Available online at www.sciencedirect.com
Journal of Pragmatics 44 (2012) 1764--1782 www.elsevier.com/locate/pragma
So as a multifunctional discourse marker in native and learner speech Lieven Buysse * Hogeschool-Universiteit Brussel, Belgium Received 2 February 2012; received in revised form 15 July 2012; accepted 22 August 2012
Abstract This article gauges the extent to which so is used as a discourse marker by Belgian native speakers of Dutch who have almost reached the end of formal instruction in English. The interview corpus compiled for this study is further diversified in order to determine the potential influence of distinct learning objectives in foreign language acquisition, with half of the learner participants majoring in English Linguistics, and the other half in Commercial Sciences. Not only is the use of discourse markers in these two sub-corpora juxtaposed from a quantitative and a qualitative perspective, the learner corpus is also set off against a comparable native speaker corpus. The investigation shows that the language learners use so significantly more often than their English peers, and the students of English Linguistics use so slightly more often than those of Commercial Sciences. All ten discourse marker functions of so, which can be situated in three different domains (ideational, interpersonal and textual), are found both in the learner and the native sub-corpora. An initial tentative account of the interrelatedness of these functions points in the direction of polysemy. © 2012 Elsevier B.V. All rights reserved. Keywords: Discourse markers; English as a foreign language; So
1. Introduction Discourse markers1 are optional linguistic items that fulfil an indexical function, in that they connect an utterance to its co-text and/or the context. Since the late 1980s research on discourse markers has boomed, but only in recent years has the scope been broadened to include non-native speaker discourse as well (see e.g. Romero Trillo, 2002; Fuller, 2003; Aijmer, 2004, 2011; Müller, 2005; Fung and Carter, 2007; Hellermann and Vergun, 2007; Buysse, 2009, 2011; Mukherjee, 2009; Polat, 2011; Wei, 2011). The relatively late onset of this complementary focus can be related to two parallel evolutions. First, following Schiffrin's (1987) groundbreaking monograph much attention was paid to the delimitation of the domain of discourse markers, including key definitional and taxonomic issues (see e.g. Blakemore, 1988; Redeker, 1991; Fraser, 1999; Schourup, 1999; Fischer, 2006). The scope of research can only be extended to a wider range of language varieties and corpora if the theoretical framework has been solidified. Although the theoretical debates are on-going, the discipline seems to have sufficiently matured to investigate discourse markers in languages other than English (see e.g. Hansen, 1998 on French; Wouk, 1999 on Indonesian; Cuenca and Marín, 2009 on Spanish), and to other language varieties, such as learner language (for an interesting overview of this evolution see Aijmer and Simon-Vandenbergen, 2009). Second, although the first learner corpora date back to the 1980s, it took until around the turn of the century for them * Correspondence address: Faculty of Languages & Literature, Hogeschool-Universiteit Brussel, Warmoesberg 26, 1000 Brussel, Belgium. Tel.: +32 485 45 10 29; fax: +32 2 210 13 52. E-mail address:
[email protected]. 1 The term discourse marker is preferred over a plethora of competing terms (see Fraser, 1999), first and foremost because it most aptly conveys what a linguistic item like so does (viz. signal discourse relations), and secondly because it probably has the widest currency in the field. 0378-2166/$ -- see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.pragma.2012.08.012
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
1765
to come into prominence (Granger, 2009), and spoken learner corpora, as vital resources for studying discourse markers in learner language, are more laborious and time-consuming to collect than written corpora. In concentrating on so in learner and native English, this article aims to contribute to two areas which have been little researched. As its first aim, it seeks to lay bare the discourse marker functions so fulfils in a particular speech context. Although so is highly frequent in English discourse (Biber et al., 1999:887; O’Keeffe et al., 2007:35), research on its use as a discourse marker has been ‘‘surprisingly scarce’’ (Bolden, 2006:663), particularly that which relies on corpus data and reaches beyond reflections on the traditional distinction between its status as a marker of ‘result’ and of ‘inference’ (see inter alia Van Dijk, 1979; Blakemore, 1988). Yet, a marker as frequent as so is likely to perform a variety of functions, which raises questions as to the nature of the relation between these functions. Broadly speaking, three main approaches have been put forward to account for the multifunctionality of discourse markers (Hansen, 1998): homonymy, i.e. the different functions are unrelated except for their form; monosemy, i.e. all functions are merely context-induced surface realisations of one single meaning; and polysemy, i.e. the functions are all related as extensions from a prototype. Obviously, which of these accounts applies to so ought to be the object of a diachronic study of the marker. The present investigation can, however, tentatively suggest which approach is most likely to apply to so, without the ambition of formulating a full-fledged proposal for one of these approaches at this point. As its second aim, this investigation sets out to expand the widening scope of discourse marker research in learner language to the English speech of native speakers of Dutch, which has so far remained untouched. Müller's (2005) analysis of so in the English of native speakers of German has borne out that these learners use so quite extensively yet with a considerably lower frequency than in native speaker discourse. The present article takes two distinct types of learners into its scope, viz. students of English Linguistics and students of Commercial Sciences. Both types have nearly reached the end of formal language instruction in English, but they evidently do not share the same aims for their proper language learning tracks. It can be hypothesised that the former use so more often than the latter, but, following Müller (2005), that so still has a lower incidence in both learner corpora than in a comparable native speaker corpus. It goes without saying that the findings of any corpus study have to be read in its context; hence, the results of this investigation cannot be generalised to all contexts in which so could be used nor to any kind of native speaker or learner of English. It does hope to contribute some insights that add a piece to the complex puzzle that the functional spectrum of so constitutes. After a brief overview of relevant prior research on so in section 2 and an outline of the methodology employed to conduct this study in section 3, section 4 sketches the discourse marker functions of so that have been attested in the corpus, and present the quantitative data. Section 5 addresses the multifunctionality of so, and seeks to elucidate the observed differences between three groups of participants. Conclusions are drawn in section 6. 2. Previous analyses of so In models of discourse and coherence so often serves as a prototypical example of either ‘result’ (e.g. Schiffrin, 1987) or ‘inference’ (e.g. Fraser, 1999). Many accounts of so take the distinction between its ‘result’ and ‘inference’ meaning as their starting point (e.g. Halliday and Hasan, 1976; Blakemore, 1988; Sanders et al., 1992). Van Dijk (1979:453--454) points at the underlying basic distinction between semantic and pragmatic (uses of) connectives: the former express relations between denoted facts (so indicates a result between two states of affairs), whereas the latter express relations between speech acts (so introduces the speech act of ‘drawing a conclusion’), as these examples demonstrate: (1)
I was sick, so I stayed in bed. John is sick. So, let's start. (Van Dijk, 1979:453)
In situating so on the five planes of talk of her model of discourse coherence, Schiffrin (1987) starts from this dichotomy too: on the ideational plane so links propositions, and in the action structure it links speech acts. So is, however, also present on the other three planes, yielding a more complex picture of so: it links turns, e.g. in adjacency pairs (exchange structure), it expresses a link between speaker and hearer (participation structure), and it contributes to the organisation and management of (meta)knowledge, e.g. because so introduces new information that in its turn serves as background knowledge for what follows (information state). In general terms Schiffrin (1987:191) describes so as a ‘‘marker of result’’ and of a transition to a ‘‘main idea unit’’, i.e. a proposition which expresses an idea that is considered of primary importance in the discourse structure. Most corpus-based studies of so have concentrated on specific contexts: turn-initial tokens of so in task-oriented spoken dialogue (Byron and Heeman, 1997), so-prefaced questions in police interviews with child witnesses and adult defendants (Johnson, 2002), so as a marker of professional procedures in courtroom testimonies of expert witnesses (Stygall, 2001), its various discourse-organisational uses in computer science seminars (Rendle-Short, 2003), and so
1766
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
marking ‘‘emergence from incipiency’’ to indicate that what follows does not emerge from the prior segment but has been on the conversational agenda for some time (Bolden, 2009). This has yielded a fairly scattered picture of the functional potential that the evidently versatile discourse marker so harbours. The findings of these studies should obviously all be considered within the constraints imposed on the data by the text type. Lam's (2009) study of so across text types shows that it is a highly frequent marker across the board, but that it is slightly more common in monologic, impromptu settings such as lectures than in dialogic settings such as meetings and planned types of speech such as press briefings and fully scripted speeches. Nonetheless, all functions of so can be used in all text types, but the relative frequency of each function may differ (e.g. turn managing functions are extremely rare in predominantly monologic settings such as lectures). It has also been suggested that discourse markers like so and now are typical of classroom discourse (as they are induced by teacher talk) as opposed to markers used to build rapport like you know and I mean (Hellermann and Vergun, 2007). This would make the former a more likely candidate for widespread acquisition by foreign language learners than the latter. What about so in non-native speaker discourse then? Recent years have witnessed a fanned research interest in learner practice in the use of discourse markers. In spite of evidence that so is highly frequent in learner English (e.g. Romero Trillo, 2002; Pulcini and Furiassi, 2004; Hellermann and Vergun, 2007), so has rarely received systematic treatment in investigations of discourse markers among foreign language learners of English. To the best of my knowledge, Müller (2005) is the only researcher who has attempted to comprehensively map the discourse marker uses of so in non-native discourse.2 From her corpus of monologic and dialogic story retellings in English by native speakers of German and of English she distils the following functional classification of so. On a textual level so can mark a result or consequence, a main idea unit (e.g. after a digression), a summary, rewording or example, a new sequence in a series of events (e.g. in a narrative), and the boundary that starts the conversation. On an interactive level so can indicate a speech act of question or request, a speech act of opinion, an implied result (when so occurs at the end of an intonation unit leaving the hearer to infer an implied result), and a potential turn end (2005:68--87). All ten functions are found in the native and non-native parts of the corpus alike, without a qualitative difference in use. The native speaker participants, however, turn to so about twice as often as their German peers. In sum, the discourse marker so appears in the existing literature as a ‘resultative’ or ‘inferential’ marker that has the ability to fulfil a multitude of (predominantly) discourse-structuring functions. As such it is frequent in native and learner language, but it is still more common in the former than in the latter. 