Constructions: Emerging and Emergent 9783110229080, 9783110229073

This volume embarks on an exploration of the processual and dynamic character of grammatical constructions in emergence,

240 62 1MB

English Pages 337 [340] Year 2011

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Table of Contents
Constructions: Emergent or emerging?
Emergent grammar and temporality in interactional linguistics
Emergent grammar for all practical purposes: the on-line formatting of left and right dislocations in French conversation
Constructions vs. lexical items as sources of complex meanings. A comparative study of constructions with German verstehen
Online changes in syntactic gestalts in spoken German. Or: do garden path sentences exist in everyday conversation?
Between emergence and sedimentation. Projecting constructions in German interactions
Improvisation, temporality and emergent constructions
Verb-first conditionals in German and Swedish: convergence in writing, divergence in speaking
Action, prosody and emergent constructions: The case of and
On the emergence of adverbial connectives from Hebrew relative clause constructions
Recommend Papers

Constructions: Emerging and Emergent
 9783110229080, 9783110229073

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Constructions: Emerging and Emergent

linguae & litterae

6

linguae & litterae Publications of the School of Language & Literature Freiburg Institute for Advanced Studies

Edited by

Peter Auer · Gesa von Essen · Werner Frick Editorial Board Michel Espagne (Paris) · Marino Freschi (Rom) Erika Greber (Erlangen) · Ekkehard König (Berlin) Per Linell (Linköping) · Angelika Linke (Zürich) Christine Maillard (Strasbourg) · Pieter Muysken (Nijmegen) Wolfgang Raible (Freiburg)

6

De Gruyter

Constructions: Emerging and Emergent Edited by Peter Auer and Stefan Pfänder

De Gruyter

ISBN 978-3-11-022907-3 e-ISBN 978-3-11-022908-0 ISSN 1869-7054 Library of Congress Cataloging-in-Publication Data Constructions : emerging and emergent / edited by Peter Auer, Stefan Pfänder. p. cm. - (Linguae & litterae; 6) ISBN 978-3-11-022907-3 (alk. paper) 1. Grammar, Comparative and general - Coordinate constructions. I. Auer, Peter, 1954- II. Pfänder, Stefan. P293.C66 2011 415-dc23 2011027835

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. 쑔 2011 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany www.degruyter.com

V

Table of Contents

Peter Auer and Stefan Pfänder Constructions: Emergent or emerging? . . . . . . . . . . . . . . . .

1

Paul Hopper Emergent grammar and temporality in interactional linguistics . . . .

22

Simona Pekarek Doehler Emergent grammar for all practical purposes: the on-line formatting of left and right dislocations in French conversation . . . . . . . . . .

45

Arnulf Deppermann Constructions vs. lexical items as sources of complex meanings. A comparative study of constructions with German verstehen . . . . .

88

Wolfgang Imo Online changes in syntactic gestalts in spoken German. Or: do garden path sentences exist in everyday conversation? . . . . . 127 Susanne Günthner Between emergence and sedimentation. Projecting constructions in German interactions . . . . . . . . . . . . . . . . . . . . . . . . . 156 Thiemo Breyer, Oliver Ehmer and Stefan Pfänder Improvisation, temporality and emergent constructions . . . . . . . . 186 Peter Auer and Jan Lindström Verb-first conditionals in German and Swedish: convergence in writing, divergence in speaking . . . . . . . . . . . . 218 Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen Action, prosody and emergent constructions: The case of and . . . . 263 Yael Maschler and Susan Shaer On the emergence of adverbial connectives from Hebrew relative clause constructions . . . . . . . . . . . . . . . . . . . . . . 293

VI

Constructions: Emergent or emerging ?

1

Peter Auer and Stefan Pfänder

Constructions: Emergent or emerging ?

1

Introduction

In his contribution to this volume, Paul Hopper distinguishes between emergent and emerging grammar: “By ‘emerging’ we are entitled to understand the development of a form out of its surroundings, its epigenesis. The term ‘emerging’ is thus appropriate for the view of grammar as a stable system of rules and structures, which may ‘emerge’ (i.e., come into existence) out of a less uniform mix”. In contrast, the term “emergent” refers to “the fact that a grammatical structure is always temporary and ephemeral”. Emergent grammar is provisional, epiphenomenal to conversation and “consists not of sentences generated by rules, but of the linear on-line assembly of familiar fragments” and structure which is “constantly being elaborated in and by communication”. This volume embarks on an exploration of the processual and dynamic character of grammatical construct(ion)s in emergence, both from “emergent” and “emerging” perspectives. In both senses, grammar is modelled as highly adaptive resources for interaction. Among the questions addressed are: How can what initially appears to be construction x end up being construction y in on-line syntax? What are the local interactional needs which such processes respond to in the process of their emergence? Does the online (re-)modelling of a construction concern its syntactic or semantic side – or both? Moreover: Should emergent grammatical structures as they unfold in real time be seen as steps in the emerging of grammar? In this introduction, we will first sketch some defining elements of emergence as the term is used in disciplines outside linguistics (Section 2) before briefly introducing the notion of emergence in linguistics, mainly in the work of Hopper (Section 3). After a brief discussion of constructions as emergent gestalts (Section 4), we discuss the relationship between emergent and emerging structures in language (Section 5). We will close this introduction with a brief presentation of the contributions to this volume (Section 6) and a summary of the main features of emergence in language (Section 7).

