin's notion of markers as 'brackets' is compatible with all these examples, but not with (15), where donc appears clause-internally: (15) ... moi j'6tais sur la plage avec le reste des gosses j'avais donc euh vingt-et-un gosses que je devais surveiller ... (CV: 32) Finally, by speaking of "sequentially dependent elements" and of "relat[ions] to prior discourse", both Schiffrin and Fraser imply that markers always connect their host utterance to the linguistic co-text, rather than to the context in a wider sense (including, of course, the co-text). However, while the signalling of co-textual connections is certainly the prototypical function of discourse markers, "the relation indicated by a cohesive device is not always a relation between linguistically realized meanings", as is pointed out by Blakemore (1987: 106). Borrowing her example and translating it into French, we see that it is perfectly possible to open a conversation with someone who arrives laden with parcels by uttering (16): (16) Alors, tu as d6pens6 tout ton argent And, as Blakemore notes, there is no reason to suppose that this way of using discourse markers is any different in principle from instances where markers establish connections between linguistic entities. This is compatible with a view according to which coherence is not a property of texts as such, but of mental representations constructed on the basis of text, context, and inferences from both. 2.3. A revised definition
In view of these facts, I propose to define discourse markers as linguistic items of variable scope, and whose primary function is connective. By 'variable scope' I mean that the discourse segment hosting a marker may be of almost any size or form, from an intonational pattern indicating illocutionary function, as in (17), through subsentential utterances, as in (18), to a segment comprising several utterances, as in (19): (17) A. ce n'est pas parce que votts r6p6tez chaque vendredi soir que la gauche a 6chou6 que la re~ilit6 est celle-l~ ce n'est pas vrai B. & s'il n'y avait que moi && § qa ne serait pas grave §§ A. & ce n'est pas vrai & & § ce n'est pas vrai §§ B. Estier mais Estier s'il n'y avait que moi qa n'aurait aucune importance mais il y a des millions de Franqais qui pensent comme moi & je le r6p~te chaque occasion && A. & prenez prenez prenez && les r6sultats de la politique men6e depuis deux ans et demi comparez-l~ honn~tement et objectivement B. oui eh bien? (VSI: 10)
M.-B. Mosegaard Hansen / Journal t)f Pragmatics 28 (I 997) 153-187
161
(18) ... doric on en a vu un petit morceau par avion, puis on est all6s jusqu'5' Abou Simbel parce que y a l e s deux temples-15': euh:, de: c o m m e n t Rams~s euh: Rams~s II, oui, et p u i s sa f e m m e qui sont c6te 5' c6te quoi (VE: 31) (19) A. B. A. B. A. B. A.
on a vu par exemple § pour euh: §mm c o m m e n t elle s'appelle, Hatchepsout oui mais c'est q u ' o n a martel6 son visage partout mm euh ils ont fait dispara~tre un tas de textes ou de choses qui la concernaient donc y a des zones d'ornbres 15. terribles, et p u i s a l o r s un t r u c / o n n ' a rien vu, dessus mais on a vu simplement au mus6e, c'est euh: l'6poque ils appellent qa L ' H E R E S I [ E : : amarnienne, avec c o m m e n t il s'appelle euh: B: Akh6naton (VE: 5 5 - 5 6 )
It i~ m o r e o v e r part of the defil3ition o f markers that they do not contribute to the propositional content of their ho~,;t units (in other words, they belong to that part of the utterance which is ' s h o w n ' rather than 'asserted' [cf. N01ke, 1 9 9 4 : 1 1 5 - 1 1 6 ] ) , and that they function as instructions from the speaker to the hearer on how to integrate the host unit into a coherent mental representation o f the discourse. This, incidentally, makes them a subclass of conventional implicatures, as suggested by both Levinson (1979: 214) and Blakemore (1987: 76). The non-propositional character o f markers entails the exclusion from the class of markers of, for instance, the adw',rbs n o w and then, when these function as temporal anaphora (any use of these adverbs is included in Redeker's (1991: 1169) notion of discourse operators), and of conjunctions such as b e c a u s e when it operates on propositions (all uses of this item are included in Schiffrin, 1987: ch. 7). The inclusion of items which contribute to propositional content is usually warranted by the observation that they m a y also help to signal or enhance inter-utterance coherence. This stance, however, in m y view makes it almost impossible to draw a principled distinction between items that can function as discourse markers and items that cannot: after all, anaphoric pronoun:~, left- and right-dislocated elements, lexical chains and a great many other devices are also cohesive and hence involved in signalling coherence .6
~' This is not to say that the distinction between what does and does not belong to the propositional level is entirely unproblematical. As one anonymous reviewer has noted, certain instances of for instance done, such as that in (10) below, might ~eem to be of a propositional nature. However, the crucial difference between parce que and done in utterances such as the following: Jean a ~pous~ Marie parce qu'elle est riche and Marie est fiche. D~.nc. Jean l'a dpous6e , is that the first utterance constitutes one speech act, expressing one complex proposition which asserts a causal link between the two component propositions. The cause can thus be the focus of both negation and interrogation, and the structure can be embedded as a whole under a matrix predicate such as I1 est dvident que ... With done, we have a link which intuitively is not asserted but implicated between two separate speech acts, and negation/interrogation/embedding touches only the first utterance.
162
M.-B. Mosegaard Hansen / Journal ~t Pragmatics 28 (1997) 153-187
The instructional character of markers implies that their semantics is procedural, i.e. they have no conceptual core, but are basically instructions on how to process their host utterance in a given context. This has the advantage of making individual items compatible with a large number of different contexts, since, as Ducrot et al. (1980: 33) put it, "there are a thousand possible ways of complying with the instructions conveyed by [a] sentence" (my translation). This, of course, does not imply that 'anything goes': the meaning potential of the marker must, as it were, 'unify' with that of the structure in which it occurs. Having said this, 1 must emphasize that not all uses of the morphemes treated in this paper, and in previous ones, are properly analyzed as discourse markers according to the above definition. Instances of a given item may, in some contexts, be closer to other, related categories such as modal particles, interjections, and the like. The reason why such uses are included is that the methodology followed in my work thus far has been basically semasiological and inductive, starting with individual expressions and trying to arrive at a description of their coded content, rather than onomasiological and deductive, i.e. starting with a definition of possible content categories and then attempting to find out how these categories are expressed in a given language (in casu French). Both methods are presumably equally valid in the abstract, but as far as discourse markers are concerned, 1 find the former more fruitful for two reasons. One is that the semasiological method allows one to explain how apparently different senses of what is materially the same linguistic item may be related (which, on the lexicalsemantic level, allows us as far as possible to opt for polysemy rather than homonymy), while at the same time making it easier to distinguish between materially different items which happen to have (partially) overlapping distributions. The other reason is that, compared to many other areas within linguistics, the study of discourse markers is a relatively recent phenomenon, and attempting an exhaustive taxonomy of content categories in this domain (as is done, for instance, by Roulet et al. (1987: ch. 2), and by Heltoft (1987/1989: 39)) simply seems premature as long as there is little consensus both about the function of individual morphemes, and about exactly which items should be included in the class of markers.
3. Donc and alors as discourse m a r k e r s
The two markers done and alors have two features in common, which justify their being treated together. One is that, diachronically, both originate in temporal anaphoric expressions. The other feature is that, in contemporary French, they are both frequently used in argumentative structures, where they mark a result or a conclusion. Hence, they have partially overlapping distributions, and may, in fact, occasionally occur together in the same utterance, as in (20): (20) ... c'est moi qui doit commencer bon (h) alors d o n c il s'agit de la biographie de MONsieur le Pr6sident de la R6publique (MP: 4)
M.-B. Mosegaard Hanson / Journal q/'Pragmatics 28 (1997) 153-187
163
Interestingly, it seems that, when they do co-occur as (part of) a single intonation unit, they may only do so in the order shown here, i.e. with alors preceding done (unless alors is used in its temporal sense, with which 1 am not concerned here). However, they are apparently not synonymous, even when occurring in utterances that function to indicate a result e r a conclusion, as is demonstrated by the following examples: (10) C'est le cousin de ma femme, et donc/?alors mon cousin par alliance (from Jayez, 1988: 136) (! 1) Tu sais tout, alors/?donc donne-moi le tierc6 (from Roulet et al., 1987: 151) In what follows, I propose a new analysis of these markers, based on data adduced from my corpora, which will hopefully allow us to explain the difference evidenced by the above examples. I start with done, which is the 'simpler' of the two in the sense that it seems possible to propose a single basic sense for this item. But first, a few remarks about the diachronic development of the markers will be useful. Done originates in Latin DUM, either via the elaboration DUMQUE, or via DUNC, a late Latin form which is assumed to be the result of analogy with the pair TUM/TUNC (cf. von Wartburg, 1934: 179; Ernout and Meillet, 1967: 187; Bloch and von Wartburg, 1968: 201). DUM (and its sister adverb DUNe) was a temporal particle/subordinating conjunction indicating the., simultaneity of two states of affairs, and which moreover carried a sense of duration (= at that time~while; cf. Emout and Meillet, 1967). Old French done had several u~';es (cf. Tobler and Lommatsch, 1936: 2005ff.), the most important of which seem to be, first, a temporal one, in which it may, like DUM, be glossed as at that time. Secondly, it could be used in hypothetical si-constructions, where it would introduce the apodosis. Thirdly, it had, as in Modern French, the function of introducing result:~ or conclusions. And fourthly, it could be used as an emphatic particles with imperatives. The latter use of course also still exists (see below), and was in fact one possible function of Latin DUM. According to Gamillscheg (1957: 587), of the first three uses mentioned, the temporal sense is the original one, from which developed first the use in apodoses, and subsequently the resultative use (see also Stempel, 1964: 341). 7 Of these three, it is only the resultative sense which has carried over into Modem French, whereas the temporal and hypothetical uses have been entirely lost. s The evolution of the particle may then be considered a case of semantic shift (Holm, 1988: 101), whereby the meaning of an item is extended, with eventual lc,ss of the original meaning. Alors, on the other hand, originates in Latin ILLA HORA, an ablative meaning at that hour. Like done, its original meaning in Old French was that of temporal simul7 It may be relevant, however, that late Latin DUM seems already to have had a resultative sense in some contexts (cf. von Wartburg 1934). x Note that Ferrari and Rossari (1994: 33-34) adduce some examples of done inside the apodosis of a conditional. These structures, however, have certain peculiar properties which serve to distinguish them from 'canonical' conditionals.