3. Data and methodology The learner corpus analysed in the current study consists of informal interviews with 40 Belgian native speakers of Dutch in their second or third year in higher education (aged 19--26). At the time of corpus compilation half of the interviewees were undergraduates majoring in Commercial Sciences (henceforth CS), and the other half were majoring in English Linguistics (henceforth EL). Both groups were in the final stage of formal language instruction in English, but the latter can be considered to have been more intensively exposed to the target language and to have had a keener interest in the language than the former. Obviously the learning/teaching objectives set for the respective programmes are different for both groups, with the CS group aiming for a more functional command of the language and the EL group for a close approximation of native speaker practice. The interviews were conducted by members of the teaching staff at the students’ institutions according to the format of the Louvain International Database of Spoken English Interlanguage (LINDSEI; Gilquin et al., 2010). Each interview lasted about 15 min (average word count of interviewee speech: 1587) and followed a set pattern: the interviewee first talked for 1 or 2 min about a topic such as a film, book, travel experience or a life-changing experience, which sparked a conversation with the interviewer; at the end of each interview the students were asked to tell a short story based on four pictures. The CS corpus is almost balanced for gender (11 female and 9 male interviewees), whereas over two thirds of the EL corpus (14 out of 20) are made up of female speech. These ratios largely reflect the student populations of the respective programmes, and the voluntary nature of participation in the study. A third component was added to the corpus in the form of a sample of 20 interviews from the Louvain Corpus of Native English Conversation (LOCNEC), LINDSEI's native speaker reference corpus, which was compiled in the same format as the learner corpus. All native speakers of English were British students majoring in the Humanities at the University of Lancaster; the gender ratio matches that of EL. In total the corpus comprises almost 15 h of recordings, amounting to 115,530 words of interviewee speech: 30,134 in CS; 33,360 in EL; and 52,036 in NS. For comparative purposes, all quantitative data have been normalised to a
2 Lam (2009) also maps functions of so in a corpus that incorporates non-native discourse, viz. the Hong Kong Corpus of Spoken English. However, the aim of this corpus was to investigate the English spoken in Hong Kong, and hence comprises discourse of speakers of English as a native, second and foreign language. These are not treated separately in Lam's (2009) investigation.
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
1767
relative frequency expressed in number of tokens per 1000 words. Observed differences in frequency between the sub-corpora have been subjected to a log-likelihood test (henceforth LL) to gauge their statistical significance (Rayson and Garside, 2000). After the corpus was compiled, it was analysed with a moderately corpus-driven approach, taking its cues from the corpus yet without refuting external frameworks of analysis (e.g. notions from discourse analysis). After an initial analysis, so appeared to be the most frequent discourse marker in all three sub-corpora (discounting and and but), and together with well it was the only marker that occurred with a sufficiently high frequency to merit in-depth analysis (Buysse, 2010). In order to distinguish so's discourse marker functions from its other uses I have relied on Schourup's (1999) account of discourse markers in the form of their key characteristics: a discourse marker fulfils a connective function, does not contribute a propositional meaning to the utterance, and is semantically and syntactically optional. The non-discourse marker functions fall beyond the scope of the present article, and are therefore only briefly mentioned here for completeness’ sake (for a more elaborate account see Buysse, 2009): so as an adverb of degree or manner, so marking ‘purpose’, as a part of fixed phrases (e.g. and so on) and as a pro-form (e.g. I think so). Subsequently, the discourse marker tokens of so in the present corpus were grouped in nine categories. A mere two per cent of tokens of so (viz. 36 out of 1583) proved unclassifiable because a part of the utterance was unintelligible. 4. Results 4.1. Discourse marker categories attested in the corpus The discourse marker functions of so attested in the corpus can be classified according to the types of discourse relations they index, as inspired by Halliday and Matthiessen's (2004) metafunctions. Ideational relations are those linking two states of affairs in the world described in the discourse, such as a result (Halliday and Matthiessen, 2004:29; Degand, 2001:79; Redeker, 1990:369). Interpersonal relations link ‘‘the illocutionary meaning of one of the discourse units with the locutionary meaning of the other’’ (Degand, 2001:79), such as a conclusion. In a similar vein, Redeker (1990) -- calling them ‘‘rhetorical relations’’ -- describes these relations as linking ‘‘the utterances themselves or, to be more precise, (. . .) the beliefs and intentions motivating them’’ (1990:369). Contrary to ideational types of relations, interpersonal relations can be both interactive and personal and therefore involve a personal involvement from the speaker (Halliday and Matthiessen, 2004:30). This is underscored by Traugott (2010:34), who organises expressions on a cline from those that are non-subjective (or less subjective) to those that are subjective and/or intersubjective. She aligns non-subjective expressions to the ideational metafunction and (inter)subjective expressions to the interpersonal metafunction. Textual relations, finally, are discourse-organising relations between ideationally or interpersonally loosely or indirectly related elements, such as a list or digression (Degand, 2001:79; Halliday and Matthiessen, 2004:30). Table 1 provides an overview of all discourse marker functions identified for so in the corpus at hand, presented according to the type of discourse relation they mark. Although more than one relational type may operate in one and the same token, one of these is always more salient than the other (see also Redeker, 2006:354). For example, just about all discourse marker tokens of so can be interpreted as discourse-organising and hence textual (e.g. in indicating a ‘result’ relation (ideational) so also functions as a discourse-organising device (textual)). If an ideational or interpersonal relation can be discerned, this is considered more salient in our categorisation. In what follows each function is explained in greater detail, illustrated with examples from the corpus and related to similar, if not identical, functions identified in previous research. Table 1 Discourse marker functions of so. Type of relation
Discourse marker function
Ideational
Indicate a result
Interpersonal
Draw a conclusion Prompt Hold the floor
Textual
Introduce a summary Introduce a section of the discourse Indicate a shift back to a higher unit of the discourse Introduce a new sequence Introduce elaboration Mark self-correction
1768
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
4.1.1. Ideational: Indicate a result A ‘resultative’ relation holds between discourse units that denote states of affairs in the world described in the discourse (and is hence ideational), the first of which causes the second. In a sentence X so Y, it can thus be paraphrased as: ‘‘state of affairs Y is the result/consequence of state of affairs X’’. For example, in (2) the fact that some actors were playing several parts is a result of a shortage of actors, and the inability of the audience to figure out what was going on in the play is a result or consequence of the fact that some actors were playing several parts. (2)
yeah . and they had a shortage of actors . so they had some actors playing several parts so you just couldn’t figure out what was going on <\Ie> (NS09; 01:30 & 01:32)3
In the semantics--pragmatics dichotomy the ‘result’ relation that so expresses is taken to belong to the semantic domain. In some investigations this function of so has been denied discourse marker status (e.g. Stygall, 2001; Pulcini and Furiassi, 2004), although such tokens of so adhere to the three criteria of non-truth-conditionality, optionality and connectivity, as posited by Schourup (1999). I will follow Müller (2005:72) and Lam (2009:362) in positing that so indicating a ‘result’ or ‘consequence’ should be awarded discourse marker status because it is both syntactically and semantically optional in this context. Almost all participants cast so in its ‘resultative’ role at least once (19 each in EL and CS and 20 in NS). 4.1.2. Interpersonal: Draw a conclusion In contrast with a ‘result’ relation, a ‘conclusion’ relation does not pertain between states of affairs in the world the discourse describes, but rather between two propositions, the first of which serves as the ground for the speaker to posit a claim in the second. Moreover, as this category belongs to the interpersonal domain, it can incorporate personal involvement from the speaker in the message conveyed (Traugott, 2010). Aaron (2004) remarks that ‘‘[w]hile many authors collapse [these functions], they are fundamentally different’’ (2004:169), suggesting that in analyses of corpus data ‘resultative’ and ‘conclusive’ instances of so are rarely discriminated between (see e.g. Müller, 2005; Lam, 2009, 2010) although there are solid grounds to do so. Most significantly, there are two distinct types of discourse relations that underlie them, the ‘resultative’ belonging to the ideational domain and the ‘conclusive’ to the interpersonal. A possible paraphrase for a conclusive structure is: ‘‘from state of affairs X I conclude the following: Y’’. Take excerpt (3), in which the interviewee has just responded negatively to the interviewer's question whether she would ever have her portrait painted. When the interviewer enquires whether her answer would be different if someone asked her to paint her, she argues that this might indeed change matters: if someone asks to paint you, they must be convinced that the painting would turn out nicely. The so-prefaced segment offers an inference drawn by the interviewee from the prior segment, on the basis of the presupposition that a painter will only approach beautiful women (or women that would look good in a painting). The epistemic use of must in the so-prefaced segment underscores the speaker's involvement in the message. (3)
uhu . would you ever have your portrait painted <\Ir> . no <\Ie> why not <\Ir> erm . . . I would feel like er okay look at me look at me how beautiful I am because I’m em . no . . . em <\Ie> suppose someone asked you . whether . he or she could draw your painting <\Ir> maybe because . in that way he . he or she asked me so he or she must be convinced . that it would be beautiful <\Ie> (CS13; 12:13)
Excerpt (4) was taken from the picture-based storytelling section of the interview. In the story a woman is displeased with the portrait an artist has painted for her, after which he redoes the portrait and the woman shows it to her friends. The interviewee here infers from the observation that the woman is pleased with the second version of the portrait and that she is now eager to show it to her friends that the second version does not resemble her. The evidential clause it's obvious that points at the inferential process underlying the following statement. 3 Each excerpt from the corpus is identified with the interview code (which starts with the abbreviation of the sub-corpus to which the interview belongs) and the time reference at which the relevant token of so occurs in the interview (in minutes and seconds). The transcription conventions are provided in Appendix.