2

2

Peter Auer and Stefan Pfänder

Emergence: Elements of a definition

Emergence is traditionally defined as an effect that is different from the sum of the effects of each causal conjunct (Mill 1843); it is thus a feature of complex systems. Emergence is a non-teleological process in which multiple factors are operative which lead to a new yet unpredictable element or feature in a given system. The non-teleological character of emergence means that the process is for the most part unintentional. The factors in play can function either synchronically or diachronically. In both cases, however, emergent structures have the quality of oversummativity, i.e. knowledge of the factors involved in the process of change does not lead to a complete understanding of the emerging or emerged phenomenon. This also holds for long-scale selection processes in evolutionary scenarios (Piattelli-Palmarini 1989). For instance, in hierarchically structured systems, the more complex properties of the higher levels cannot be reduced to the more primitive properties of the lower levels, although they are functionally dependent on them. Weak and strong perspectives of oversummative (emergent) processes can be distinguished (Stephan 1999). The weak claim is that the oversummative aspects of higher levels are currently unexplainable given our limited knowledge of the working of the lower levels, but that such an explanation is possible in principle. The strong position challenges this possibility and argues that it is a priori impossible. Emergent phenomena are often self-reinforcing (auto-referential) and thus show circular causality. This can be illustrated by many events in the physical world. The surface of sand dunes, for example, is an emergent phenomenon in which the interplay of the air steams and ripples in the sand creates the waves. Starting with a slight unevenness of the surface, the grooves become deeper and deeper by means of self-reinforcement. Emergentist approaches differ from mechanistic ones, according to which the results of a development consist of an increase in complexity of certain systems, but which does not result in a qualitative change (cf. Beckerman, Flohr and Kim 1992). Contrary to this view, emergentist approaches assume the novelty of certain properties which hence alter the system as a whole. Novelty occurs whenever the first exemplar of a new type is instantiated (Stephan 1999: 18). Emergence has become a much-discussed and disputed concept in different disciplines. For instance, it plays an important role in the life sciences. A living organism is constituted by organs, which are constituted by cells, etc. But life cannot be explained by the functionality of the organs or their constituting elements alone. There has to be a complex interplay between the organs in order for life to emerge. Another illustrative example of (synchronic)

Constructions: Emergent or emerging ?

3

emergence in complex biological systems is the swarming behaviour of birds and fish. The swarm has a “behaviour” of its own which cannot be grasped by describing the movements of one of the individuals (Mielke 2008: 79, Blitz 1992). As mentioned before, emergence can be studied using either a synchronic or diachronic approach: Synchronic emergentist approaches describe the structure of a system that has features which are not shared by any of the constituting parts of the system or their sum, while diachronic emergentist approaches focus on the development of novel features in time (cf. Stephan 1999: 68). Diachronic emergentist views in the life sciences maintain that new entities or phenomena can come into being via evolution in nature. They therefore opt against a crude form of reductive materialism, according to which all change is solely due to regroupings and reorderings of existing elements. The temporal dimension of emergence emphasizes the novelty of the higher-level properties and the unpredictability of the moment in which such a novel phenomenon occurs. In terms of evolutionary biology, for instance, genetic mutation happens by chance, i.e. its systemic effects and its time of occurrence is unpredictable. The question is whether this change will stabilize, thus causing a permanent change in the systemic structure. In neuroscience (Gregg 2003; Racine & Illes 2009), consciousness and mental capacities are often considered to be emergent from the neurological properties of the brain. While the neurons are the functional units which make up the material basis necessary for the computations involved in cognitive processing, the content of this process (e.g. the attitude towards an object, a belief or a desire) cannot be explained merely by looking at the functioning of the cells. It is much debated whether it is possible, as a matter of principle, to draw valid inferences about such content on the basis of a neurological inspection of brain states, i.e. patterns of neural activity associated with certain cognitive processes. In linguistics, finally, the term “emergence” has also been used in both a diachronic and synchronic sense. In diachronic studies of variation and change, emergence refers to the development of new linguistic forms or even new varieties (in particular, dialects). From a synchronic perspective, it refers to the unfolding of syntactic projects in real time.