164
M.-B. Mosegaard Hunsen / Journal q/Prugmatics 28 (1997) 153-187
taneity, sometimes, but not invariably, with a sense of duration, and again like donc it later came to be used in hypothetical si-constructions, as well as with results or conclusions. Unlike done, alors has retained both the former uses in Modem French, witness the following examples: Temporal anaphoric sense: (21) ... mais quand j'ai entendu Georges Marchais dire ~t i'usine SKF d'Ivry, on ne construira pas d'usine nouvelle en commenqant par saborder celles que nous poss6dons actuellement,, lh alors j'ai plus du tout cru ~ votre d6claration commune ... (VS2: 26) Introducing apodosis in si-construction: (14) ... je crois que si l'on commence ~ partir du principe, que y a pas d'unit6 politique possible au Liban, alors, on on est parti pour on ne sait quoi ... (VS 1 : 29) Hence, it is more appropriate to speak of semantic broadening (Holm, 1988: 101), rather than semantic shift, in this case. Now, it would appear that this broadening took place at a much later stage in the case of alors, than in the case of done. In fact, there seem to be no instances of alors with a clearly non-temporal sense in Old French texts. Two hypotheses thus present themselves. One is that donc lost its ability to appear in the types of structure exemplified in (12) and (4) because it was supplanted by alors. This would accord with the Principle of Contrast (formulated as such by E.V. Clark, cf. for instance Clark, 1988, although it is, of course, a traditional assumption in lexical semantics, cf. Cruse, 1986: 270), which states that absolute synonymy is avoided in language. The alternative is that the meaning of alors was extended to fill a gap left by the shift in the sense of done. It is worth pointing out that the general evolution of the two markers, away from the temporal sense, in which they contribute to truth-conditional meaning, and towards a metadiscursive function, provides support for Traugott's (1982: 256) hypothesis that "[i]f there occurs a meaning-shift which, in the process of grammaticalization, entails shifts from one functional-semantic component to another, then such a shift is more likely to be from propositional through textual to expressive than in reverse direction".
4. Donc: Marker of manifestness
Z6none (1981) identifies five different environments for done in contemporary French: (a) marking the resumption of something mentioned earlier, as exemplified in: (22) Donc, pour revenir ~ ce disait la dame pr~c6demment, il est difficile d'envisager une solution ~ br~ve ~ch6ance (from Z6none, 1981:118)
M.-B. Mosegaard Hanson / Journal ~f"Pragmatics 28 (1997) 153-187
165
(b) a 'discursive' use: (23) Que ta maison est d o n c jolie! (Z6none, 1 9 8 1 : 1 1 9 ) (c) an 'argumentative' use (24) B ne l'a pas lu doric il ne peut rien dire (Z6none, 1981: 122) (d) a 'metadiscursive' use: (25) C ' e s t votre point de vue d o n c que vous 8tes en train d ' 6 n o n c e r ? (Z6none, 1981: 131) and (e) a 'recapitulative' use: (26) X: paragraph describing the history o f the attribution of a natural dignity to any individual A: (following paragraph) Tout individu humain a d o n c une dignit6 naturelle que le sto'l'cisme a th6matis6e ... (Z6none, 1981: 132) In Z6none (1982: 129), she collapses these functions into two: a co-textual use, covering (c), and a contextual one, covering the rest. She considers these as instantiating two distinct morphemes, a'oncl and donc 2 . I will attempt to show that this ' h o m o n y m y ' analysis is unnecessary, and that the five functions identified above are not as clearly distinct as Z6none a s s u m e s ? Moreover, I will argue that the 'argumentative' function as such is not specified directly by the semantics of the marker, but is simply highly compatible with it. Specifically, I propose to analyze the semantics of donc in terms of Sperber and W i l s o n ' s (1986: 3 8 - 4 6 ) notion o f 'mutual manifestness', a notion which is intended by these authors to provide an alternative to the more widely used 'mutual knowle d g e ' : According to Sperber and Wilson, something is manifest to an individual if (s)he is "capable o f representing it mentally and of accepting its representation as true or probably true". In other words, it should be either perceptible or inferrable. Manifestness is much weaker than knowledge in that it admits o f degrees and allows for false assumptions. Importantly, it also allows for something to be manifest to an individual without requiring that it be currently attended to. Thus, donc occurs overwhelmingly in two broadly defined environments (which are not mutually exclusive): one. in which it marks a conclusion or a result, as in (27), and one in which it marks a repetition o f something said earlier, as in (28): ~0
In an article which has come to my aT:tentiononly after i wrote the first version of this paper, Ferrari and Rossari (1994) make the same point and interestingly, their analysis of Italian dunque has much in common with my analysis of dora'. When analysing donc, however, they follow Z~none in taking the 'argumentative' use of the morpheme as central, and they assume that its other uses are derived from this. "~ Note that when using done, speakers are not necessarily repeating themselves; allo-repetitions are found as well.
166
M.-B. Mosegaard Hansen /Journal of Pragmatics 28 (1997) 153-187
(27) A. ... c'6tait d'ailleurs tout h fait la position du RPR en 1979 qui avait ~ ce moment-l~ compl6tement refus6 une liste unique avec I ' U D F B. c'6tait pas une position & tactique && A. & qa lui a pas bien r6ussi && d'ailleurs parce que le r6sultat n ' a pas 6t6 tr6s brillant mais enfin B. non mais c'6tait une position encore une fois fond6 sur des principes, je tes ai Rappel6s tout ~ I'heure A. d o n e vous avez chang6 de principes B. non, nous n'avons pas chang6 de principes A. ah ben si puisque maintenant vous voulez une liste unique (VS2: 18) (28) A . . . . je suis all6e j e suis arrivde au village de Sainte-Barbe il 6tait dix heures sans mentir c'est-~-dire la nuit tombait, en plus Kiki m'avait dit le v61o dans les dunes t'y vas ~ pied tu peux pas rouler parce que sinon tu vas p6ter les pneus je lui dis bon d'accord mais moi qu'est-ce que j'ai fait je me suis dit oh la la dans les dunes d6sertes si j'am6ne mon v61o toute seule ben je dis ~t p / e n plus y a des tas de mecs qui passent le long de la plage tu vois B. ah ouais ouais A. SEULE en plus B. bonne ambiance A. je me suis dit oh la la non je je fais pas B. mais Sophie A. attends attends done je suis arriv6e au village de Sainte-Barbe ... (CV: 17-18) We don't invariably find word-for-word repetitions, as in (28): sometimes, the repetition may be only partial: (29) A . . . . et il tuait alors par exemple y avait euh comment elle s'appelait Myriam, qui jouait dedans qui jouait son r6le & tu vois && B. & m m mm && A. de Myriam, Collomb et puis un autre petit un tout petit Marseillais qui faisait craquer toutes les petites nanas je sais pas si tu te rappelles B. oui oui oui je vois A. & h peu prbs && B. & je sais pas && son pr6nom mais A. et bref alors d o n e ils jouent leurs r61es et tout ... (CV: 9-10) Often, done marks what is both a conclusion and a (partial) repetition: (30) ... alors nombre de roots minute,, e'est un peu c o m m e l'histoire des pauses, eourtes ou moyennes ou longues,, c'est assez difficile ~t dire hein que ce soit pertinent parce que bon comme je vous ai dit,, moi j'ai quelqu'un de la Ni~vre qui pour moi parle extr$mement lentement,, mais pour lui c'est tout h fait normal, alors c'est peut-$tre une caractEristique de la Nibvre,, du moins lui c'est ce
M.-B. Mosegaard Hansen / Journal q] Pragmatics 28 (I 997) 153-187
167
qu'il m ' a dit euh:: mais c'est s~r que d'un locuteur ~ un autre, dans la m~me langue les gens parlent pas a la m~me vitesse, y a des difficult6s moi qui travaille en situation plurilingue [...] d o n c vous voyez que: l'interpr6tation de: du nombre de mots minute, pe~at euh: c'est c'est sujet ~t caution (C: 4-5) The marker is also found in structures such as reformulations, paraphrase, and summary, which can be seen as forms of repetition ~ (31), and with requests for confirmation or questions, when these are the result of inferencing, and thus related to conclusions (32), (33): (31) A. oui le Lac Nasser j'ai lu quelque part que: au moment de sa plus grande 6tendue d o n c au & moment B. & m m A. o?a la crue est la plus fo~rte B. oui A. il fait cinq cents kilom6tres de long (VE: 29-30) (32) A. et vous &es rest6s combien de temps alors au Caire B. attends je me souviens plus on a fait deux passages oui en arrivant on a dO rester deux jours, je crois A. oui B. et puis au retour on est rest6s une journ6e A. oui B. c'est peu & en fait && A. & d o n c && vous avez fait le le, & d'abord deux jours au Caire && B. & XXX au Caire && (VE: 16-17) (33) ... mais quand j'ai entendu Georges Marchais dire a l'usine SKF d'Ivry, on ne construira pas d'usine nouvelle en commenqant par saborder celles que nous poss6dons actuellement, la alors j'ai plus du tout cru ~t votre d6claration commune, parce que saborder, les industries que nous poss6dons qa veut dire que c'est un acte volontaire le sabordage, d o n c euh qui est responsable du sabordage, le Ministre de l'Industrie? ... (VS2: 26) Thus, Z6none's exx. (22) and (26) can be subsumed under the broad category of 'repetition', and exx. (24) and (25) under that of 'result/conclusion'. Word-for-word repetitions like that in (28) above of course inherently express manifest information. Reformulations and partial repetitions, like those in (29)-(31), are certainly highly inferrable, although sometimes, as in (29), the discourse context must be taken into account, and, in my material at least, done is never used to mark results or conclusions which are in any way unexpected, given what has already transpired in the talk. ~l In fact, Tannen (1989: 54) proposes a "scale of fixity in form" for repetitions, going from exact repetition at one end to paraphrase at the: other.