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
(4)
1769
erm . . and then she s= she seems pleased with that picture . and she's showing it off to all her friends so it's obvious that the picture looks quite different from her <\Ie> (NS01; 13:45)
Clearly, in this example the fact that ‘‘the picture looks quite different from her’’ was not caused by the fact that ‘‘she seems pleased with that picture and she's showing it off to all her friends’’ (quite the contrary), and neither is the fact that ‘‘s/he would be convinced that it would be beautiful’’ by the fact that ‘‘s/he asked me’’ in example (3). This runs counter to the analysis of the ‘resultative’ excerpt (2), in which the fact that ‘‘several actors were playing several parts’’ was indeed caused by the fact that ‘‘they had a shortage of actors’’. ‘Conclusive’ so, therefore, rather links a claim to the argument that justifies its positing, and therefore requires a reasoning process. A ‘resultative’ relation does underlie the ‘conclusive’ relation between a claim and its argument, but it does so on a more abstract, cognitive level (viz. the positing of the claim is the result of the argument) than its ‘resultative’ counterpart. All participants, across the three sub-corpora, used ‘conclusive’ so at least once. 4.1.3. Interpersonal: Prompt Some tokens of so at first glance appear to fulfil little or no function other than to indicate the desire to yield the floor. The discourse marker does not introduce a segment then, as the segment stops after so, and carries a falling or level tone with a short vowel.4 Typically in this category so is followed by an empty pause or, if the interlocutor claims the floor right away, by a turn transition. It can be preceded by an empty pause, which in combination with a following empty pause creates a separate intonation unit for so; if it is not, so is incorporated in the intonation unit of the preceding segment, which it closes. Hence, tokens of so that have a ‘prompting’ function do not normally start an intonation unit. Müller (2005), who observes the same intonation pattern for so in a similar function (‘‘Marking implied result’’), contends that this pattern ‘‘gives an additional cue for the hearer that s/he may take over the floor’’ (2005:85). Schiffrin (1987), Aaron (2004), Müller (2005) and Lam (2009, 2010)5 assert that in such non-prefatory instances so incorporates an implicit proposition that stands in a resultative or inferential relation with the prior co-text. The discourse marker serves as a prompt for the hearer to recover this implied message, and because of this appeal to the hearer it simultaneously constitutes a transition-relevance place (i.e. a point in the interaction at which the turn may but need not shift to another interactant; cf. Sacks et al., 1974). In excerpt (5) the inferred conclusion is: ‘‘so no I haven’t seen any films recently’’. So marks this inference as self-explanatory given the argument presented in the prior segment, and hence as readily retrievable for the interlocutor, making its explicit utterance superfluous. Because of its indexicality to the hearer that a turn-shift is expected this function of so belongs to the interpersonal domain. (5)
uhu uhu . em any films that you’ve seen recently <\Ir> . erm well I hate films so <\Ie> you hate [films <\Ir> [yeah I hate films <\Ie> (CS16; 06:32)
Because the implied meaning is not uttered, it cannot always readily be brought to the surface, to the extent that in some instances the only message may be that the speaker has nothing else to say on the matter. In excerpt (6), for example, the interviewee has reached a point of termination in the explanation of a project he worked on, and indicates a desire to relinquish the floor with the prompt so. A medium-length empty pause is only met with the interviewer's backchannelling, which makes the interviewee signal turn-termination with a token of so that is uttered with level intonation. This is followed by a short empty pause before the interviewer eventually claims the turn with a new question. In such cases so rather seems to prompt a turn transition than the reconstruction of an implied resultative proposition. (6)
and you should never do that and stuff like that it wasn’t like that at all it was more of a it was well it was a definite story .. erm .. and it started off with this guy in his home and then he went out to a night-club with his friends and then he got .. followed by some .. scary guy and he had his bleeper with him so . you know <\Ie> uhu <\Ir>
4 In line with the transcription conventions of the LINDSEI corpus (see Appendix), the interviews that form the corpus for the present study have not been transcribed prosodically, but the original recordings were consulted to disambiguate the function of non-prefatory tokens of so. 5 Aaron (2004:170) chooses the opaque term ‘‘cut-off so’’ to describe this function, whereas Schiffrin (1987:223) and Müller (2005:84) dub it ‘‘marking implied result’’ and Lam (2010:670) considers it a ‘‘turn managing’’ device (and discusses it in the same category as turn-initial instances of so).
1770
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
always carry your bleeper with you <\Ie> mhm <\Ir> and stuff like to scare him away .. erm yeah just to make him as normal as possible as I said you wouldn’t know he was mentally handicapped .. er except the title . <\Ie> mhm <\Ir> and things like that <\Ie> mhm <\Ir> . so <\Ie> . and I guess there were quite a few pictures in the book as well <\Ir> (NS18; 07:11) All language learners and all but three native speakers turned to this prompting non-prefatory use of so at least once. 4.1.4. Interpersonal: Hold the floor In the interpersonal domain non-prefatory tokens of so may also indicate the speaker's desire to hold the floor. Contrary to those with a prompting function, floor-holding tokens of so are usually pronounced with a rising tone, preceding and/or following pauses (which may be filled or empty), and a prolonged vowel.6 Lam (2009) hints that in this ‘‘processing’’ function so can be used ‘‘as a delay strategy and signal that the speaker is undergoing some processing problem and requires extra time’’ (2009:364). Example (7) illustrates how so can be used to this end. The interviewee has been talking about her commitment to a youth movement called Chiro. She indicates that she is no longer an active leader in the organisation, but former leaders still on occasion participate in some of its activities. She explains what this involves (‘‘we meet once in a while and we help the Chiro’’), and then the interviewee takes a brief moment to reflect, filled by so, after which she gives an example of the assistance they offer. Here so's prosodic features signal that it does not introduce the following segment, but rather follows the previous without an immediate intention of adding a segment. Indeed, there is only a brief pause between the prior segment and the marker, which is filled by the interviewer's backchannelling, but a longer pause between so and the following segment. The rising intonation indicates an intention to continue the turn, while the lengthened vowel in so stresses hesitation or reflection. Moreover, the following segment starts at a high pitch, which marks the onset of a new intonation unit. (7)
it was very important but now I’m er I quitted <\Ie> why <\Ir> because . it is . it was too much and now I really want to go for school so: <\Ie> uhu <\Ir> you have to make p= priorities so: <\Ie> uhu <\Ir> for me it was enough but we have an old leaders working now and . yeah . we: we meet once in a while and we help the: . the Chiro <\Ie> uhu <\Ir> so: .. in two weeks it's a party and . fo= while the children make a little dance or . <\Ie> uhu <\Ir> at the stage and . yeah we: . we help to organise this for . the: Chiro team <\Ie> (CS06; 07:10)
In (8) too so is prosodically set off from both the preceding and the following segment with a pause, it carries rising intonation and its vowel is lengthened. The interviewee responds to the interviewer's advice to start thinking about finding a job that she is actually in the middle of an application process for a job at Boots. The turn might have ended after the segment ‘‘I’ve been doing it this afternoon’’, but the interviewee holds on to the floor with so, and then starts a new segment with the contrastive marker but. It would seem that she has realised that the interviewer might get a false picture of the kind of job she is aiming for at Boots, and therefore adds that information. So is not the appropriate marker to relate this segment to the previous, creating the need for the insertion of an alternative marker (but). 6 Research into the prosody of discourse markers remains scarce (but see e.g. Wennerstrom, 2001:96ff.). Yet, Local and Kelly (1986:195--196) confirm that a combination of a discourse marker like so or well and surrounding silence is typical of seeking closure if the marker is pronounced with level or falling tone, and of seeking continuance if pronounced with level or rising tone.