3

Emergent, but not emerging: Paul Hopper’s approach to grammar

Hopper’s notion of “emergent grammar” falls within this latter domain. It has been developed and controversially discussed since the late 1980s as an approach to the study of (spoken) syntax (Hopper 1987, 1998, 2004, this vol-

4

Peter Auer and Stefan Pfänder

ume). Its main object of analysis is the creation of syntactic structures in real-time interaction. In contrast to most theories of grammar, and certainly those of a generative orientation, the emergent approach does not posit a priori linguistic knowledge in the human mind “which operate[s] on fixed categories like nouns and verbs, specif[ies] the forms of additive categories like those of case, tense, transitivity, etc., and restrict[s] the possible orders in which words can occur in a sentence” (Hopper 1987: 141). Rather, emergent grammar focuses on the collective sum of actual speakers’ experiences which is seen as the basis for the creation of new utterances without determining their structure. The notion of emergent grammar is therefore an oversummative theory in the sense discussed above. Hopper agrees with most current research in the usage-based paradigm which maintains that routines of language use are the basis of grammar. Language does not simply instantiate grammatical blueprints which are given by some kind of I-language in a theoretically non-interesting way; rather, the structure of language develops out of talk in real time. However, Hopper goes beyond most usage-based theories of language, questioning the very existence of “grammar” as a structure outside language use. The only thing we know for sure about grammar is that it is continuously changing with use. The grammatical resources we have at our disposal are the structures we have experienced in concrete speech situations before – obviously a highly malleable and individualistic part of our “knowledge”; we rearrange these bits and pieces anew every time we speak. Language, much like culture as a whole, is always “temporal, emergent, and disputed” (Clifford 1986: 19), and “its structure is always deferred, always in a process but never arriving” (Hopper 1998: 156). Grammatical structures “come and go in the speaker’s awareness according to whether they are often or rarely heard, and are not totally and simultaneously available to the speaker without regard to context (Hopper 1988: 164).” What remains of grammar then? At best, “vast collection[s] of subsystems” (Hopper 1988: 158) which constitute a more or less provisional and negotiable framework for communication. They are like Lego blocks which can be used to build something; the structure of this emergent building is only constrained, but never determined by, the shape of the blocks. Grammar cannot be a stable synchronous state, a system où tout se tient; rather, it is “epiphenomenal to the outgoing creation of new combinations of forms in interactive encounters” (Hopper, this volume: 26). As such, it is an abstraction of usage, sometimes useful but not the primary object of linguistic analysis. For Hopper, then, “emergent grammar” is not a theory of how new grammar comes into being (diachronic emergence) as it is, for instance, in gram-

Constructions: Emergent or emerging ?

5

maticalisation theory. He uses the terms “emergent” and “emerging” to distinguish the synchronic from the diachronic approach to emergence in language: The two follow a very different kind of temporality. Emerging grammar focuses on the resultative states and investigates how they are reached in time, while emergent grammar focuses on the processuality of an ongoing, temporally structured, never-finished process of “languaging”. Or, to paraphrase Oesterreicher (2001), emerging grammar research often starts out with an “inverted teleology”: Since linguists already know which forms finally made it into the canonical grammar of a given language, they filter out all aspects of variation in previous language stages that cannot be linked to this final outcome. What looks like a well-ordered and even logical process of structural emergence in time is in fact constructed by the linguist, since all competing, alternative or contradictory structures that also exist in the data, but have not reached the same kind of sedimentation, are simply disregarded. Although Hopper does not deny that sedimentation exists and indeed is the foundation of grammar (cf. Hopper and Thompson 2006), his interest in emergence is not historical. Rather, the fundamental assumption of emergent grammar is that structures are “unfinished and indeterminate” (Hopper, this volume: 28); consequently, the aim of emergent grammar research is not to filter out ongoing processes of grammaticalisation, but to show how speakers go about producing structured utterances which cannot be explained entirely by the rules of canonical (or even spoken) grammar. It is this interest in the non-explained and non-explainable bits and pieces, the seemingly ungrammatical, peripheral or ad hoc forms which perhaps most clearly distinguishes emergent grammar from emerging grammar research. If one considers Hopper’s examples for his approach in more detail, they seem to be of two kinds; the first kind deconstructs the grammatical patterns of standard grammars with their written basis in addition to theories (usually of structuralist–generativist provenience) which are based on notions of canonical, introspective syntax. The second examples are much more radical in that they focus on utterances that operate seemingly without grammar, i.e. those in which speakers arrange constructional patterns in a novel and improvised way. For instance, speakers may superimpose various utterances, or they may change their constructional orientation midway in the course of the production of an utterance. The first kind of argument is well known from corpus-based spoken language research (cf. e.g. work on English matrix clauses containing verba sentiendi by Thompson and Mulac (1991); on relative clauses by Fox and Thompson (2007); or on biclausal constructions in German by Günthner (2008). For instance, Hopper (2001, 2004) argues that the pseudocleft construction

6

Peter Auer and Stefan Pfänder

of the canonical format what we need is more money – which like all “extractions” has been a favourite topic of generative syntax for many decades, but has usually been investigated on the basis of introspection only – is not widely found in corpora of spoken language. Here, the constructional fragment what +Subj-V is often followed by a stretch of talk that is only weakly (if at all) integrated with the what-part (cf. also Günthner 2006; Auer 2009). Hopper concludes that the structure of a pseudocleft as it emerges in discourse has more to do with temporal planning and serialisation than with grammar. The second, more radical kind of argument can be exemplified by the following example, reproduced here from Hopper’s contribution to this volume for convenience: DORIS:

... Sam has been, .. has taken such an interest in this retirement bit. .. (H) ... That it– .. it really surprises me. ANGELA: .. Well she’s begun to listen. DORIS: .. Yes she has.