168
M.-B. Mosegaard Hansen / Journal ~)f'Pragmatics 28 (1997) 153-187
This leaves us with the so-called 'discursive' use, as in: (23) Que ta maison est donc jolie! (34) Qu'ont-ils donc ~ me regarder c o m m e qa? (35) Donne-le-lui donc! (as said by a mother to her son who is fighting with his younger sister over a toy) In this type o f examples, the particle connects to extralinguistic information, but seems to indicate that the propositional content is obvious to both interlocutors. When pointing thus to the extralinguistic context, done is always integrated into the syntactic structure o f the utterance, i.e. it does not occur as an independent tone unit at the beginning of the host utterance. Moreover, it has the additional 'emphatic' function o f reinforcing the illocutionary force of the utterance. This is not surprising, inasmuch as, if the speaker considers it obvious that, for instance, an older brother should be mature enough to let his younger sister play with his toys, then one reason for uttering (35) anyway would be to allow the speaker to express insistence. In some instances, in fact, the connective component seems to take a back seat to the expression of emphasis, to the point where it may be almost absent: (36) Pourquoi donc? (37) Allons donc! (38) Dis donc, mon vieux, t'es pas dou6 pour qa ... - the latter two expressions probably being (close to) lexicalized as such. Hence, it might ultimately be more appropriate to treat donc in this 'discursive' use as a modal particle, rather than as a discourse marker, the more so as the particle tends to be pronounced without a final [k], i.e. as [d6], in such contexts. As is well known, phonetic erosion often accompanies a greater degree of grammaticalization (cf. Lehmann, 1985; Heine et al., 1991 : ch. 1). Done may thus be analyzed as indicating that the information contained in its host utterance is treated by the speaker as being mutually manifest to interlocutors.~2 This means that the relevance of the utterance should not be sought in the 'news-value' o f the proposition being expressed. ~3 Note here that the notion of relevance applies, not just to propositions, but to ostensive stimuli (of which utterances are a subset) as such (cf. Sperber and Wilson, 1987: 744). This means that the utterance of a proposition which is already mutually manifest to both speaker and hearer could nevertheless achieve relevance, either directly by for instance indicating the speaker's attitude or by the mere fact of its being uttered in particular circumstances, or indirectly by reducing processing effort. ~2 Interestingly, Kroon (1989: 238-240) analyzes Latin ERGO,whose distribution seems very similar to that of donc, along similar, albeit not identical, lines. That there should be this parallelism between Classical Latin and Modem French is of course all the more interesting, as ERGOand done are, of course, not etymologically related at all. ~3 It should be noted that, as (27) shows, a speaker may be mistaken in assuming that the information is mutually manifest.
M.-B. Mosegaard Hansen / Journal ~?fPragmatics 28 (1997) 153-187
169
If this is the correct semantic description of donc, it comes as no great surprise that the particle is particularly appropriate as a marker of conclusions in syllogisms a use which is often felt to be its canonical one (cf. for instance Weinrich, 1989: 457; Z6none, 1982:115-116): by definition, such a conclusion must be contained in the premises, and hence can convey no new information. Alors, on the other hand, is quite unsuited for this type of environment. There is a very small group of apparent exceptions to the generalization stated above. These are exemplified in (39) and (40): 14
-
(39) A. y avait combien de personnes sur la sc6ne qui faisaient qa? B. alors doric y e n avait y avait doric trois soeurs, le l'espbce de cur6 maudit qui 6tait cette esp~ce de petit mec qui s'appelait St6phane ... (CV: 4) (40) ... oui tu as tout ~t fait raison de l'indiquer le premier tome c'est le rebelle qui va doric de euh dix-huit cent quatre-vingt-dix h mille neuf cent quarante-cinq n'est-ce pas puisqu'il est n6 en quatre-vingt-dix ... (MP: 11) In (39), the information conveyed is manifestly unknown, and therefore inaccessible, to the hearer, who has to ask for it explicitly. In (40), taken from a radio debate, the topic is the first volume of a new biography of General De Gaulle. The time-span covered by this volume has in fact not been mentioned earlier in the discussion. However, in (39), the preceding discourse does contain mention of both the nuns and the cursed priest, although the exact number of the former has not been not made explicit, and in (40) the speaker's use of n 'est-ce pas and puisque indicates that, for him, the information contained in the utterance ought at least to be manifest to any educated listener. 15 Now, given the etymology and older, obsolescent uses of donc, how may we explain its evolving into a marker of manifestness? There is nothing particularly surprising about a particle signalling the simultaneous existence of two states of affairs taking on the meaning of a temporal anaphor. As for its use in conditionals, Hopper and Traugott (1993: 179) note that, cross-linguistically, conditional markers are often derived from temporals expressing duration, and they explain this by the fact that "conditionality presupposes an extant (durative) condition". Furthermore, Haiman (1978: 576) argues that then in English if... then ... constructions is not part of the meaning of the condition~tl, but rather a pronominalized form of the protasis, and he supports this by observing that resumptive personal or demonstrative pronouns may be used in the same way, as in (41) (Haiman's (42)): (41) a. If you could bring a ball, that would be good b. If you tidied up your things, it would make me very happy ~4 Examplesof this type are also noted by ZSnone (1981: 132n), who does not, however, attempt to provide an explanation of them. 15 Note also that this use of done is extremely infrequent in my material - in fact, the examples number only three out of a total of 116.
170
M.-B. Mosegaard Hansen / .lournal of Pragmatics 28 (1997) 153-187
A morpheme like donc would then be equally suited for the anaphoric resumption of a conditional protasis. And modus ponens being a rather basic pattern of inference, once an item is allowed to mark apodoses, it is only a very small step to extend its use to consequences and conclusions in general.~6 Its being apparently confined to fairly obvious conclusions, and its use with repetitions, may then be explained by the persistence of the original sense of simultaneity: in X (... Z ...) donc Y, the proposition expressed in the utterance X is seen as automatically making the proposition expressed in Y available. It is this insistence on the simultaneous availability of two propositions that makes done appropriate in analytical or quasi-analytical (given a certain cultural context) statements, such as (10') and (42): (10') C'est le cousin de ma femme, et doric mon cousin par alliance (from Jayez, 1988: 136) (42) Tu as gagn~ 10.000 francs, donc tu as gagn6 un million ancien (from Roulet et al., 1987: 149) I hope to have shown then, that donc in contemporary French is not primarily a marker of result or conclusion, as is commonly assumed (cf. Z6none, 1981; Roulet et al., 1987: sect. 2.3.4; Ferrari and Rossari, 1994), but that its semantics makes it compatible with at least some such environments. In the next section, I will argue that the same is true of alors, although this morpheme, contrary to present-day donc, will be seen to be multiply polysemous.
5. Alors: A radial category Judging by my data, alors is by far the most common discourse particle in contemporary spoken French, rivalled only by bon. One might therefore expect it to occur in a wider range of contexts than other markers, and that is indeed the case. As already noted, its evolution to some extent resembles that of donc, but I will argue that its extension to conditional and resultative environments is motivated by another set of factors, which provide for its further extension to a number of environments which are seemingly quite different from the ones mentioned. Hence, the content of this particle seems best represented as a conceptual network (Heine et al., 1991: 98) or radial category (Lakoff, 1987: 65), i.e. a category comprising central and less central members, where the latter are not predictable from the former, but nevertheless motivated by them. One important characteristic of radial categories is that there t6 This evolution was probably facilitated by the confusion between Old French done and dont (Togeby, 1974: §272.5; Reynaud de Lage, 1990: §329), a relative-interrogativeadverb with a locative meaning expressing the origin of something (from Latin DE UNDE= "from where'). As Stempel (1964: 140) notes, "[d]ie Verfknfipfungmit dont steht in unmittelbarer Nahe des Konsekutivverhfiltnisses,da DE UNDEexplizit auf die Ursache hinweist". Old French dont was, in many of its uses, equivalent to Modem French d'o~, which may also be used in consecutive structures such as the following: ll ne m'avait pas pr~venu de sa visite: d'oiL mon ~tonnement.