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
1771
you: you should start thinking about it [I mean you’re in your last last year <\Ir> [yes . . well I am thinking about it I’m applying to: Boots at the moment I’m doing the[i:] application form I’ve been doing it this afternoon . so: . . but management type of jobs and <\Ie> management <\Ir> yeah <\Ie> (NS12; 06:26)
(8)
In CS and NS ‘Hold the floor’ surfaces in 12 interviews each, and in EL in 11 interviews. 4.1.5. Textual: Introduce a summary As Redeker (1990:373) and Müller (2005:76--78) point out, so can introduce a segment that sums up (a part of) the prior discourse. The main difference with ‘conclusive’ so is that in introducing a summary the speaker does not posit a new claim based on the prior co-text or the shared background knowledge of the interactants, but rather restates the main argument(s) that can be distilled from the prior discourse in more general terms, allowing the speaker to round off a stretch of speech. For example, the interviewee in (9) explains what the film The Big Lebowski is about, and at various points in this exposé (underlined in the fragment) he points out the aspects he particularly likes about it. At the end of this turn he brings the most important of these features together in a proposition that states what he likes about such films. On the one hand, he summarisingly reiterates the core elements of his preceding explanation, and on the other hand, he generalises his claims about this particular film to other instances in the same genre. (. . .)7 it's a very funny movie . my . it's my favourite movie because . it doesn’t have a storyline at all . in the[i:] end it's like nothing really happened somebody died . of course there were some moments in which in which you thought now . he's learning something or: something's changing in his life (. . .) erm what I also like about the movie is that there .. is a severe sort= shortage of good communication <\Ie> uhu <\Ir> there isn’t any communication at all . people try to talk to each other but each of them has a . specific .. form of . communication failure (. . .) so that's what I really like about . movies like that it's . kinda like . Pulp Fiction and having not a real story <\Ie> uhu <\Ir> and kinda like . . . yeah well just . a messed up .. messed up . yeah language use and <\Ie> mm <\Ir> other . behaviour of the . main personalities <\Ie> (CS11; 02:34)
(9)
Excerpt (10) offers a so-prefaced summary of the story the interviewee has just recounted on the basis of the four pictures put to her by the interviewer. Here the summative function of so is further strengthened by the adverb basically. (10)
(. . .) because she doesn’t like the way she looks . so he makes up . a totally false impression of her . . and erm . . then she shows it off to all her mates and she's really made up with it . . so it's basically about . he draws a good picture of her mhm <\Ir> a realistic picture of her she can’t face the truth of it . and so he has to start all over again and she's got like this false image of herself <\Ie> (NS09; 15:23)
So introduces a summary in all NS, 19 CS and 15 EL interviews. 4.1.6. Textual: Introduce a section of the discourse So has the ability of opening the interviewee's first turn in (a section of) the conversation. At the beginning of each interview the interviewee is expected to talk briefly about a topic (see section 3). Before the recording device was started, 7 Some passages have been cut out from the excerpt for reasons of space, viz. the introduction as to which film the interviewee wishes to discuss and the two long lists of examples that buttress the two underlined claims.
1772
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
all participants received a sheet with a brief explanation of the task. After they had read it, the recording was started, so the topic they wanted to address was not discussed in advance. In example (11) all that precedes the interviewees’ first turn is a short prompt by the interviewer to start talking. A similar situation occurred near the end of the conversation: each interview contains a point where the conversation is interrupted to allow the participants to briefly study four pictures and subsequently tell a story on the basis of these pictures, as in (12). (11)
okay thank you go ahead <\Ir> so er . I have family in England and I’ve visit them a few times erm and I went last year for two weeks [. . .] <\Ie> (CS05; 00:06)
(12)
erm so there's this man painting a woman first he he makes a portrait of her which is very realistic [. . .] <\Ie> (EL14; 12:56)
Müller (2005:80) observes a similar function for so (as a ‘‘boundary marker’’8), in which it relates instructions to the beginning of the narrative in her experimental setting. About a quarter of native and non-native speakers in her study make use of it, which is more than in the present investigation, because the specificity of the environment in which so takes on this function heavily restricts the frequency of ‘discourse initial’ so. It only occurs 15 times in the corpus as a whole, 7 of which in CS, 6 in EL and 2 in NS (in 4, 5 and 2 interviews, respectively). The relatively high number in CS is largely due to the way in which the interviewers prompted the interviewees to start (a part of) the conversation: the CS interviewer almost invariably used an imperative as in (11), whereas the EL interviewers only did so on occasion, and the NS interviewer never. The presence of this prompt gave the interviewees an opportunity to ‘‘respond to’’ or ‘‘build on’’ what the interviewer said, which can be marked with the discourse marker so. 4.1.7. Textual: Indicate a shift back to a higher unit of the discourse A speaker may use so to signal that s/he shifts the conversation back to a higher textual level, either after a brief interruption by or an exchange with the interviewer or after a turn-internal digression. This function is illustrated in excerpt (13), in which the interviewer interrupts the interviewee to ask for clarification as to the location of the interviewee's and the assistant's flats. The interviewee provides the requested information and then resumes her anecdote with a segment that starts with so and a repeated part of the last segment before the interruption. This repetition obviously facilitates utterance interpretation by the hearer. (13)
and then I had a lot of form filling to do I had to get an insurance for it erm but luckily we: sort of . protested quite a lot myself and the German assistant it it was good that there was two of us . erm and we managed to get a rent rebate [for students <\Ie> [it it was it was in the same building <\Ir> yeah <\Ie> [oh yes <\Ir> [it was like a building for young workers and students <\Ie> mhm <\Ir> erm . so we managed to get erm a rent rebate by claiming that even though we were paid we were still students <\Ie> mhm <\Ir> erm and that made it a lot that made about fifty pounds a month which was excellent it was very cheap then <\Ie> (NS07; 09:56)
Excerpt (14) is an example in which the interviewee is interrupted by the interviewer, who runs ahead of the former's second argument why he does not intend to visit Tanzania again any time soon. The interviewee briefly acknowledges this remark with ‘‘yeah sure’’, after which so, in combination with the last (unfinished) segment before the interruption, prefaces the second argument.
8 This term is not taken over here, because it can also refer to so's function as a marker of a return to a higher level of the discourse (Kyratzis and Ervin-Tripp, 1999:1324), discussed in section 4.1.7, which corresponds to Müller's (2005:74) category of ‘‘main idea unit marker’’.
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
(14)
1773
[yeah . . and do you . do you plan to go there again or <\Ir> I’d hope so . <\Ie> yeah <\Ir> I hope so . erm . but I don’t know when . <\Ie> yeah <\Ir> I mean . erm . I’m quite sure that it . it won’t be for in a very short time now because erm . my parents have been there and I don’t think they’ll go there again they’ve enjoyed it a lot but I’m not sure if they’ll go again . <\Ie> uhu <\Ir> and . [by the time <\Ie> [well it's expensive as well [I suppose <\Ir> [yeah sure so by the time I’ll have the money to pay it for myself . <\Ie> yeah <\Ir> I think I’ll have to work a lot it's [not for . <\Ie> (EL10; 13:05)
Polanyi and Scha (1983:265) consider so the prototypical example of a ‘‘resumption’’ marker, and Schiffrin (1987:191) as well as Müller (2005:74) calls so a marker of ‘‘main idea units’’. Similar instances of so have also been identified by Vandepitte (1993:145), who observes that so can indicate a return to the main story, and Redeker (2006:345), who reports that it can mark a return from a parenthetical segment. In CS 9 interviewees used so to this end, 10 in EL and 13 in NS. 4.1.8. Textual: Introduce a new sequence Redeker (1990) already pointed at so's potential to mark sequential relations ‘‘between successive elements in a chain of events’’ (1990:373--374), and Müller (2005:78) noticed a similar phenomenon in the narrative parts of her corpus. In the present corpus too so can indeed introduce a new sequence in a narrative or a new step in an explanation. Apart from the fact that it does not mark a clear ‘resultative’ relation, this category distinguishes itself from so's ‘resultative’ function in that ‘sequential’ so starts a new sequence within the turn, whereas ‘resultative’ so brings a sequence or a turn to a closure. In (15) the interviewee tells the story of the film The Big Lebowski. Jeffrey Lebowski is confused with a millionaire with the same name whose wife owes money all over town. ‘‘Some thugs’’ claim the money from the wrong guy in all but a subtle way, which leads to the next scene, in which he goes to see the millionaire to complain about the situation. This scene is prefaced by so in the interviewee's version of the story. Similarly, in (16) one of the scenes from the picture-based story is prefaced with so. (15)
[. . .] and so em . some thugs come over to his place smash the entire place up because they think he's . the millionaire and er that they have to: get the money out of him through the means of physical violence <\Ie> uhu <\Ir> however . it's not . it's not him so he starts complaining to the millionaire . about how the thugs have peed on his rug and the rug really tied the room together so9 he wants the rug from the millionaire . and he goes over there . and . through a matter of circum= er events . he starts . to become a private detective in finding the missing millionaire's wife [. . .] <\Ie> (CS11; 04:20)
(16)
. . erm she's not happy with it because he's painted her . as she is <\Ie> <\Ir> she obviously wants to look a bit more glamorous . . er so he repaints the picture and she hangs it up very impressed . . er to show all her friends . . who per= who perhaps aren’t quite as impressed as her <\Ie> (NS16; 15:09)
A total of 17 NS interviewees had so perform a ‘sequential’ function, compared with 11 in EL and a mere 8 in CS. 9 Notice that this token of so does not fulfil a sequential function. It ends the sequence that explains the underlying reason for Lebowski to visit the millionaire, and indicates a ‘result’: because the thugs have urinated on the rug that was an important decorative element in his house, he wants to be compensated by the millionaire.