There is good evidence that Doris starts out in the second line by making use of the constructional scheme intensifier + indef.art. + noun. The intensifier such is stressed, and the meaning of the whole utterance is that of an evaluation or assessment. The intensifier is not anaphoric or cataphoric (cf. Auer 2006). Since line 2 ends with falling intonation, the utterance is also prosodically complete, and turn transition is possible at this point. However, it happens that (perhaps due to Angela’s non-response) Doris changes her project and retrospectively recategorizes … such an interest … as the first part of a bipartite construction in which stressed such is a cataphoric device which projects a following that-clause. This projection is fulfilled by the clause produced in lines 3 and 4. The example proves Hopper’s point that the production of an utterance (and therefore, its grammatical structure) is not simply an instantiation of an underlying grammatical pattern, but that speakers can change their plans “on the fly” and shift from one constructional scheme to the next. But there are even more interesting examples for grammatical improvisation than such ad hoc reanalyses. These are examples in which there seems to be no underlying grammatical pattern which can account for the emerging structure although it comes close to one or more of them. Take the following examples from Hopper (2004): (a) Well, that is what we were trying to decide is whether there were any of those or whether we felt - - (CSPAE) (b) I mean this is what worries me is the evidence you see.

Constructions: Emergent or emerging ?

7

In these apo koinu structures, the speaker starts out with a presentative construction (this/that is what X), which could already terminate the utterance. S/he then takes the last part of the utterance (the what …-clause) and turns it into the first element of a pseudocleft construction which needs as a continuation a copula and a predicative clause (in the first example) or a noun (in the second example). The beginning of the utterance including the koinon (i.e. what we were trying to decide/what worries me, respectively) is therefore wellformed, as is the stretch of talk beginning with the koinon and continuing until the end of the utterance; however, the utterance as a whole, although delivered as one prosodic unit, does not “represent” an underlying format. Many researchers following the paradigm of interactional linguistics subscribe to Sacks’ (1995) dictum that there is “order at all points” in interactional language (cf. particularly Schegloff ’s introduction). Still, they also tacitly agree that there are always phenomena in our recordings or even in our transcripts in which word order is very difficult to explain. In other cases, it seems that no matter what analysis we arrive at, the structure still remains ambiguous. Often we leave these examples aside and do not talk about them, at least in our published work; if only for reasons of space, we focus on the “good” examples which instantiate the phenomenon in question in a more or less clear-cut way. Against this practice, Hopper argues that it creates the erroneous impression that any utterance found on a tape or in a transcript can unambiguously be assigned to a grammatical pattern. As the examples show, this is not true, even when the grammar we use for the description of conversational utterances is one which is based on and suited for the analysis of spoken language and does not have a written bias, and even when we leave out obvious repairs. Emergent grammar, on the other hand, is more interested in the vague boundaries of grammatical categories and units than in their prototypical centre, and its aim is to “explor[e] the leading edges and the territory around” (Hopper, this volume: 28); it wants to do justice to language structures that do not follow canonical patterns, that are not entrenched or sedimented, and that may be composed in an ad hoc fashion. The question is how they arise, not whether they can be discarded as irrelevant on the basis of an abstract notion of grammar. Although such an approach seems indeed radical in linguistics, it is of course not difficult to trace its roots outside the discipline. In recent papers, Hopper repeatedly mentions Giddens’ sociological theory of “structuration” as an inspiration and even suggests replacing the term “emergence” (in the sense of “emergent grammar”) with this term. Giddens’ structuration is indeed meant as a criticism of the way structure is described in structural and functionalist models of society. He only concedes to social structure the

8

Peter Auer and Stefan Pfänder

status of a “virtual order”, by which he means that “social systems, as reproduced social practices, do not have ‘structures’ but rather exhibit ‘structural properties’ and that structure exists, as time-space presence, only in its instantiations in such practices and as memory traces orienting the conduct of knowledgeable human agents” (Giddens 1984: 17). This is very much Hopper’s position, for whom language has no reality outside the practices and memory traces of its knowledgeable speakers. Even more evident is Hopper’s indebtedness to the Bakhtinian notion of dialogicity (cf. Bakhtin 1986; Linell 1998) and its post-structuralist adaptations. Hopper’s notion of deferral of structural closure echoes dialogical approaches which also stress that “our” language – the utterances we produce – is not really ours, but stems from a network of other voices that we have experienced on previous occasions and that are, however faintly, reflected in our words.

4

Constructions as emergent gestalts

While Hopper speaks of “emergent grammar”, the title of this volume replaces “grammar” with “construction”. The term “construction” is ambiguous (as is the term “grammar” in “emergent grammar”). It can be understood in a pre-theoretical and in a theoretical way. Pre-theoretically, it refers to any utterance which is complete in the sense that it constitutes an independent turn or at least a turn constructional unit, i.e. a self-contained turn component. It is a term which refers to the level of speech production and interpretation, not the level of grammatical knowledge. In spoken language, syntactic structures often do not conform to sentences in the sense of schoolbook grammar. They may be highly elliptical and often lack all the ingredients of a “proper sentence” and often consist of only one word. In this context, the term “construction” offers a convenient way of avoiding the problematic and presupposing notion of a sentence. In addition, constructions can be seen as emergent gestalts, i.e. units whose non-completion or completion is hearable on the basis of projections operating at any level of their unfolding in time, but which, at the moment they are completed, have all the qualities of an oversummative structure. Temporality and projection are essential components of emergent grammar. On a more theoretical level, “emergent constructions” also alludes to construction grammar which, of course, comes in many forms. In most variants, construction grammar is part of cognitive linguistics and not very prone to conceding central status to issues of on-line emergence (cf. the overview in Croft and Cruse 2004). Rather, a grammar is usually defined by construction grammarians as a structured inventory of interrelated conven-

Constructions: Emergent or emerging ?