M.-B. Mosegaard Han3en / Journal of Pragmatics 28 (1997) 153-187
171
need be no one property which is common to all members, rather such categories are structured by chaining links from one member to another. While the original temporal sense still exists in Modern French, and thus precludes the univocal classification of alors as a discourse marker according to the definition proposed above, it is likely that it is no longer the central sense for most speakers. As a matter of fact, this sense is found in only one out of a total of 268 examples in my corpora, namely the one already adduced in (21): (21) ... mais quand j'ai entendu Georges Marchais dire a l'usine SKF d'Ivry, on ne construira pas d'usine nouvelle en commenqant par saborder celles que nous poss6dons actuellement,, 1~ alors j'ai plus du tout cru a votre d6claration commune ... (VS2: 26) Interestingly, the speaker apparently felt it necessary to reinforce the anaphoric meaning with 1~ in this example. Prototypical examples of this use, such as that in (43) seem to belong to more formal, probably mostly written registers: (43) I1 avait attendu l'arriv6e du marin, qui commandait alors le Saturnia (E. Peisson) According to the Tr~sor de ra langue fran~'aise (1973: 606), alors in this use "cr6e, par r6f[6rence] a u n proc6s d6j~ 6nonc6 ou ~t une situation connue ou suppos6e telle, une actualit6 distincte de l'actualit6 du locuteur et situe donc le proc~s dans le pass6 ou dans l'avenir". Now, Ducrot (1972: 168) insists that an utterance of the form "si p, q n'a pas pour signification premiere ' p e s t cause de q', ni ' p e s t condition de q' [...] Sa valeur fondamentale est de permettre la r6alisation successive de deux actes illocutoires: 1° demander h l'auditeur d'imaginer 'p', 2 ° une fois le dialogue introduit dans cette situation imaginaire, y affirmer ' q ' " . A hypothetical condition is by definition a 'reality distinct from that of the speaker', and it is my claim then that it is this aspect of temporal alors which allows its extension to contexts like that of (14): (14) ... je crois que si l'on commence h partir du principe, que y a pas d'unit6 politique possible au Liban, ah~rs, on on est parti pour on ne sait quoi ... (VS 1 : 29) This would seem to account for the fact that alors is, as Franckel (1986/1987: 24-25) notes, not appropriate in non-hypothetical si-constructions: (44) S'il a fait qa, (*alors) c'est pour ton bien (on an interpretation that is roughly equivalent to 'He only did that to help you') (from Franckel, 1986/1987: 24) On the required reading, the si-clause of (44) is not intended to create an imaginary situation - the proposition expressed is already known to be true by both speaker and hearer. The insertion of alors would impose a hypothetical reading, and is therefore inappropriate.
172
M.-B. Mosegaard Hansen / Journal of Pragmatics 28 (I 997) 153-187
By thus anaphorically referring to a 'distinct reality' established in the previous clause, one which sets the stage for the information conveyed by its host utterance, alors may be said to have a re-perspectivizing or reorienting effect. This seems compatible with Franckel's (1986/1987: 26) description of the semantics of the particle: "Etant donn6e une relation pr6dicative P construite b. partir d'un premier rep~re SITi, alors constitue comme rep~re SITj d'une seconde proposition Q la reprise d'un rep6re associ6 ~ P, tout en 6tablissant une forme de disjonction entre SITi et SITj [...] alors constitue SITj comme fronti6re de l'espace 6nonciatif associable ~t SITi". However, assuming that I have correctly understood what he means, Franckel's account appears to imply that alors is always anaphoric, and that, I believe, is hard to maintain for many of the uses to be considered below. In any case, the re-perspectivizing or reorienting effect of alors provides for one major class of extensions, whereby its overall distribution comes to differ markedly from that of donc. ~7 5.1. Alors marks re-perspectivization o1" reorientation
Alors is not infrequently used to mark shifts to new topics, particularly subtopics or digressions. Although, in this type of use, the marker cannot meaningfully be called an anaphor, its original anaphoric properties probably helped to license the extension, since, as pointed out by linguists in the Conversation Analytic framework, there is a fairly strong pressure in interactive talk to make new topics seem to grow out of and cohere with old ones (cf. Heritage, 1989: 28). A few examples of relatively abrupt topic shifts introduced by alors may be found in debates, but these are on the whole attributable to the moderator, who of course has special privileges in this respect, and are usually effected by means of a metadiscursive comment (see below). The following instances are less blatant: (45) A . . . . apr~s tout le monde se balanqait des trucs en pleine gueule et tout tu vois qa craignait bien & tu vois B. & oui A. et alors bon B. c'est g6nial ~a A. donc aprbs qa a dur6 comme qa alors si tu veux le probl6me c'est que moi je faisais le narrateur j'assurais pas du tout ... (CV: 8) In this excerpt, A is telling B about a play staged by himself and the children attending a summer-camp where he worked as a counselor. After having outlined the plot, he shifts the topic to that of his own role in the play, but stays within the same overall topical framework. (46) A . . . . y a encore plein de zones bien stir qui ne sont pas 61ucid6es, puis y a des zones d'ombre parce que & pour euh: parce que && 17 Note that I am of course not claiming that Old French done, when used in contexts similar to (21), (43) and (14), did not also have this type of effect, merely that this was not the aspect of its semantics which was apparently focused on in extending its use to conclusions and repetitions.
M.-B. Mosegaard Hansen / Journal of Pragmatics 28 (1997) 153-187
173
B. & y a des 6poques qu'on conna~t plus mal, des p6riodes qu'on conna~t plus mal && A. on a vu par exemple § pour euh: B. § m m A. comment elle s'appelle, Hatchepsout B. oui A. mais c'est qu'on a marlel6 son visage partout B. m m A. euh ils ont fait dispara]tre un tas de textes ou de choses qui la concernaient donc y a des zones d ' o m b r e lh terribles, et puis alors un t r u c / o n n ' a rien vu, dessus mais on a vu simplement au mus6e, c'est euh: l'6poque ils appellent qa L ' H E R E S I E : : amarnienne, avec comment il s'appelle euh: B. Akh6naton A. ouais, alors IL bon il a fix6 sa capitale beaucoup plus haut & B. & m m A. et enfin plus vers le Nord et: nous on n'y est pas all6s, c'est Amarna on n'est pas all6s (h) d'abord il para~t q u ' y a TRES peu de restes ... (VE: 55-56) Here, alors introduces an additional illustration of A ' s general statement about lesser known areas of Egyptian history. This is clearly a subtopic, which closely parallels the one preceding it. (47) A. ... 6videmment la vie du G6n6ral on pouvait penser qu'on savait tout 1~ encore une fois l~t on pouvait croire que vraiment euh si y a bien un homme sur lequel on s'est pench6 cela dit on s'6tait jamais pench6 dans sa continuit6 en France sauf Franqois Mauriac qui alors lui avait fait un livre tout fait euh: hagiographique pour reprendre le mot de tout h l'heure qui avait valu un pamphlet tr6s r6ussi d'ailleurs de Jacques Laurent qui s'appelait Mauriac sous De Gaulle B. pour lequel il a 6t6 condamn6 d'ailleurs
A. B. A. B. A.
pour lequel il a 6t6 & condamn6 && & il a 6t6 condamn6 pour offense au chef § de l'Etat §§ § et §§ c'est pas la meilleure page de l'~euvre de Mauriac donc non mais c'est les meilleurs de celle de Laurent absolument et on peut dire que y avait rien en France sur ie G6n6ral de Gaulle ... (MP: 11)
In this example, taken from a radio debate, we find a digression. The current discourse topic is a new biography of General De Gaulle, and there is never any doubt that this topic will eventually be; resumed. Alors thus functions as what Polanyi and Scha (1983: 264-265) refer to as a PUSH-marker, i.e. a device which signals movement into an embedded discourse unit, and which these authors describe precisely as "forc[ing] an interpretation of a clause in a 'frame of reference' other than the one appropriate to the immediately preceding clause".
174
M.-B. Mosegaard Hansen / Journal of Pragmatics 28 (1997) 153-187
Closely related to these are examples where alors signals the return to a previous topic, most commonly, as in (48), by effecting what Fox (1987: 27) calls a 'return pop', i.e. by tying its host utterance not to the immediately preceding one, but to one that occurred earlier in the same on-going sequence, or in rare cases, as in (49), by connecting to a sequence which has seemingly been closed off at an earlier stage. (48) T1 A.