1774
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
4.1.9. Textual: Introduce elaboration ‘Elaboration’ occurs when ‘‘one clause elaborates on the meaning of another by further specifying or describing it’’ (Halliday and Matthiessen, 2004:396). There are two ways in which so introduces such segments. First, it can preface the explanation of a prior general statement: the so-prefaced clause supplies a non-argumentative explanation, which is merely meant to supplement the information the speaker has just provided. In (17), for example, the so-clause clearly does not express the result of learning a language non-artificially but it explains what the interviewee means by this. (17)
. French is quite [usef= useful I think <\Ir> [yeah yeah yes useful <\Ie> if you want to go to Canada [to Quebec <\Ir> [yeah <\Ie> it might be useful <\Ir> but I’d like . just to not learn it artificially as well so I’d just be around [lots of French people and <\Ie> [be around and pick up the <\Ir> [sort of pick . yeah pick things up yeah <\Ie> [real . real French <\Ir> (NS02; 15:26)
Second, the so-prefaced segment is inserted parenthetically in a turn, as in (18). The main argument seems to be that although the interviewee had not initially planned on majoring in Italian, he chose the course because he loved the language so much. However, he adds parenthetically what his initial plan was and why he dismissed it in the end. This information is still a specification of what precedes -- as with the first type of elaborative so -- but here the specification is not vital to the argument: the focus is on why the interviewee chose Italian, not on what he did not choose. The information on Linguistics as a major course appears to be inserted merely for clarity's sake, so the interviewer does not have to wonder about what the student had initially planned on choosing. The general extender and what not at the end of the parenthetical segment reflects the secondary weight the segment carries in his decision-making process. After a mediumlength empty pause the interviewee again picks up the main thread of his turn. (18)
mm I wh= when I came here I hadn’t planned on majoring Italian <\Ie> mhm <\Ir> so I was I was gonna do linguistics but cos you’ve got to choose three subjects and what not .. I enjoyed Italian so much so I decided to carry on with it and you know .. it's not worth doing unless you can do it properly so I thought I’d do a year abroad and <\Ie> mhm <\Ir> see how it goes <\Ie> (NS13; 06:06)
Similar uses have been identified for so by Schiffrin (1987), Rendle-Short (2003), Müller (2005) and Lam (2009, 2010). The ‘elaborative’ function of so stands out, because it contradicts the prototypical depiction of so as a ‘‘marker of main idea units’’ (Schiffrin, 1987:191), or in other words a hypertactic device by which speakers signal relatively important parts of their speech (e.g. a result, a conclusion, a return to the main thread of a conversation, the start of a conversation). In its ‘elaborative’ guise so brings out a hypotactic relation between segments, or in Schiffrin's terms introduces not the ‘‘position’’ of an argument but rather its ‘‘support’’ (1987:223). The speaker clearly feels that a claim has been insufficiently substantiated prior to its utterance, and hence offers some justification. By inserting so as a preface to this support the speaker gives the impression that the justification follows from information that has been previously provided. This tallies with so's inferential meaning: a conclusion that is based on (co- or contextually) prior information. Thereby the justification seems a logical step in a typical reasoning, although it is not what it pretends to be, viz. a consequence or a conclusion (Buysse, 2009). This function is spread across 16 CS, 14 EL and 18 NS interviews. 4.1.10. Textual: Mark self-correction A clause introduced by so can correct (a part of) a prior utterance by reformulating it, i.e. it acts as a textual device for self-repair or ‘‘self-editing’’ (Lam, 2009:364). This may concern rectification of the contents of what precedes, or of a language-related problem, as in (19), where the interviewee rephrases a clumsy formulation, or in (20), where the interviewee is searching for the correct English equivalent of the Dutch phrase zwakke weggebruiker (‘vulnerable road user’), which she starts translating with weak, and then attempts to paraphrase.
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
1775
(19)
er I really lo= it was about erm . well this girl and a guy and em they they’re in a relationship and er well eventually they break up and the girl decides she wants to: have her brain washed so she wants to have a brainwash and em . well then . it's . really difficult to follow the movie but in the end it's really like oh yeah that's how it went [. . .] <\Ie> (CS14; 00:28)
(20)
uhu they don’t follow the rules <\Ir> no and er . the: . weak. er . so the when you are with a bicycle erm <\Ie> uhu <\Ir> and by feet . erm zwakke weggebruiker <\Ie> (CS17; 05:22)
‘Mark self-correction’ is limited to 6 interviews in both CS and EL and 4 in NS. 4.2. The discourse marker functions of so in numbers In all, the corpus under investigation contains 1583 tokens of so, four fifths of which fulfil a discourse marker function (1258 tokens). The non-discourse marker tokens are responsible for 18 per cent of cases, and the unclassified utterances for 2 per cent. As Fig. 1 shows, in NS 9.32 tokens of so perform a discourse marker function per 1000 words (n = 485), but in CS this incidence is 21 per cent higher, with 11.28 tokens per 1000 words (n = 340) and in EL even almost 40 per cent, with a relative frequency of 12.98 (n = 433). Hence, the language learners in the corpus use so considerably more often as a discourse marker than their native peers, and the students of English Linguistics do so to a greater extent than the students of Commercial Sciences. The difference between CS and NS achieves statistical significance at p < 0.01, and that between EL and NS is significant at p < 0.0001 (LL at 7.21 and 24.82, respectively). No statistical significance can be found for the CS--EL difference (LL at 3.76). All participants in the study, native and non-native speakers alike, use so at least 3.33 times per 1000 words. Fig. 2 breaks these numbers further down per discourse marker function for each sub-corpus. In all three sub-corpora the interpersonal domain is most commonly explored by so, followed by the textual and then the ideational, with overall relative frequencies of, respectively, 4.27, 3.78 and 2.84. It should be borne in mind, though, that the latter domain consists of a single function. The only statistically significant difference for this function holds between EL and CS (at p < 0.05; LL 6.41), the former using it on average about one-and-a-half times as often as the latter. In spite of its first place in the ranking of relational domains across the board, the interpersonal functions are on the whole still used significantly more often in EL and CS than in NS ( p < 0.001 and p < 0.0001, respectively; LL 12.32 and 16.36). The incidence of ‘conclusive’ so in NS is three-and-a-half times lower than that of EL ( p < 0.001; LL 12.44) and three times lower than that of CS ( p < 0.01; LL 7.67), whereas so occurs significantly more often as a ‘prompt’ in both CS and EL than in NS, by a factor of 1.75 ( p < 0.001; LL 12.53) and 1.41 ( p < 0.05; LL 4.78), respectively. No significant differences could be established for ‘Hold the floor’.
[(Fig._1)TD$IG]
14
n tokens per 1,000 words
12 10 8 6 4 2 0 CS
EL
NS
Fig. 1. Relative frequency of so as a discourse marker in each sub-corpus.
[(Fig._2)TD$IG] 1776
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
3.50
n tokens per 1,000 words
3.00 2.50 2.00 1.50
CS EL
1.00
NS
0.50 0.00
Fig. 2. Relative frequency of the discourse marker functions of so in each sub-corpus.