9

tional patterns – linguistic signs of varying complexity – which are located in the speaker’s mind as simple words (Langacker 1987). However, there are also more usage-based types of construction grammar (cf. Fox and Thompson 2007; Auer 2006; Birkner 2008), which share a number of Hopper’s concerns. For instance, one of his arguments against canonical syntax patterns which are defined based on the linguist’s intuition is that corpus-based analysis often reveals a strong amount of idiomatisation or lexical fixation of the pattern. For instance, the English pseudocleft construction mentioned above is usually used with a small number of verbs (to be, to say and a few others; Hopper 2004). If this is the case, one might consider an alternative grammatical description which posits a number of clefting constructions, containing one of these verbs each, and each tied to a particular discourse function. Hopper’s approach and certain variants of construction grammars share scepticism towards broad-sweeping generalisation in linguistic description and abstract, maximally general linguistic “rules”; they share a concern with the more specific, sometimes lexically specified and even prefabricated types of constructions. There is much less agreement, though, as to whether a construction is real as a cognitive unit or whether it only is an abstraction derived by the linguist from language use. Hopper definitely favours the second alternative: “Canonical constructions should […] be seen as highly stylized cultural artifacts, amalgamations of fragments put together” (2001: 125–126). A similar position is also taken by Linell, who expresses his reservations vis-à-vis construction grammar by arguing that setting up inventories of constructions is merely the result of “decontextualising activities by linguists and other language cultivators” (2005: 43). According to Imo (this volume), linguists are the only ones involved and interested in “a more or less irrelevant – almost artistic – game of inventing structure”; the language users themselves simply build on previously heard or used utterances, and they reuse these structures in ways which always vary slightly. The issue is not whether language comes into existence anew in each speaking situation, for Hopper would readily concede the role of “prior text” – after all, it is this prior text which the present utterance refers back to dialogically. Furthermore, since this prior text needs to be remembered, it is beyond question that memory (and therefore cognition) plays a role in languaging. Rather, the question is how prior text is represented in the mind: as concrete utterances remembered in their individual shape, or as more or less abstract patterns filtered out of this prior experience?1 1

The question is also discussed in so-called exemplar theory, particularly in phonetics and phonology (cf. Pierrehumbert 2001).

10

Peter Auer and Stefan Pfänder

Hopper would agree that there is a continuing tension between sedimentation and innovation or improvisation (also cf. Günthner, this volume), but it is not clear whether he would also concede that abstraction is a necessary condition for the indexicality of language. Some contributions in this volume (such as Deppermann’s) take exactly this latter view and argue that the concrete utterance which is produced at a certain moment in an interaction is contextually overdetermined by it; for grammar (constructions) to be a useful instrument in such over determined contexts, it therefore needs to be under determined – i.e. constructions need to be abstract in order to be flexible enough to be used in a multitude of singular situations (also cf. Auer and Günthner 2005, with reference to Hartmann 1959). It is precisely a grammar’s abstractness which could be argued to render it “a resource for rhetorical concerns of local (re)interpretation” (Deppermann, this volume: 120). This is also the meaning of the term we follow in this introduction.

5

Emergent vs. emerging – how large is the gap really?

As we have seen above, Hopper argues for a strict separation of emergent grammar research and emerging grammar research. Not all contributors to this volume subscribe to this strong dichotomy, however. For instance, Pekarek Doehler (this volume), while situating her paper squarely in the Hopper paradigm of emerging grammar, hints at the possibility that emerging and emergent grammar might be inextricable after all. She shows how a given grammatical format can be reconfigured according to “locally occasioned interactional needs” and sometimes looks “patched together within a moment-by-moment temporally organised process” (81). But she also argues that this process of adaptation in the end might lead to a different canonical structure, i.e. to language change. This raises the question of whether emergent and emerging approaches are indeed antagonistic, an issue we turn to in this section. To begin with, it is not clear whether Hopper’s portrayal of “emerging grammar” is based on a contingent critique of (some parts of) research on grammaticalisation and language change in general as it is practiced today, or whether it is an a priori argument against any kind of diachronic analysis. It seems to us that the “inverted teleology” which is typical of much of diachronic research is not inherent to the investigation of language change as such; it is possible to do research on language change and investigate nonteleological types of variation. It is obvious that in the beginning of any kind of grammatical or phonological change, we will find a certain amount of

Constructions: Emergent or emerging ?