... et il m ' a jamais dit Sophie ouais bon t'as tel petit probl6me qa va pas trop bien tu devrais faire comme ~a jamais j ' a i ,IAMAIS rien eu T2 B. non mais qa il le dit ~t & personne && T3 A. non && si si il il l'a dit ~t Didier T4 B. pas en question animation et pas en question & de rien du tout && T5 A. & mais si && parce que Didier me l'a r6p6t6 apr~s T6 B. ah qa m'6tonnerait T7 A. non je t'assure je te mens pas hein il a dit qa & alors moi j'ai && T8 B. & qa m'6tonnerait && T9 A. non non ne ne crois- & moi vraiment T10 B. & parce que nous il nous a jamais aid6 de cette mani6re-l~t T l l A. alors moi j'ai jamais su que finalement j'6tais prise pour deux mois ... (CV: 25-26)
In this extract A makes two attempts, one unsuccessful and one successful, and both by means of alors, to resume her account of the troubles she had with the leader of the summercamp where she has been working. The example differs from the previous one insofar as the sequence beginning with B's first reply in turn 2 and ending with turn 10 is only retroactively embedded, and does not initially involve any topic shift. This demonstrates the dynamic nature of informal talk and underscores the importance of local management. Note that in all the examples adduced so far, alors is accompanied by a left-dislocated element (45, 48), a nominativus pendens (46), or an emphatic pronoun (47), w h i c h help to accomplish the reorientation inherent in topic change. Similar devices (including fronted spatial and temporal adverbials) can in fact be found in a majority of examples of this type. (49) ... et ensuite, nombre de mots minute,,, et enfin divers, alors nombre de mots minute,, moi je suis/ c'est un peu comme l'histoire des pauses, courtes ou moyennes ou longues,, c'est assez difficile h dire,, hein, que ce soit pertinent [...] c'est comme l'histoire des" pauses il y a des choses qui sont relatives c'est pas absolu mais c'est ~ vous de moduler, en fonction de ce que vous savez du locuteur, de la situation, de son attitude, de la particularit6 de sa rdgion ou sa langue, alors vous avez enfin en dernier un exemple de texte transcrit comme qa hein ... (C: 4-7) The example is taken from a university lecture on transcription practices. The initial et ensuite, hombre de roots minute,,, et enfin divers refers to a handout containing a list of factors which should be noted when making a transcription. The enfin divers seems to indicate that the speaker has come to the end of the handout. The
M.-B. Mosegaard Hansen /,lournal of Pragmatics 28 (1997) 153-187
175
alors following it (in italics) effects a 'return pop' to the preceding hombre de mots minute, which is subsequently commented on through three and a half pages of text, with no references to the handout, until a tie back to the latter is finally achieved by the mere use of the second alors (note that the pause before this alors is very brief, suggesting that the marker is quite a powerful reorienting device). Whether or not it is appropriate to speak of 'retroactive embedding' here, too, is an open question, since the return, despite the enfin en dernier which ties back to the previous mention of the handout, does not actually lead to a resumption of the previous topic, but rather to a topic transition. As a matter of fact, examples such as the last two raise the important question of whether 'embedding', although a useful descriptive notion, should be accorded any theoretical status in discourse analysis. While speakers clearly orient to topical coherence, it is much less obvious that they orient to the structure of the discourse as such. It is highly likely that whatever structure may be observed post hoc by analysts is in fact an 'emergent', rather than an a priori one (cf. Hansen, 1996b: ch. 6 for a detailed argument to this effectl. ~ If this is so, the notion of discourse structure might then more usefully be seen as derivative of extralinguistic pragmatic factors such as the already mentioned topic coherence, politeness considerations (for instance in the case of so-called pre-sequences (Levinson, 1983: 345ff.)), global interactional goals, and the like. In any case, examples like the last three could form the basis for using alors as a bracketing device (PUSH- and POP-markers in Polanyi and Scha's (1981: 264-265) terminology), ~ either preceding or following another type of discourse 'embedding', namely parenthetical comments or asides: (50) ... et il me semble que c'esl dans celui-ci/elle nous a / s u r des: des colonnes, y a vraiment, des cartouches, alors assez profonds avec des dessins, r6p6titifs, alors je sais plus si c'est ie crocodile,, ou,, ouais avec des t&es et vraiment pour l'instant c'est une 6nigme ... (VE: 54) (51) A . . . . j'ai lu un peu en rentrant parce que qa stimule un peu qa, et puis j'ai vu que: apparemment c'est vrai ils disaient que c'6taient les paysans pendant les p6riodes & de crues ouais && B. & les grands travaux && c'6tait qa A. ouais ouais qa qa j'ai l,~ ouais, alors apparemment bon c'est peut-~tre vrai elle avait peut-~tre raison quand m6me (VE: 43-44) Related to the topic-shifting t~ses are ones where alors signals shifts of frame in the sense of Goffman (1974), for whom a 'frame' is an interpretive perspective imposed on a strip of experience (including talk) in order to make sense of it. This is ~ The present formulation is obviously inspired by, although not identical to, and does not necessarily imply an endorsement of, Hopper's (1987) notion of "emergentgrammar'. t9 In view of the preceding paragraph, my use here of Polanyi and Scha's terminology should of course not be seen as a commitment to their overall notion of a "syntax of discourse'.
176
M.-B. Mosegaard Hansen / Journal of Pragmatics 28 (1997) 153-187
not identical to the notion of perspectivization employed above, insofar as Goffman, as a sociologist, is concerned to elucidate the mechanisms by which we identify and/or transform the nature of what we experience, thus, in the case of talk, whether it should be interpreted as serious, playful, ironic, etc. Change or resumption of a discourse topic, or marking an utterance as being relative to a past or imagined world, of course does not alter the nature of the talk as such. Now, one very frequent use of alors, especially in debates, is initial or final bracketing o f metadiscursive utterances, 2° which by definition are situated on a different plane of talk, and thus constitute frame-breaks: (52) A. B. A. B. A. B. A. B.
ben elle 6crit elle 6tait journaliste elle le fait dans l'actualit6 oui hein non mais vous avez l'air de iui reprocher mais moi je reproche rien vous voulez qu'elle fasse une biographie de Barre pr6ventive? bon alors on p e u t p a r l e r on peut p a r l e r du contenu? (MP: 4)
(53) A . . . . moi je le rEpbte aujourd'hui nous sommes contre l'61argissement de la Communaut6 Economique Europ6enne, et: euh: tout ~t fait r6solument contre & son 61argissement & & B. & contre l'Espagne et le Portugal & & § contre l'entr6e de l'Espagne et du Portugal §§ A. § contre l'entr6e de l'Espagne et du Portugal §§ oui oui C. vous vous mettez d ' a c c o r d avec les Socialistes A. j e le dis trOs tranquillement p o u r ee qui est de l ' a c c o r d entre les Socialistes et nous laissez-nous en discuter entre nous et ne vous en m~lez pas trop p a r c e que ~'a vous r~ussit p a s vraiment C. mais ~'a m ' a m u s e tellement A. alors euh: euh: donc nous sommes contre l'61argissement ... (VS2: 8-9) T w o important, and closely connected, aspects of the framing of talk in Goffm a n ' s sense are the identification o f the source of an utterance, and the status of participants (Goffman, 1974: 5 1 6 - 5 2 3 ; the latter referred to as 'footing' in Goffman, 1981). 2j Alors is often found iotroducing reported speech, especially when exchanges are being reported, making it all the more essential for the hearer to identify exactly who said what: (54) ... parce que j ' tu m ' a s dit euh: tu attendais mon coup de t616phone et alors moi je t'ai dit mais non j'6tais en train de dormir (CT: 4) -'t~ In identifying metadiscursive utterances and sequences, I have relied on Morel's (1985) taxonomy. 2~ Readers familiar with Ducrot's theory of 'polyphony' (Ducrot et al., 1980; Ducrot, 1984: ch. VII) will no doubt recognize a certain affinity between the latter and Goffman's ideas, an essential difference being that, for Ducrot, "polyphony' inheres in the language itself.
M.-B. Mosegaard Hansen / J o u r n a l ~/" Praematics 28 (1997) 153-187
177
(55) A. ... et elle m ' a dit euh: ne me ne dis pas merci c'est vrai 22 B. non mais allez vas-y continue A. que euh: ce qui lui semblait bizarre c'est que quand elle t'a vue elle avait l'impression qu'elle t'avait d6j~ vue quelque part B, ah bon? A. m m mm, et elle m ' a dit alors je lui ai dit mais comment qa? (CT: 5) In (54), we have a reported exchange between the current speaker and hearer, and the pronouns je and tu are thus ambiguous. In (55), A is about to continue reporting what the third party has said, but changes her mind and uses alors to make the switch to her own reply. In both these examples, the change of speakers is of course redundantly specified, since alors is accompanied by verba dicendi in both cases. 23 However, that the marker does indeed signal that the following utterance should be attributed to a different source is supported by (56), where alors is the sole item which linguistically marks the utterance as reported: (56) A . . . . alors pour eux, pour eux les Coptes c'est en meme temps que Isis Osiris B. ah oui,, & ben oui && A. & ils ont beau && ils ,ant beaucoup de mal alors "mais les Chrdtiens ils vivaient comment ~ l ' 6 p o q u e ? " (VE: 63-64) 24 Similar to these examples is the following, from a university lecture: (57) ... vous verrez que l'oral, quand on le transcrit, euh: met en 6vidence que les gens ne parlent pas d'une fa~;on trbs construite, il y a dnormdment de reprises, ils s ' y prennent h plusieurs lois pour construire un dnonc6, "oui alors comment peut-on", alors lh c'est un cas o?a on ne sait pas ce qu'il a dit "en tirer ou attirer", hein, 1"~ o~ on avait mis deux X s6pards par une virgule dans les conventions 1~ vous avez un K hein on n'arrive pas apr6s "a l'enregistrement savoir s'il a dit "oui alors comment peut-on en tirer" ou, "alors comment peuton attirer",, "des: des st6rd,atypes en particulier" alors 1~. hdsitation entre 'des' et zdro, "en particulier stdrdotypes valables et ajustables h l'esp~ce humaine'" et puis alors l~t deux syllabes inaudibles, qui se sont chevauchdes avec le d6but du deuxi~me locuteur, c'est pour qa que c'est soulignd ... (C: 8-9) The speaker is alternately reading aloud from a transcription and commenting on it. At the end of all three quotations, he marks the transition to himself as the source of the upcoming utterance with ¢tlors.