The textual domain is one-and-a-half times more frequent in EL than in NS, which achieves statistical significance at p < 0.001 (LL 12.87). Since the frequencies for ‘Introduce a summary’ and ‘Indicate a shift back to a higher unit of the discourse’ closely approximate each other for these sub-corpora, and the discrepancies for the discourse-initial function of so are irrelevant given the slight differences in context created by the interviewers (see section 4.1.6), the overall dissimilarity for this domain must be down to the other three functions. It would seem, however, that a major role is reserved for ‘elaborative’ so. A significant difference can be observed for the ‘self-corrective’ function between EL and NS ( p < 0.05; LL 6.01), but with a mere 21 tokens in the corpus as a whole this category is too small to carry any meaningful weight. Uncharacteristically, ‘sequential so’ surfaces twice as frequently in NS than in CS ( p < 0.05; LL 4.14), and only slightly more than in EL (not statistically significant). ‘Elaborative’ so, on the other hand, has been attested almost 2.3 times more in EL than in NS ( p < 0.001; LL 14.35) and 1.7 times more in CS than in NS ( p < 0.05; LL 5.31). 5. Discussion 5.1. In search of an account for the multifunctionality of so The qualitative analysis of so in section 4.1 has evinced ten functions of the discourse marker, which demonstrate the versatility of the discourse marker so. The main common and distinguishing features between these different functions can be derived from the degree to which the prototypical ‘result’ relation of so can be discerned in each function. Three aspects are key in this: the level at which the relation holds (a), the hierarchy between the segments that so joins (b), and whether the so-prefaced segment provides closure to a discourse sequence (c). First, the ‘result’ relation that so indexes may range from a concrete reason--result relation between two states of affairs in the real world (‘resultative so’), over a more abstract argument--position relation (‘conclusive so’ and ‘prompt’) to a textual, wholly procedural relation that signals the so-prefaced segment as a logical next step in the discourse (e.g. ‘Shift back to a main unit’). Second, as Schiffrin (1987:191) observes, a relation of ‘result’ between discourse segments incorporates a shift from a subordinate to a main unit, in which the cause in the utterance is taken to be the subordinate unit, and the result the main unit. This hierarchy springs from the syntactic position of the corresponding markers: the prototypical ‘causal’ marker is because, which is a subordinating conjunction and hence introduces a subclause, whereas the prototypical ‘resultative’ marker is so, which is either a coordinating conjunction or an adverb and hence always introduces a main clause. Moreover, a subclause is not meant to occur independently; likewise a cause cannot readily occur without its result or consequence. The reverse is possible, however: a result or a consequence can be mentioned, even if the speaker (or writer for that matter) does not go into detail about the event that caused this state of affairs. Therefore, if both cause and result are mentioned, the former can be taken to be subordinate to the latter, and ‘resultative so’ thus marks a shift from a subordinate to a main unit, which is a hypertactic transition. Likewise, in ‘conclusive so’ a claim is superordinate to its argument, and if so functions as a prompt, it
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
1777
prefaces an implicit result or conclusion. A summary can also be considered superordinate to the more specific segments that it summarises. The remaining functions, on the other hand, are rather ambivalent in the hierarchy of the relations so indexes. For example, ‘Shift back’ and ‘Elaboration’ are each other's opposites, the former leading out of a digression back to the main thread of the discourse (hypertactic) and the latter leading into a digression (hypotactic). Also starting a new part of the discourse, a new sequence or a corrected version of an infelicitous utterance is hypertactic, albeit that the distance between the subordinate and main units is less pronounced than with ‘Shift back to a higher unit of the discourse’. Thirdly, by its very nature a result constitutes a point of closure, which can pertain to various levels: in any event that of the discourse sequence made up of the causal and resultative segments, but also that of a stretch of sequences that together form an explanation or a story, of a turn, etc. As such ‘resultative’ so holds the (secondary) potential of a pivotal element: it states the most important element of what precedes, which allows the interlocutor(s) and the conversation to carry on, either to continue a topic based on the information that has just been provided, or to start off in a new direction of the same topic (e.g. to start a parallel part), or to launch an entirely new topic. This potential is even more outspoken in the ‘conclusive’, ‘prompting’ and ‘summarising’ functions, as these are more prone to end not only a distinct segment of a turn but also a turn altogether. The other functions, however, rarely constitute a point of closure: only ‘Self-correction’ could be considered as providing closure to an infelicitous part of the discourse, whereas the other functions either start a new section of the discourse (‘Initiate discourse’, ‘New sequence’) or continue one (‘Shift back’, ‘Elaboration’) or could do either (‘Hold floor’). The mere existence of these common features for the discourse marker functions of so seems to rule out both a monosemous and homonymous approach to explaining so's multifunctionality. The variation among the ten discourse functions of so attested in the corpus along the lines of these three features, rather points at a loss of semantic meaning as the spectrum moves from ideational to textual functions. This tendency could reflect a diachronic process of grammaticalisation or pragmaticalisation (cf. inter alia Schourup, 1999; Andersen, 2001; Waltereit and Detges, 2007), whereby so has undergone processes of semantic bleaching (loss of semantic meaning) and pragmatic strengthening (gain in pragmatic functions), which can point at a polysemous network of functions for so. This, in its turn, raises the question of the existence of a core meaning (cf. Aijmer, 2002 for among others sort of and oh, and Fox Tree and Schrock, 2002 for you know and I mean). Schiffrin's (1987) depiction of so as a marker of ‘result’, on the one hand, and of hypertactic transitions on the other, appear to befit our description of so's common features. The polysemy hypothesis is, however, solely based on the scrutiny of synchronic data, and can only be falsified in a diachronic approach (which also takes so's non-discourse marker uses into account), which has so far remained a blind spot for so.10 Moreover, teasing out a core function of so would equally involve a diachronic study, as well as a broad investigation of the marker in a wide variety of registers and setting it off against other markers fulfilling similar functions (Aijmer, 2002:23ff.). 5.2. In search of an account for the quantitative analysis of so Some tentative explanations can be proffered for the quantitative results sketched in section 4.2, viz. the differences in the frequencies of the ten functions of so, the differences between the native speakers and the learners in the corpus, and those between the students of Commercial Sciences and English Linguistics. 5.2.1. Frequencies of so's functions The frequencies of the functions attested for so in the corpus seem to tally with their relative position in the multifunctional spectrum of so: the more proximal a function is situated to ‘resultative so’ the higher its incidence, and the more distal it is the lower its incidence. For example, ‘Draw a conclusion’ solely differs from ‘resultative so’ in terms of the type of relation that it indexes (interpersonal rather than ideational), whereas ‘discourse-initial’ tokens of so are situated on the textual level, do not lead to closure but rather its opposite and do not even appear to link two discourse segments. Interestingly, the four most prominent functions of so in the corpus (‘result’, ‘prompt’, ‘summary’ and ‘conclusion’) are closely aligned in terms of the three distinguishing features for so's function (section 5.1): they index a hypertactic relation, provide closure to a stretch of speech, and are the functions in which so's ‘resultative’ meaning is most readily recognisable. The relative incidence of so's functions indubitably also ties in with the ‘‘employability’’ of these functions in discourse: the textual functions, in which so carries least semantic weight, appear to be used for specific environments that are less common than indicating a result or drawing a conclusion. For example, opening a conversation and leading the discourse out of a digression are not transformations that a speaker has to signal with a particularly high frequency in a single discourse event. The most notable exception to this tendency is the summative use of so, which ranks third in the corpus overall in spite of its status as a textual function. This too can be related to this function's employability in the speech
10 The entry of so in the Oxford English Dictionary Online suggests, though, that so has been grammaticalised from its earlier sense ‘in this way’ to its ‘resultative’ use. The further developments are more recent, and the processes of grammaticalisation appear to be on-going, which might imply that the textual functions have not reached their full potential yet.
1778
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
context. In an interview setting interviewees answer questions in often fairly lengthy turns in which they try to make their points clear. Providing a more general summary of an argumentation or story allows them to state the main point of what precedes to bring closure to their answer, their turn or to a distinct part of their answer or turn. In other words, this category is a prime example of how so can fulfil a pivotal role in discourse. Finally, tokens that fulfil an ideational or interpersonal function simultaneously help shape the discourse, and therefore also contribute to the textual domain -- albeit not as saliently as those solely fulfilling a textual function -- and are therefore more likely to carry more currency. 5.2.2. Native versus learner practice The language learners employ so with a similar functional scope as the native speakers, in that no discourse marker functions are found in just one of the sub-corpora and in each sub-corpus almost all functions are found in a similar number of interviews. However, some functions, most notably ‘elaborative’ and ‘self-corrective’ so, occur with an insignificant frequency in native speech but are significantly more prevalent in learner speech. On the one hand, it could just be that the language learners resort to these strategies because they use more vague terms that require elaboration than their native peers, and they are certainly more likely to require repair of their own utterances. On the other hand, functions with a peripheral status in so's spectrum of use may see their incidence increased if learners have so take them on more than native speakers due to an avoidance strategy. As has been demonstrated in previous investigations (Buysse, 2010, 2011), the learners in this corpus steer clear of elements that manage social rapport like ‘‘involvement discourse markers’’ (Romero Trillo, 2002:84) such as you know and I mean, because these are typically associated with informal speech and their use can be socially sanctioned. ‘‘Operative discourse markers’’ (Romero Trillo, 2002:84), like so and now, on the contrary, are inherent to teacher talk (Hellermann and Vergun, 2007) and are less register-sensitive because they first and foremost contribute to the coherence of the interaction. Since there is no clear boundary between these two types, limited functional overlap between certain markers may occur. In this vein, I mean characteristically introduces an ‘elaboration’ or ‘reformulation’ (see e.g. Fuller, 2003; González, 2004), in much the same way as so does in the categories ‘Introduce elaboration’ and ‘Mark self-correction’ (Buysse, 2009). Similar to the non-prefatory use of so with a ‘prompting’ function, you know typically ‘‘invites the addressee to complete the argument by drawing the appropriate inferences’’ (Jucker and Smith, 1998:196). The language learners, however, choose a safe option by expanding the frequency of use of peripheral functions of a discourse marker they are familiar with and is not deemed highly register-sensitive, i.e. of so. The reverse process may explain the one discourse marker function that is more popular in NS than in the two learner sub-corpora: ‘sequential’ so. In ‘sequential’ uses of so the ‘result’ meaning has faded to a more general ‘temporal’ meaning -- segment Y does not ‘follow from’ X but simply ‘follows’ it -- which is at the core of then, another operative discourse marker that is frequent in English conversations (Biber et al., 1999:796--797). Interestingly, then also has the (limited) capacity to convey a ‘resultative’ or ‘inferential’ meaning (Biber et al., 1999:883). Speakers are, therefore, more likely to choose then if they want to introduce a new sequence in a story than so, and vice versa. In the corpus at hand then in its ‘sequential’ function indeed occurs significantly more often in the learner sub-corpora than in the native corpus.11 Obviously, an in-depth investigation of then as a sequential marker is called for to ascertain whether this is the only factor in play to explain this difference, but at the very least it offers some insight into the underuse of ‘sequential’ so by the language learners in the corpus compared to native speaker practice. Interestingly, Müller's (2005) report of underuse of so among German learners of English (see section 1) makes her results diametrically opposed to those of the current investigation. How could this difference be explained? First, the text types of these studies are not identical: Müller's corpus was based on an experimental setting with monologic and dialogic story retellings, which did not involve the same kind of question--answer structure nor the same kind of relationship between the interlocutors. This may indeed explain a difference in frequency between the two corpora in the use of so in general across speaker profiles, as it may explain differences in relative order between the functions of so. For example, ‘Draw a summary’ ranks highly in the present corpus, whereas Müller (2005:76--79) merged this function with ‘rewording’ and ‘giving an example’, among others because on their own they would not be sufficiently frequent to constitute a proper category (e.g. ‘summarising’ was only attested 16 times in Müller's corpus). The specific activities the participants in Müller's study engaged in (particularly dyadic story retellings) probably did not give them as many opportunities to use so in this function as in an interview setting. What speech context is unlikely to have influenced, though, is the overuse by the Dutch and the underuse by the German learners of English. Incidentally, a rough preliminary analysis of the German sub-corpus of LINDSEI, in which the discourse marker tokens of so have merely been singled out and counted, reveals a significantly lower incidence of so in
11 Then has a relative frequency of 4.85 (n = 146) in CS, 4.05 (n = 135) in EL and 2.84 (n = 148) in NS ( p < 0.0001 for CS--NS and p < 0.01 for EL--NS; LL 20.59 and 8.69, respectively).