11

variation between old and new forms, or between pragmatic and syntactic solutions to communicative problems/functions. The new forms will always start out as an invididual’s or a group of individuals’ innovations, patched together from elements of their previous experiences and ad hoc formulating practices, including improvisation. Many of these idiosyncractic forms will never survive, let alone sediment into grammatical structures of a language system. It may be very worthwhile to investigate these incipient forms of language change and to ask why they fail to win out against competing structures. In fact, at least two of Hopper’s examples show characteristic features of such incipient sedimentation as grammatical patterns (constructions); both, however, have not yet quite made it into the grammar. They seem to have been “locked” at the stage of incipient grammaticalisation for a long time, and have remained marginal in quantitative terms as well. The first example is apo koinu utterances as discussed above for English; from research on similar forms in German (Scheutz 1992, 2005) and Swedish (Norén 2007), we know that they have been an option throughout most of the history of these languages, with varying stages of popularity. Characteristically for incipient grammaticalisation, we find a whole range of variants of this format, from clearly pragmatically conditioned on-line phenomena (where hesitations and prosodic breaks during the delivery of the construction make it clear that some kind of reorganisation in time is taking place) to half-grammaticalised sedimentations of some particular variants out of the many apo koinu formats, which show a specific prosodic packaging and are linked to a single pragmatic function. In German, the most construction-like of these apo koinu formats are so-called mirror constructions, such as Das ist ein solcher Idiot ist das! ‘That is a such idiot is that’ Das ist ja unglaublich ist das! ‘That is PART incredible is that!’

in which the initial copula construction with an anaphoric element in the first position (das) ends in a predicative noun or adjective, which is then used as the first element in an inverted copula construction with the predicative element in first position. The predicative element functions as the koinon (boldface). The pattern shows all the features of a full-fledged construction such as delivery without a break before the koinon, as it is found in online composed variants, and a typical emphatic function resulting from the movement of the topical evaluative element from last to first position. Other apo koinu formats are much less frequent, while others are less generally ac-

12

Peter Auer and Stefan Pfänder

cepted and more likely to be heard as self-corrections, such as topicalizing apo koinus as in2 wie_s wegkommen sind war_er (-) zehn zwölf Jahr so was wird_er gwesen sein gell ja ‘when they were taken away he was (-) ten twelve years or so he must have been right yes’

In this case, the speaker hesitates in the production of the sentence when s/he reaches the age description. After the hesitation, s/he resumes the sentence by recategorizing the sentence-final predicative noun phrase into a sentence-initial one. (Also note that the emerging syntagm is less rigidly structured since the final part does not exactly mirror the beginning, but changes it by adding an epistemic auxiliary, wird.) There is some evidence that this pattern used to be more accepted than it is today even in writing, particularly in Middle High German texts, and that it was grammatically less restricted at that time; for instance, apo koinu formats in which two propositions are expressed which share one noun phrase (here in the role of the object) are unusual today but were widespread in earlier forms of German (here MHG)3 Rˇ olant uie mit paiden hanten/den guten Oliuanten/satzer ze munde ‘Roland took with both hands/the good olifant [a horn]/put he to his mouth’

A full historical account of the ups and downs of the apo koinu construction in German (or other Germanic languages with flexible word order) still needs to be written (however, see Scheutz 1992); what seems clear, though, is that the format has been around for a long time, that it has shown tendencies towards grammaticalisation, but that despite the “naturalness” of the inherent topic/ comment inversion which lends itself to functions such as emphasis or shift of perspective, the pattern has never completely sedimented, perhaps as a consequence of prescriptive grammars and their overt sanctioning of constructions containing two predicates. Apo koinu utterances have always remained somewhere between mere on-line emergence and grammaticalisation. The second example of an incipient grammaticalisation which has not fully made it into the grammar (presumably also for quite some time) is the hendiadyn or serial verb construction discussed in Hopper (2002, 2005). We know – if only from other languages in which grammaticalisation has proceeded much further – that the hendiadyns have the potential of developing 2 3

Example from Scheutz (1992: 258). From: Rolandslied des Pfaffen Konrad, 6653–6655 (from 1170).

Constructions: Emergent or emerging ?

13

into a full-fledged grammatical device; serial verb constructions are indeed a central feature of many languages (for instance in West Africa and South East Asia; cf. Crowley 2002; Aikhenvald and Dixon 2006). German, English and Swedish all contain beginnings of such a process. While fully developed serial verb constructions consist of two juxtaposed verbs not linked by a conjunction which express only one predication and often add aspectual meaning, the Germanic languages have only developed “weak” forms of serialisation of two verbs linked by a conjunction, such as in (a) Swedish4 Peter gick och läste en bok. ‘Peter went and read a book.’

(b) English5 They took the same design as before and enlarged it by including a library and a gymnasium.

(c) German6 (…) ich höre, wie alle zuschauen von dieser Regierung von ÖVP und FPÖ, wie die Bundesregierung hergeht und sagt, nein, wir haben keine Initiative, dass etwas besser wird, sondern wir schicken alle mit 50 Jahren in Pension ‘I hear how everybody in this government of ÖVP and FPÖ just watches, how the federal government comes and says, no, we have no initiative to make things better, but we send everybody into retirement at the age of 50.’