22 Note that the ne me ne dis pas met+c; c'est vrai is an aside and hence not part of the quote. 22 Redundancy of information is, in fa,zt, both c o m m o n and desirable in (especially) informal speech, where it aids the hearer in coping with tile pressure of on-line processing. 24 Although, strictly speaking, one ou~tht not to add quotation marks to a transcription, as they obviously are not to be found in speech, I have nevertheless taken the liberty of adding them here, and in the following example, to make it clear to tile reader what is and is not quotation.
178
M.-B. Mosegaard Hansen / .lour,al ~f Pragmatics 28 (1997) 153-187
Another kind of quotation is found in so-called 'diaphonic' structures (Roulet et al., 1987: 70), where speakers incorporate some portion of their interlocutor's talk into their own: (58) A. B. A. B. A.
... je pense que y a probablement un [sic] espbce de: de palier "~ franchir & & oui au niveau § intellectuel §§ § et certains ne Font §§ pas encore franchi en sixi6me & ouais && & alors & & certains ne I'ont pas franchi mais ce que je constate moi ... (VE: 65)
(59) A. ... c'est tout "~ fait la lecture la seule lecture l'unique lecture possible puisque il existe un texte et y a qu'une faqon de le lire B. bon alors & euh:: && C. & Henri Amouroux && B. mettons qu'il n'y air qu'une lecture et il y a plusieurs interpr6tations ... (VS2: 22) In all such instances found in my material, the discourse move seems to incorporate concession. Instances such as these probably explain why alors is more felicitous than done in the reductio-type argument found in (11'), already adduced above: (11') Tu sais tout, alors donne-moi le tierc6 (from Roulet et al., 1987: 151) In such structures, the premise is always at least implicitly attributed to someone other than the speaker (in casu the hearer). In broadcast debates, alors sometimes marks a change in the status of addressees, as in the following: (60) A . . . . et en tout cas sont 1"~ physiquement pr6sents et je les en remercie vu l'6poque ofa nous enregistrons cette 6mision c'est-'~-dire en plein weekend de la Toussaint Fran~oise X6nakis qui reprdsente Le Matin et FranceMusique car aussi Frangoise Xdnakis est aussi sur France-Musique et Jdr6me Garcin que vous avez bien stir reconnu qui est h FR3 mais aussi qui vient de signer ses premi6res collaborations au journal que vous avez tous lu et qui est sorti y a trois jours et qui s'appelle L'Ev~nement c'est bien qa L 'Evdnement du." jeudi B. L'Ev~nement du jeudi A. oui alors c'6tait:: c'fitait comment, on en parle d6j'~ h l'imparfait (MP: I) (61)
... nous sommes en direct de Bourg-en-Bresse ofa se d6roule vous le savez le congr6s du Parti Socialiste domin6 cet aprbs-midi par l'intervention de Lionel Jospin que vous avez entendu tout ~ l'heure, Lionel Jospin qui a interpell6 trbs directement le Patti Communiste,, lui a demandd en quelque sorte de foumir des explications sur les divergences exprim6s depuis un moment et puis cet
M.-B. Mosegaard Hansen /,hmrnal of Prugmatics 28 (1997) 153-187
179
aprbs-midi aussi, ofJ ici on a r6ussi un peu h s'informer sur cet article du Monde qui faisait 6tat peut-~tre 6tal d'fime de Monsieur Fiterman, alors Roland Leroy sans vous demander peut-&~:e tout de suite j'allais dire le contenu de l'6ditorial de L'Humanit~ de demain matin, j'ai tout de m~me envie de vous demander si vous avez pass6 un bon apri~s-midi (VS 1 : 1) Although the participants in a debate, the listeners at home, and in (60) the studio audience, are all ratified listeners, they are not usually all being directly addressed at the same time. Here, the moderator starts out in both cases by addressing the studio audience (60) or the listeners at home (61), and then abruptly switches to one particular active participant, marking the switch with alors. 2~ Note especially that in (60), B is explicitly not addressed in the opening remarks, as he is being referred to in the third person. Many, although not all, of the uses reviewed in this section correspond to the uses described by Bouacha (1981) in an article about alors as a meta-operator, i.e. an item which connects discourse sequences, in pedagogical discourse. The framing hypothesis pqt forward here, however, allows us to give a more plausible explanation of what he terms "la fonction d'attaque de discours" (Bouacha, 1981: 44): here the marker is used at the very beginning of a lecture, and, according to Bouacha, it "permet de poser [...] une op6ration d'anaphorisation qui reprend un 6nonc6 du cours/discours pr6c6dent" (1981: 45). Indeed, the example he gives starts as follows:
(62) ... bon alors/je vous avais dit qu'on allait//parler maintenant un peu plus de /classification//heu ... des constructions (from Bouacha, 1981: 45) Here, the speaker does explicitly refer to the preceding lecture. Nevertheless, it would seem that (s)he might equally well have used the alternative formulation in (62"), where there is no such reference: (62') ... bon alors aujourd'hui nous allons parler de classification des constructions Now, Goffman (1974: 251) notes that "[a]ctivity framed in a particular way especially collectively organized social activity - is often marked off from the ongoing flow of surrounding events by a special set of boundary markers". It is clear that in classrooms, as in many loci of 'collectively organized social activity', the lecture itself is not co-terminous with the interaction between those present (both teacher and students have to enter and leave the classroom, sit down, take out their books etc., and some amount of talk is likely to take place during this time), so alors in this use is not anaphoric, but simply a device which allows the speaker to signal the shift to a 'lecturing frame'. 25 Unfortunately, 1 cannot say whether the marker is also used in this way off the radio, as the only multi-party recordings that are at my di,,;posal are precisely radio debates. However, it seems that there may be a special need, in radio talk, to ~,.ignalsuch changes in participant status linguistically, since the listeners at home cannot rely on body posture, gaze direction and the like to identify the addressee.
180
M.-B. Mosegaard Hansen /Journal qf"Pragmatics 28 (1997) 153-187
Having come to the end of this particular chain of extensions, we are now in a position to explain the ordering constraint on co-occurrences of alors and done mentioned in Section 3 of this paper, and exemplified by (20): (20) ... c'est moi qui doit commencer bon (h) alors donc il s'agit de la biographie de MONsieur le Pr6sident de la R6publique (MP: 4) In fact, whenever the two markers co-occur in my data, the alors is of the re-perspectivizing or re-framing type, and done marks a repetition. If we limit ourselves to the example given, the host utterance embodies both a resumption of a previously announced topic after a digression and a return to an 'unmarked' frame of talk after a metadiscursive exchange, and the information it conveys has already been given earlier. Signalling the shift in perspective before the assumed manifestness of the content would seem simply to be the most cooperative strategy for a speaker who wishes the addressee to be able to construct a mental representation of the discourse with a minimum of effort. In terms of scope relations, in other words, the mutual manifestness signalled by dora' is part of an utterance that constitues a new departure, and not vice versa. If we now return to the anaphoric temporal and hypothetical uses of alors, which were claimed to form the basis for the uses reviewed above, we see that they also provide for another, smaller, class of extensions. 5.2. Alors marks Joregrounding When functioning anaphorically, alors has as a major function that of casting previously introduced material in the role of background for the information contained in its host utterance, and thereby foregrounding the latter. This seems to be the basis for extending the use of alors to marking (sometimes gradual) transitions from more backgrounded to more foregrounded material, especially, but not exclusively in narrative: (63) A. B. A. B. A. B. A.