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
1779
the German interviews in the project than in NS (8.06 tokens per 1000 words for the German component of LINDSEI; p < 0.05), in spite of the identical speech context. Second, Müller (2005:251) puts the lower frequency of so in her learner corpus down to negative L1 interference, viz. the learners may have avoided so because of its close formal resemblance to the German particle so, which can also be used at the start of an intonation unit, and the German adverb also. Dutch-speaking learners of English are unlikely to turn to a similar avoidance strategy, because the functional equivalent of the discourse marker so in Dutch (dus) does not resemble it. Third, the authoritative Dutch explanatory dictionary Van Dale Groot woordenboek van de Nederlandse taal (Van Dale, 2005:887) reports that dus can occur as a conjunction that indicates that the second segment in coordinated clauses or phrases is a consequence of the first, or that introduces a conclusion. It can also occur as an adverbial conjunct expressing consequence, or in informal speech it can be used as ‘‘stoplap’’ (2005:887; ‘‘a stop gag’’), which often has a summarising and even an expletive function, in which case it can occupy just about any slot in a clause. Dutch dus would, therefore, appear to be highly frequent in Dutch speech, which might have had an influence on the learners’ use of its equivalent in English. For want of research on dus in Dutch spoken language, a systematic comparison of the two markers’ functional scope is still lacking. 5.2.3. CL versus EL Contrary to what might be expected, the participants studying English Linguistics overuse so to a greater extent than those studying Commercial Sciences. This should not be considered problematic, because the difference in frequency between the two learner groups does not yield statistical significance (except for ‘resultative’ so), and overuse of an operative discourse marker like so is unlikely to make language learners sound non-nativelike nor is it likely to be socially sanctioned. Nonetheless, it seems remarkable that in making more intensive use of the discourse marker functions of so than their CS peers, the EL participants have struck a wider gap with the native speakers in the investigation than the CS participants, resulting in a more strongly significant difference between EL and NS than between CS and NS. As they use discourse markers more frequently in general (Buysse, 2010), it could be conjectured that on the whole the EL interviewees are more aware of the need to employ pragmatic devices than their CS peers. Four observations support this thesis. First, only one difference between CS and EL achieves statistical significance (‘resultative’ so). Therefore, it is statistically possible that the frequencies for the two learner sub-corpora actually diverge less from each other than the current numbers show. Second, statistically significant discrepancies between EL and NS can be observed for only four of the ten functions (‘conclusive’ so, ‘elaborative’ so, ‘Self-correction’ and ‘Prompt’), which curtails the scope of the gap between these two sub-corpora. Third, for six of the ten functions EL frequencies are closer to NS practice than CS frequencies, the exceptions being ‘conclusive’ so, ‘summary’, ‘elaborative’ so and ‘self-correction’. Fourth, ‘elaborative’ so and ‘self-correction’ stand out, because so shares these functions with involvement discourse markers like I mean (see section 5.2.2). In this light, some of the practices laid bare for the EL sub-corpus could well have been meant to meet these interviewees’ pragmatic needs. Also ‘Prompt’ could be added to this list, given its partial resemblance to the involvement discourse marker you know. As a result, EL as well as CS interviewees turn to so, one of the most frequent words in English, more often than native speakers do, and the former do so to a greater extent because they are more acutely aware of the beneficial nature of paying sufficient attention to the pragmatics of the interaction in the target language. 6. Conclusions So is among the most popular discourse markers in native and learner speech, making it particularly suited for an investigation of discourse marker use in learner language. In a corpus made up of English interviews with native speakers of English and of Dutch I have identified ten distinct functional categories of so, thereby revealing it as a decidedly versatile marker that demonstrates an aptitude to index ideational, interpersonal as well as textual relations. The interrelatedness of these functions depends on the degree to which each displays three main characteristics: indicate a ‘resultative’ relation, a hypertactic shift and a point of closure. All functions are found in native and learner speech alike, but contrary to previous findings for learners of English with a different mother tongue background, the Dutch-speaking EFL learners show a higher incidence for the marker than their English peers. Those learners that major in English Linguistics turn to so even more than those that major in Commercial Sciences. If individual functions are considered, the differences between both of the learner sub-corpora on the one hand, and NS on the other, are only statistically significant for ‘Draw a conclusion’, ‘Prompt’ and ‘Indicate an elaboration’; for EL and NS ‘Mark self-correction’ should be included too. I have argued that this may be explained by extensive exposure to so in settings familiar to foreign language learners, as well as to an avoidance strategy, whereby language learners steer clear of markers associated with informality (such as you know and I mean) and use so instead for functions that they have in common. Some limitations to this study should be noted. First, its findings are confined to the context of an informal interview, which is a context all foreign language learners are acquainted with (if only from its resemblance to oral exams) but is likely
1780
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
to yield different results from a more informal setting. Second, for practical purposes the number of interviews in each subcorpus is fairly limited, although the learner sub-corpora can be claimed to be quite representative of the population of Belgian, Dutch-speaking students of English Linguistics or Commercial Sciences. Third, a thorough investigation mapping all functions of so's Dutch counterpart dus is desperately needed to gauge the potential extent of positive L1 transfer. Finally, the multifunctionality of so is to be accounted for in a diachronic study, which should be able to falsify the polysemy hypothesis to lay bare the relations between its various discourse marker functions. Much work remains on the shelf, but this investigation is merely meant to give a modest initial impetus to many more. Acknowledgement Thanks are due to the Centre for English Corpus Linguistics (Université Catholique de Louvain-la-Neuve, Belgium) for granting access to the recordings of the LOCNEC corpus. My gratitude also goes to Anne-Marie Simon-Vandenbergen and two anonymous referees for their invaluable comments on an earlier version of this article. Appendix. Transcription conventions The transcription conventions have been taken over from those initially laid out for the LINDSEI Project at the Centre for English Corpus Linguistics at the Université Catholique de Louvain-la-Neuve. <\Ir> <\Ie> ... [] : [ = italics
Interviewer turn starts Interviewee turn starts Interviewer turn ends Interviewee turn ends Empty pause: . (short), .. (medium), . . . (long) Phonetic annotations (e.g. the[i:], a[ei]) Vowel lengthening (e.g. so:) Overlapping speech Truncation Foreign words Non-verbal sounds (e.g. coughing, laughing, sighing) and contextual comments (e.g. someone entering the room) are specified in angle brackets
References Aaron, Jessi Elana, 2004. ‘‘So respetamos un tradicio´n del uno al otro’’: So and entonces in New Mexican bilingual discourse. Spanish in Context 1 (2), 161--179. Aijmer, Karin, 2002. English Discourse Particles. Evidence from a Corpus. John Benjamins, Amsterdam/Philadelphia. Aijmer, Karin, 2004. Pragmatic markers in spoken interlanguage. Nordic Journal of English Studies 3 (1), 173--190. Aijmer, Karin, 2011. Well I’m not sure I think. . . The use of well by non-native speakers. International Journal of Corpus Linguistics 16 (2), 231--254. Aijmer, Karin, Simon-Vandenbergen, Anne-Marie, 2009. Pragmatic markers. In: Ӧstman, Jan-Ola, Verschueren, Jef (Eds.), Handbook of Pragmatics 2009 (13th Installment). John Benjamins, Amsterdam/Philadelphia. Andersen, Gisle, 2001. Pragmatic Markers and Sociolinguistics Variation. John Benjamins, Amsterdam/Philadelphia. Biber, Douglas, Johansson, Stig, Leech, Geoffrey, Conrad, Susan, Finegan, Edward, 1999. Grammar of Spoken and Written English. Longman, Harlow. Blakemore, Diane, 1988. ‘So’ as a constraint on relevance. In: Kempson, Ruth M. (Ed.), Mental Representations: The Interface Between Language and Reality. Cambridge University Press, Cambridge, pp. 183--195. Bolden, Galina B., 2006. Little words that matter: discourse markers ‘‘so’’ and ‘‘oh’’ and the doing of other-attentiveness in social interaction. Journal of Communication 56 (4), 661--688. Bolden, Galina B., 2009. Implementing incipient actions: the discourse marker ‘so’ in English conversation. Journal of Pragmatics 41 (5), 974--998. Buysse, Lieven, 2009. So as a marker of elaboration in native and non-native speech. In: Slembrouck, Stef, Taverniers, Miriam, Van Herreweghe, Mieke (Eds.), From Will to Well: Studies in Linguistics Offered to Anne-Marie Simon-Vandenbergen. Academia Press, Gent, pp. 79--91. Buysse, Lieven, 2010. Discourse markers in the English of Flemish university students. In: Witzcak-Plisiecka, Iwona (Ed.), Pragmatic Perspectives on Language and Linguistics, Vol. 1: Speech Actions in Theory and Applied Studies. Cambridge Scholars Publishing, Newcastle upon Tyne, pp. 461--484. Buysse, Lieven, 2011. The business of pragmatics: the case of discourse markers in the speech of students of Business English and English Linguistics. ITL: International Journal of Applied Linguistics 161, 10--30. Byron, Donna K., Heeman, Peter A., 1997. Discourse marker use in task-oriented spoken dialog. In: Proceedings of the Fifth Biennial European Conference on Speech Communication Technology (Eurospeech ‘97), Rhodes, Greece, September 1997.