In the three examples given here, it is clear that the first of the two conjoined verbs is not used in its conventional meaning but has undergone semantic bleaching: neither is the design in the English example literally taken to any place, nor does the government in the German example “move” anywhere to say something. On the other hand, it is difficult to pinpoint the semantics of these quasi-serial verb constructions, and there are many examples (as shown by Hopper 2005) in which it is unclear whether we are already dealing with a serialised package of two verbs or two independent predications. As a third example of an emergent and perhaps also emerging construction, consider the cosa de QUOTE structure in (Argentinian) Spanish recently discussed by Ehmer (2009: Ch. 5.2.). He finds in his data “unusual” ways of introducing hypothetical or generalized (impersonal) direct speech by the noun cosa, as in 4 5 6

Fabricated example from Wiklund 2009: 181. From Hopper (2005). Quotation from the protocol of the Steirmark parliament, session of 16. 11. 2004, MP Schrittwieser.

14

Peter Auer and Stefan Pfänder No sé por qué la gente tiene esa cosa de “pero es tu cumpleaños tenés que hacer algo” ‘I don’t know why people have this thing of ‘but it is your birthday you need to do something’’

There are reasons to speak of an established construction of spoken language here, which obviously is modelled on the N de N construction in Spanish (the standard way of forming compounds) but extends it to whole sentences of quoted speech in the position of the second noun. In the most construction-like version, as in the example above, the first noun is the dummy element cosa (‘thing’) and the meaning of the construction is equivalent to that of a quotative verb. But again, in addition to this relatively wellestablished pattern, similar structures can be found in which it is much less clear whether we are dealing with a sedimented pattern. For instance, the structure often occurs in a syntactic context in which the previous utterance contained the standard construction N de N, with the second N-position filled by a non-finite verb form, i.e. the infinitive: Porque yo tengo esa cosa: de salvar; de “ay po::bre, vamos a sacrificarnos por”. ‘For I have this thing of saving of ‘oh poor you, let’s sacrifice ourselves for (it)’’

This suggests that the N de QUOTE structure copies and adapts a previously introduced pattern, i.e. it emerges on-line. Also, Ehmer finds related structures in which a semantically more rich noun fills the first N-position (such as: la actitud … de ‘no quiero a nadie’ ‘the attitude … of ‘I don’t like anybody’’), or in which a verbum dicendi is inserted after the first N (as in: esa sinceridad de decir ‘que me pasa con vos’ ‘this openness of saying ‘what have I got to do with you’’). All these cases testify to the ambiguous status of the format: Although there are some clear cases of an emerging construction, there are also cases in which the structure is clearly emergent, put together on the spot out of several pre-existing patterns in discourse. The three examples combine features of grammar as an emergent, on-line phenomenon, with features of incipient grammaticalisation, i.e. grammar emerging in time. Analysing them in terms of the first paradigm does not prevent us from pointing out that language change may also be taking place. No inverted teleology is needed and even possible, since the outcome of the process is unclear. Grammaticalisation has its beginnings in emergent grammar although the inverse does not hold: Emergent grammar may or may not lead to grammatical change.

Constructions: Emergent or emerging ?

15

Another argument for a reconciliation of emerging and emergent grammar starts from the opposite angle, i.e. improvisation. For improvisation to work, speakers and hearers must have a shared stock of expectations of what is to come next in the syntactic project. This is not possible without categorized linguistic experience, i.e. grammatical knowledge. Syntactic projects during their very emergence have to rely on certain expectations shared by hearers and speakers alike. These expectations are based on linguistic and social routines of interaction. There is no need to exclude these routines from an emergentist approach to spoken syntax; rather, this approach presupposes some categorized linguistic knowledge. An uncategorized set of previously heard utterances does not explain how improvising speakers play with expectations.

6

The contributions in this volume

Paul Hopper introduces the volume with his chapter “Emergent grammar and temporality in interactional linguistics” in which, on the basis of various examples, he shows how linguistic interaction unfolds in real time and how commonly used expressions get recycled in this process. Emergent grammatical structure is hence understood as ephemeral and epiphenomenal to the ongoing interaction. Speakers reassemble familiar fragments, as in the case of the so-called sluicing construction (we knew we were loosing oil, we did not know where). Another construction, the “such a/n”-construction, shows how the speaker is creating her grammar as she goes. Hopper elaborates previously neglected aspects of emergent grammar, emphasizing the openness of structure, i.e. its transformative aspects. The familiar fragments are not only put together in various well-known ways; a surprising combination may also lead to the constant modification and negotiation of constructions during use. Building directly on Hopper’s claims, Simona Pekarek Doehler also highlights the processual character of grammatical constructions in her chapter “Emergent grammar for all practical purposes”. Taking up the theme of left and right dislocations in French, she shows how speakers revise the syntactic trajectories on the fly. What initially appears to be a given construction type thus ends up as another construction type. Emergent grammar appears to be distributed over speakers and time and thus becomes a shared, yet highly adaptive resource for interaction. The recalibration of constructions as they unfold in real time allows speakers to address practical issues, i.e. local interactional needs, such as displaying alignment or inviting recipient action.