... en franqais ils ont ils travaillent sur l'Egypte aussi, & et && & mm c'est bien qa && ,, euh: ils sont avec Rougy la & Jean-Guillaume Rougy mm, m m et alors il leur a parl6 des Coptes ah oui, & les Chr6tiens coptes && & alors pour eux & & , pour eux, les Coptes c'est en m6me temps que Isis Osiris (VE: 63)
(64) A . . . . ils avaient tous une portion de fric qu'6tait 6gale au d6but et puis ils le gagnaient ils l'amassaient et & tout qa && B. & voil~ && c'est qa A. voil~ et moi j'6tais au 421 tu vois je faisais jouer tout le monde alors quand les m6mes faisaient 421 ils avaient gagn6 tu vois alors y avait Kader tu vois qui gagnait sans arr6t ... (CV: 11-12)
M.-B. Moseguurd Hanson / Journal of Pragmutics 28 (1997) 153-187
181
(65) ... la semaine dernibre on a beaucoup parle de la declaration commune et la rencontre entre Communistes et Socialistes j'avais demandE la semaine derni~re Roland Leroy s'il n ' y avait, effectivement qu'une lecture ou s'il y aurait deux lectures possibles de cet accord, alors c'est une question qu'on se pose, aujourd'hui encore apr~s les declarations de Georges Marchais hier ... (VS2: 20) Such uses are, of course, not totally unrelated to those analyzed in the previous section. In (63) and (64), we find shifts in sentence topic within the same overall discourse topic, and in (65), we see a shift from past to present reference. AIors could then be considered to mark minor changes in perspective. The foregrounding use of alors seems to provide for a small number of clauseinternal uses, where the particle ~,;eems to mark, and foreground, a constituent rather than the whole message: (66)
... et ii me semble que c'esl dans celui-ci/elle nous a / s u r des: des colonnes, y a vraiment, des cartouches, alors assez profonds avec des dessins, rEpEtitifs ... (VE: 54)
(67) A. ... mais en tout cas c'est tr6s amusant h lire parce que c'est vrai que c'est un Malraux en veritE dans des entretiens qui sont pas vraiment des entretiens & Roger StEphane && B. & mais dans le genre petite && biographie alors euh tr6s concise et tout c'est les portraits de Madeleine Chapsal (MP: 32) In these uses, alors seems to have a function akin to that of a 'pure' focus particle, and would not be classified as a discourse marker according to the definition proposed in Section 2.3 of this paper. Such examples are, however, quite infrequent in my data. Finally, we have a third group of extensions, namely the uses of alors as a resultative marker. 5.3. Alors marking results or conclusions
It is not clear from the literature whether the extension from temporal to conditional uses of alors preceded, followed, or was perhaps simultaneous with the extension to resultative structures. Both are, in fact, equally plausible. If the marker was first extended to conditional environments, this can, as in the case of donc, probably be justified by the existence of certain fundamental patterns of inference, in conjunction with the backgrounding function of temporal and conditional anaphora: when one state of affairs is seen as forming the background for another, the assumption will frequently be made that the former is also the cause of the latter. On the other hand, the semantic broadening to resultative environments may well have been an independent one. As Gerecht (1987:73) points out, it is not easy to tease apart the temporal and the resultative sense in certain examples, such as the following:
182
M.-B. Mosegaard Haplsen / .hmrmd
~/' Pragmatics 28 (1997) 153-187
(68) Jean a tir6. Alors Pierre s'est 6croul6. When the marker connects clauses with perfective verbs, the original sense of simultaneity easily gives way to one of sequentiality, inviting the inference that the event expressed in the first clause is the cause of that expressed in the second (we are, in other words, dealing with the common post hoc ergo propter hoc fallacy). AIors is found in resultative structures on both the content, the epistemic, and the speech act level (cf. Sweetser, 1990:11 et passim). Content level: (69) A. et t'as 6t6 au rendez-vous avec ta soeur? B. ben oui je l'ai attendue euh: je suis arriv6e une demi-heure plus t6t alors j'ai dfi attendre pendant une demi-heure (CT: 4) Epistemic level: (70) ... car le Liban est aujourd'hui quelque chose de parfaitement artificiel il faut rappeler qu'il est occup6 en partie par les Isra6liens et il faut rappeler que le Liban ce/le Pr6sident Gemayel ne r6gne que trbs difficilement et ~ peine, sur Beyrouth, alors, ce n'est pas une situation qu'on puisse trancher trbs rapidement (VS 1 : 27) On this level, the marker is often found with statements made on the basis of inference from prior discourse by the interlocutor, and which therefore usually function pragmatically as requests for confirmation: (71) A. mais Estier c'est peut-&re de la mauvaise tactique 61ectorale, car deux listes, une du RPR et une de I'UDF peuvent faire, 53 ou 54%, 55% tandis qu'une liste unique, qui va susciter h c6t6: des listes de faible importance qui peut faire que 48 ou 49 qa serait donc, une mauvaise tactique & 61ectorale && B. & vous ~tes pour && deux listes alors (VS2: 16) Interestingly, whenever alors is utterance-final in my corpora, it functions in this way. 26 The examples I have are too few to allow generalizations, but they suggest that utterance-final alors may be a candidate for grammaticalization as a modal particle, indicating a lower degree of speaker commitment. Speech act level: (72) A. y a ton oncle? B. non 2~, Note that I consider a/ors to be utterance-final, only if it forms an intonational unit with the message preceding it: thus, (73) and (74) below are not examples of utterance-final alors in this sense.
M.-B. Mosegaard Hans~n / Journal qf Pragmatics 28 (1997) 153-187
183
A. alors pourquoi t'es 6clal6e (CT: 2) Alors marks A's second question as being a result of B's answer to the first. Note that without alors, there would not necessarily be any direct connection between the two turns (A might, for instance, be reacting to B's starting to laugh immediately after having answered the first question). On all three levels, the conclusion may sometimes be left implicit:
(73) A .... et h Etampes c'est ju~,;te avant? B. Etampes c'est juste avant? avant quoi? A. avant euh: Saclas B. ben Etampes c'est la: le terminus, alors euh: A. ah, non je connais pas hein (CT: 5) Related to these are examples where alors is used with rising intonation, to ask about a conclusion which does not seem to be forthcoming: (74) A. ... puisque nous sommes ~ la veille du: l'attribution du Goncourt & quel && B. & ah && A. en est votre pronostic § pour demain §§ C. § je crois que c'est 6vident §§ A. oui, alors? (MP: 40) In all of these uses, the marke, r retains some of its original anaphoric properties, and is again not completely divorced from the re-perspectivizing use. Thus, Z6none (1982: 117) notes a difference betwen alors and done in resultative structures such as the following: (75) I1 pleut. Donc/alors je ne vais pas me promener (from Z6none, 1982:117) According to her, alors seems, in contrast to done, to indicate that the decision not to go for a walk is a subjective one. In the context of the present analysis, it is perhaps preferable to say that alors marks the conclusion as valid only in the particular perspective introduced in the previous utterance. Thus, if the two parts of (75) are uttered by different speakers, the: use of done becomes much less felicitous (cf. also the comments on example [ 11'], Section 5.1 above): (75') A. I1 pleut. B. Alors/?donc je ne vais pas me promener. Here, speaker B is relying on A's report in making her decision. The difference in acceptability between the two particles is entirely consonant with the 'canonical' use of alors in logical implications, and of donc in logical conclusions: in the former, nothing is presupposed about the truth of neither the protasis, nor the apodosis,
184
M.-B. Mose~aurd Hal ~e, / .Iourual qf Pra~mutics 28 (1997) 153-187
whereas in the latter, the premises are assumed to be true, and hence, the truth of the conclusion follows of necessity. 5.4. The semantic network of' alors
We can now draw up a semantic network for alors as in Fig. 1 (numbers refer to examples instantiating each particular use): source (56)
change in participant status
modal particle ? (71 ) result/conclusion (69)
frame break (52)
/
A
temporal (21) - -
conditional (14)
addressee (60)
topic change/resumption (45)/(48)
\
\
parenthesis (50)
foregmunding (63) focus particle? (66) Fig. I
It is important to note that the uses represented by the different nodes and branches in the network are not mutually exclusive: alors frequently seems to cumulate more than one function. This should not be too surprising if we accept that we are dealing with a polysemous item, rather than with a string of homonyms. A second caveat is that, apart from the initial extension from temporal to other uses, I do not wish to make any very definite claims about the precise path of the particle's diachronic development, as the necessary documentation is not available. The network is intended only, as Lakoff and Brugman (1986: 451) put it, "to account for synchronic connections in the semantic knowledge of the user". I do agree with Lichtenberk (1991 : 506) that, inasmuch as the idea of a semantic network of the kind I have proposed embodies the claim that "semantic/functional change proceeds by minimal steps", historical development ought not to be ignored. However, since a great many of the uses of alors reviewed in this paper are really only typical of spoken interaction, a diachronic study is hardly feasible. Possible alternative strategies (pointed out by Lichtenberk (1995: 506) and NOlke (p.c.)), of which I have for the time being chosen not to avail myself, might be to look at regional differences or language acquisition, but this is a matter for future research. One hypothesis that may be made with reasonable confidence is that the extensions to conditional and consecutive environments are prior to the foregrounding and more generally re-perspectivizing ones: this is based on the fact that these uses are the only ones which seem generally acceptable in written standard
M.-B. Mosegaard Hansen / Journal ~f Pragmatics 28 (1997) 153-187
185
French, which as is commonly known is a great deal more conservative than the spoken variety.