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
1781
Cuenca, Maria-Josep, Marín, Maria-Josep, 2009. Co-occurrence of discourse markers in Catalan and Spanish oral narrative. Journal of Pragmatics 41 (5), 899--914. Degand, Liesbeth, 2001. Form and Function of Causation: A Theoretical and Empirical Investigation of Causal Constructions in Dutch. Uitgeverij Peeters, Leuven. Fischer, Kerstin (Ed.), 2006. Approaches to Discourse Particles. Elsevier, Amsterdam. Fox Tree, Jean E., Schrock, Josef C., 2002. Basic meanings of you know and I mean. Journal of Pragmatics 34 (6), 727--747. Fraser, Bruce, 1999. What are discourse markers? Journal of Pragmatics 31 (7), 931--952. Fuller, Janet M., 2003. Discourse marker use across speech contexts: a comparison of native and non-native speaker performance. Multilingua 22, 185--208. Fung, Loretta, Carter, Ronald, 2007. Discourse markers and spoken English: native and learner use in pedagogic settings. Applied Linguistics 28 (3), 410--439. Gilquin, Gaëtanelle, De Cock, Sylvie, Granger, Sylviane, 2010. Louvain International Database of Spoken English Interlanguage. Presses Universitaires de Louvain, Louvain-la-Neuve. González, Montserrat, 2004. Pragmatic Markers in Oral Narrative: The Case of English and Catalan. John Benjamins, Amsterdam/Philadelphia. Granger, Sylviane, 2009. The contribution of learner corpora to second language acquisition and foreign language teaching: a critical evaluation. In: Aijmer, Karin (Ed.), Corpora and Language Teaching. John Benjamins, Amsterdam/Philadelphia, pp. 13--32. Halliday, M.A.K., Hasan, Ruqaiya, 1976. Cohesion in English. Longman, London. Halliday, M.A.K., Matthiessen, Christian M.I.M., 2004. An Introduction to Functional Grammar. Arnold, London/New York. Hansen, Maj-Britt Mosegaard, 1998. The Function of Discourse Particles: A Study with Special Reference to Spoken Standard French. John Benjamins, Amsterdam/Philadelphia. Hellermann, John, Vergun, Andrea, 2007. Language which is not taught: the discourse marker use of beginning adult learners of English. Journal of Pragmatics 39 (1), 157--179. Johnson, Alison, 2002. So. . .?: Pragmatic implications of so-prefaced questions in formal police interviews. In: Cotterill, Janet (Ed.), Language in the Legal Process. Palgrave Macmillan, Hampshire, pp. 91--110. Jucker, Andreas H., Smith, Sara W., 1998. And people just you know like ‘wow’: discourse markers as negotiating strategies. In: Jucker, Andreas H., Ziv, Yael (Eds.), Discourse Markers. Descriptions and Theory. John Benjamins, Amsterdam/Philadelphia, pp. 171--201. Kyratzis, Amy, Ervin-Tripp, Susan, 1999. The development of discourse markers in peer interaction. Journal of Pragmatics 31 (10), 1321--1338. Lam, Phoenix W.Y., 2009. The effect of text type on the use of so as a discourse particle. Discourse Studies 11 (3), 353--372. Lam, Phoenix W.Y., 2010. Toward a functional framework for discourse particles: a comparison of well and so. Text & Talk 30 (6), 657--677. Local, John, Kelly, John, 1986. Projection and ‘silences’: notes on phonetic and conversational structure. Human Studies 9 (2--3), 185--204. Mukherjee, Joybrato, 2009. The grammar of conversation in advanced spoken learner English: learner corpus data and language-pedagogical implications. In: Aijmer, Karin (Ed.), Corpora and Language Teaching. John Benjamins, Amsterdam/Philadelphia, pp. 203--230. Müller, Simone, 2005. Discourse Markers in Native and Non-native English Discourse. John Benjamins, Amsterdam/Philadelphia. O’Keeffe, Anne, McCarthy, Michael, Carter, Ronald, 2007. From Corpus to Classroom: Language Use and Language Teaching. Cambridge University Press, Cambridge. Oxford English Dictionary Online. http://dictionary.oed.com/. Polanyi, Livia, Scha, Remco J.H., 1983. The syntax of discourse. Text 3 (3), 261--270. Polat, Brittany, 2011. Investigating acquisition of discourse markers through a developmental learner corpus. Journal of Pragmatics 43 (15), 3745--3756. Pulcini, Virginia, Furiassi, Cristiano, 2004. Spoken interaction and discourse markers in a corpus of learner English. In: Partington, Alan, Morley, John, Haarman, Louann (Eds.), Corpora and Discourse. Peter Lang, Bern, pp. 107--123. Rayson, Paul, Garside, Roger, 2000. Comparing corpora using frequency profiling. In: Kilgarriff, Adam, Berber Sardinha, Tony (Eds.), Proceedings of the Workshop on Comparing Corpora, Held in Conjunction with the 38th Annual Meeting of the Association for Computational Linguistics, 7th October 2000, Hong Kong. Hong Kong University of Science and Technology, Hong Kong, pp. 1--6. Redeker, Gisele, 1990. Ideational and pragmatic markers of discourse structure. Journal of Pragmatics 14 (3), 367--381. Redeker, Gisele, 1991. Review article: linguistic markers of discourse structure. Linguistics 29 (6), 1139--1172. Redeker, Gisele, 2006. Discourse markers as attentional cues at discourse transitions. In: Fischer, Kerstin (Ed.), Approaches to Discourse Particles. Elsevier, Amsterdam, pp. 339--358. Rendle-Short, Johanna, 2003. ‘So what does this show us?’: analysis of the discourse marker ‘so’ in seminar talk. Australian Review of Applied Linguistics 26 (2), 46--62. Romero Trillo, Jesús, 2002. The pragmatic fossilization of discourse markers in non-native speakers of English. Journal of Pragmatics 34 (6), 769--784. Sacks, Harvey, Schegloff, Emanuel A., Jefferson, Gail, 1974. A simplest systematics for the organization of turn-taking for conversation. Language 50 (4-1), 696--735. Sanders, Ted J.M., Spooren, Wilbert P.M., Noordman, Leo G.M., 1992. Toward a taxonomy of coherence relations. Discourse Processes 15, 1--35. Schiffrin, Deborah, 1987. Discourse Markers. Cambridge University Press, Cambridge. Schourup, Lawrence, 1999. Discourse markers: tutorial overview. Lingua 107, 227--265. Stygall, Gail, 2001. A different class of witnesses: experts in the courtroom. Discourse Studies 3 (3), 327--349. Traugott, Elizabeth Closs, 2010. (Inter)subjectivity and (inter)subjectification: a reassessment In: Davidse, Kristin, Vandelanotte, Lieven, Cuyckens, Hubert (Eds.), Subjectification, Intersubjectification and Grammaticalization. De Gruyter Mouton, Berlin, pp. 29--71. Van Dale, 2005 = den Boon, Ton, Geeraerts, Dirk (Eds.), Van Dale Groot woordenboek van de Nederlandse taal. 14th edition. Van Dale Lexicografie, Utrecht/Antwerpen. Van Dijk, Teun, 1979. Pragmatic connectives. Journal of Pragmatics 3, 447--456. Vandepitte, Sonja, 1993. A Pragmatic Study of the Expression and the Interpretation of Causality: Conjuncts and Conjunctions in Modern Spoken British English. Koninklijke Academie voor Wetenschappen, Letteren en Schone Kunsten van België, Brussel.
1782
L. Buysse / Journal of Pragmatics 44 (2012) 1764--1782
Waltereit, Richard, Detges, Ulrich, 2007. Different functions, different histories: modal particles and discourse markers from a diachronic point of view. Catalan Journal of Linguistics 6, 61--80. Wei, Ming, 2011. Investigating the oral proficiency of English learners in China: a comparative study of the use of pragmatic markers. Journal of Pragmatics 43 (14), 3455--3472. Wennerstrom, Ann, 2001. The Music of Everyday Speech. Prosody and Discourse Analysis. Oxford University Press, Oxford/New York. Wouk, Fay, 1999. Gender and the use of pragmatic particles in Indonesian. Journal of Sociolinguistics 3 (2), 194--219. Lieven Buysse is assistant professor of English Linguistics at the Hogeschool-Universiteit Brussel (University College Brussels, Belgium), where he teaches English Linguistics, Discourse Analysis and Translation Studies. His current research interests include the study of discourse markers, foreign language acquisition, discourse analysis and the discourse of interpreting.