6. Conclusion In the preceding pages, I have, first, outlined a definition of the functional category of discourse markers, which, although illustrated with examples from French, is proposed as cross-linguistically valid. This definition took its point of departure in previous work, but suggested certain revisions which seem to be more compatible with the linguistic facts. Secondly, I have argued for an alternative analysis of the adverbs donc and alors. According to this analysis, done is basically a monosemous marker of mutual manifestness, in the sense of Sperber and Wilson (1986). Alors, on the other hand, is multiply polysemous, and is best represented as a radial category. The bulk of its uses may, however, be described as involving some form of shift in perspective, on various levels of communication. This suggests that, despite their seemingly heterogeneous distribution, there is no need, for either adverb, to postulate a set of two or more essentially unrelated homonyms, one of which would then be inherently argumentative. Rather, each can be given a single lexical entry. A further argument in favor of the account given here is that it allows us to explain certain distributional puzzles concerning the two markers. 1 have attempted to show how current uses of both markers may be derived metonymically from their respective diachronic sources, by highlighting certain aspects of meaning, while backgrounding others. As noted by Hopper and Traugott (1993: 86), metonymic change is indeed usually involved in "the search for ways to regulate communication and negotiate speaker-hearer interaction". I have, moreover, in passing, suggested possible future grammaticalizations of alors in particular. Finally, I hope to have shown how discourse markers, although often considered by purists to be redundant, and even a sign of poor elocution, nevertheless fulfill an important function in the production of discourse, not only by signalling to hearers how utterances relate to their co- and context, but also by allowing speakers to structure and restructure their messages incrementally.
References Blakemore, Diane, 1987. Semantic cons:raints on relevance. Oxford: Blackwell. Bloch, Oscar and Walter von Wartburg, 1968. Dictionnaire 6tymologique de la langue fran~:aise. 5th ed. Paris: Presses universitaires de France. Bouacha, Abdelmadjid Ali, 1981. AIors dans le discours p6dagogique: Epiph6nom~ne ou trace d'op~rations discursives? Langue franqaise 5,3: 39-52. Bybee, Joan L., 1988. Semantic substance vs. contrast in the development of grammatical meaning. Proceedings of the annual meeting of the Berkeley Linguistics Society 14: 27-264. Clark, Eve V., 1988. On the logic of contrast. Journal of Child Language 15: 317-335. Cruse, D.A., 1986. Lexical semantics. Cambridge: Cambridge University Press.
186
M.-B. Mosegaard Hansen / Journal ~/ Pragmatics 28 (1997) 153-187
Ducrot, Oswald, 1972. Dire et ne pas dire. Paris: Hermann. Ducrot, Oswald, 1984. Le dire et le dit. Paris: Editions de Minuit. Ducrot, Oswald et al., 1980. Les roots du discours. Paris: Editions de Minuit. Ernout, A. and A. Meillet, 1967. Dictionnaire 6tymologique de la langue latine. 4th ed. Paris: Klincksieck. Ferrari, Angela and Corinne Rossari, 1994. De dora "h dunque el quindi: Les connexions par raisonnement inf6rentiel. Cahiers de linguistique franc;aise 15: 7-49. Foolen, Ad, 1993. De betekenis van partikels. Dissertation, University of Nijmegen. Fox, Barbara A., 1987. Discourse structure and anaphora. Cambridge: Cambridge University Press. Franckel, Jean-Jacques, 1986/1987. Alors - alors que. BULAG 13:17-49. Fraser, Bruce, 1990. An approach to discourse markers. Journal of Pragmatics 14: 383-395. Gamillscheg, Ernst, 1957. Historische franz6sische Syntax. Ti.ibingen: Niemeyer. Gerecht, Marie-Jeanne, 1987. Alors: Op6rateur temporel, connecteur argumentatif et marqueur de discours. Cahiers de linguistique fran~aise 8 : 6 9 79. Goffman, Erving, 1974. Frame analysis. Boston: Northeastern University Press. Goffman, Erving, 1981. Footing. In: Forms of talk, 124-159. Oxford: Blackwell. Haiman. John, 1978. Conditionals are topics. Language 54(3): 564-589. Hansen, Maj-Britt Mosegaard, 1995a. Puis in spoken French: From time adjunct to additive conjunct'? Journal of French Language Studies 5( I ): 31-56. Hansen, Maj-Britt Mosegaard, 1995b. Marqueurs m6tadiscursifs en franc~:aisparl6: L'exemple de honet de hen. Le franc;ais moderne 63(1): 20-41. Hansen, Maj-Britt Mosegaard, 1996a. Eh #ien: Marker of comparison and contrast. In: E. EngbergPedersen, M. Fortescue, P. Harder, L. Falster Jakobsen and L. Heltoft, eds., Content, expression, and structure: Studies in Danish Functional Grammar, 315-342. Amsterdam: Benjamins. Hansen. Maj-Britt Mosegaard, 1996b. The function of discourse particles. A study with special reference to spoken standard French. Dissertation, University of Copenhagen. Heine, Bernd, UIrike Claudi and Frederike Hiinnemeyer, 1991. Grammaticalization. A conceptual framework. Chicago: Chicago University Press. Heltoft, Lars, 1987/1989. En plads til sprogvidenskabens hitteborn. Selskab for Nordisk Filologi. ,~rsberetning: 26-45. Heritage, John, 1984. Garfinkel and ethnomethodology. Cambridge: Polity Press. Heritage, John, 1989. Current developments in conversation analysis. In: D. Roger and P. Bull, eds., Conversation, 21-47. Clevedon: Multilingual Matters. Holm, John, 1988. Pidgins and creoles, vol. 1. Cambridge: Cambridge University Press. Hopper, Paul, 1987. Emergent grammar. BLS 13: 139-157. Hopper, Paul J. and Elizabeth Closs Traugott, 1993. Grammaticalization. Cambridge: Cambridge University Press. H61ker, Klaus, 1988. Zur Analyse von Markern. Stuttgart: Franz Steiner. Jayez, Jacques, 1988. Alors: Description et param~tres. Cahiers de linguistique franqaise 9: 133-175. Kroon, Caroline, 1989. Causal connectors in Latin: The discourse function of ham, enim, igitur and ergo. Cahiers de I'lnstitut de l inguistique de Louvain 15( 1-4): 231-243. Lakoff, George, 1987. Women, fire, and dangerous things. Chicago: University of Chicago Press. Lakoff, George and Claudia Brugman, 1986. Argument forms in lexical semantics. BLS 12: 442-454. Lamiroy, B6atrice and Pierre Swiggers, 1991. The status of imperatives as discourse signals. In: Suzanne Fleischman and Linda R. Waugh, eds., Discourse pragmatics and the verb: The evidence from Romance, 120-146. London: Routledge. Lehmann, Christian, 1985. Grammaticalization: Synchronic variation and diachronic change. Lingua e stile 20(3): 303-318. Levinson, Stephen C., 1979. Pragmatics and social deixis: Reclaiming the notion of conventional implicature. BLS 5: 206-223. Levinson, Stephen C., 1983. Pragmatics. Cambridge: Cambridge University Press. Lichtenberk, Franti,~ek, 1991. Semantic change and heterosemy in grammaticalization. Language 67(3): 475-509. Morel, Mary-Annick, 1985. Etude de quelques r6alisations de la lonction m6tadiscursive dans un corpus d'6changes oraux. DRLAV 32:93-116.
M.-B. Mosegaard Hansen / Journal of Pragmatics 28 (1997) 153-187
187
NC~lke, Henning, 1994. Linguistique modulaire: De la forme au sens. Aarhus: Aarhus Business School. Polanyi, Livia and Remko Scha, 1983. The syntax of discourse. Text 3(1): 261-270. Raynaud de Lage, Guy, 1990. Introduction h l'ancien franqais, ed. Genevibve Hasenohr. Paris: SEDES. Redeker, Gisela, 1991. Linguistic markers of discourse structure. Linguistics 29:1139-1172. Roulet, Eddy et al., 1987. L'articulation ,:lu discours en franqais contemporain. 2nd ed. Bern: Peter Lang. Schift¥in, Deborah, 1987. Discourse markers. Cambridge: Cambridge University Press Schiffrin, Deborah, 1994. Approaches to, discourse. Oxford: Blackwell. Sperber, Dan and Deirdre Wilson, 1986. Relevance. Communication and cognition. Oxford: Blackwell. Sperber, Dan and Deirdre Wilson, 1987. Pr6cis of Relevance: Communication and cognition. Behavioral and Brain Sciences 10: 697-754. Stempel, Wolf-Dieter, 1964. Untersuchungen zur Satzverkniipfung im Altfranz6sischen. Braunschweig: Georg Westermann. Sweetser, Eve, 1990. From etymology to pragmatics. Cambridge: Cambridge University Press. Tannen, Deborah, 1989. Talking voices. Cambridge: Cambridge University Press. Tobler, Alfred and Ernst Lommatsch, 1936. Altfranz6sisches W6rterbuch. Berlin: Weidemannsche Buchhandlung. Togeby, Knud, 1974. Prdcis historique c~e grammaire franqaise. Copenhagen: Akademisk Forlag. Traugott, Elizabeth Closs, 1982. From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In: W.P. Lehmann and Y. Malkiel, eds., Perspectives on historical linguistics, 245-271. Amsterdam: Benjamins. Trdsor de la langue franqaise. Dictionnaire de langue du XIXe et du XXe si~cle. 1973. Paris: Gallimard. Von Wartburg, Walter, 1934. FranzOsisches etymologisches WOrterbuch, vol. 3. Leipzig: Teubner. Weinrich, Harald, 1989. Grammaire textuelle du franqais. Transl. G. Dalgalian and D. Malbert. Paris: Didier/Hatier. Z~none, Anna, 1981. Marqueurs de cors6cution: Le cas de done. Cahiers de linguistique fran~aise 2: 113-139. Z6none, Anna, 1982. La cons6cution sagas contradiction: done, pal consequent, alors, ainsi, attssi (premitre partie). Cahiers de linguistique franqaise 4: 107-141.