16

Peter Auer and Stefan Pfänder

Arnulf Deppermann also zooms in on the local management of constructions. His paper “Constructions vs. lexical items as sources of complex meanings” suggests that the precise local meaning of a construction emerges from the interplay of meaning potential and the ongoing adaptations within the conversational history of the participant’s uses of the construction in interaction. The study of two German construction types (verstehst du ‘do you understand’, and ich kann nicht verstehen ‘I cannot understand’) reveals that not only the formal side of grammatical constructions is emergent, but also their meanings. The findings of this paper provide further evidence for the claim that participants analyse both the syntactic and semantic features of constructions during their course of production. In his chapter “Online changes in syntactic gestalts in spoken German”, Wolfgang Imo argues that in German, so-called garden path sentences, with their typically unintended ambiguity, occur very rarely in corpora of everyday talk-in-interaction. One reason for this is the relatively strong morphology of German which leads to early disambiguation in on-line production. If garden path sentences are found at all, they are related to turn continuation or incrementation; instead of causing trouble for interactional processing, they are a resource for the adaptation of the actual syntactic project to local contingencies, such as turn management. Imo relates his findings to the exploitation of some aspects of the “potentialities of the system” which are used in dialogic interaction in real time. Susanne Günthner (“Between emergence and sedimentation”) shows that the unfolding of syntax in real time heavily relies on the (degree of) sedimentation of a construction. She argues that projection constructions such as was ich wichtig finde, ist, dass ‘what I think is important is that’ and die Sache ist, dass ‘the thing is that’ are open constructions. The study shows how projections can be deferred, i.e. not be dealt with immediately, remaining valid after the insertion of different linguistic material. Participants may take advantage of this deferral of the projected continuation as a cognitive and interactional space for thinking through what they are about to say. Thus, the openness of the format mirrors its interactive suitability as a resource for solving communicative tasks that range from integrating aspects of sequential context to indexing certain activities to managing interaction contingencies. Building on a similar assumption, Thiemo Breyer, Oliver Ehmer and Stefan Pfänder (“Improvisation, temporality and emergent constructions”) focus on situated interaction in collaborative story-telling in which the participants subvert canonical grammatical formats in a playful mood. The blending of semantic and syntactic formats from constructional resources that are theoretically incompatible is referred to as “improvising

Constructions: Emergent or emerging ?

17

grammar”. The authors suggest that the notion of improvisation can help to better model the indeterminacy or openness of linguistic structure. Improvisation brings along a moment of surprise – the very moment of speech production. Improvisation is attractive for the language users because of its unexpected character and for the theorist because it explains the raison d’être for some emergent structures which are composed of (constructional) fragments. In their diachronic and synchronic comparative corpus study, “Verb-first conditionals in German and Swedish: convergence in writing, divergence in speaking”, Peter Auer and Jan Lindström show that verb-first (V1) conditionals are used hardly at all in spoken German, but very frequently in the written language; for Swedish, no such restriction regarding the use of V1 in oral language holds, and Swedish conditional V1 constructions are semantically more focussed on conditionality and thus less open to non-conditional readings; they might be considered more grammaticalised than the corresponding German construction. The authors further argue that V1 conditionals represent a case of locally specified constructions in emergent discourse. One of the reasons for the diachronically emerging differences between the two Germanic languages is that in German these constructions are “too open”. They project too vaguely, since they have to compete with a large number of other colloquial V1 constructions. The construction therefore only survives in certain genres (i.e. legal texts) and speech activities (e.g. stating law-like regularities). Here, their projectional ambiguity is low, because the competing V1 constructions are exclusively used in the spoken language. Dagmar Barth-Weingarten and Elisabeth Couper-Kuhlen’s paper by means of VP constructions with and discusses togetherness as a contributor to structural emergence. The latter, they claim, implicates not only syntactic/semantic cohesion but also togetherness of action, and togetherness in prosodic/phonetic form, i.e. only those VP conjoins can fuse and become construction-like hendiadics such as go ahead and X, sit down and X (Hopper 2001a, 2001b) which are delivered as a single action and which exhibit a high degree of prosodic/phonetic integration. On this basis then, they argue that uni-actionality and prosodic/phonetic integration may also provide a tool for identifying incipient, i.e. emerging, constructions. Yael Maschler’s chapter “On the emergence of adverbial connectives from Hebrew relative clause constructions” looks into the incipient grammaticalisation of adverbial complementizers in Modern Hebrew. According to (prescriptive) Modern Hebrew grammars, the grammatical marking of relative clauses consists of two parts: a relativizer (she- ‘that’) and an obliga-

18

Peter Auer and Stefan Pfänder

tory (though sometimes omitted) coreferential element. The lack of the second (resumptive) element in spoken discourse is the starting point of the emergence of new grammatical functions of the former relative construction. She- is being reanalyzed on-line as an adverbial connective. This process starts in syntactic contexts where the relativizer she- is preceded by adverbs of time, etc. On (the day) (that they marry) is then re-segmented as (on the day that) (they marry). The re-segmentation through re-bracketing fuels ongoing grammaticalisation processes and thus leads to new adverbial connectives.

7

Conclusion

We conclude by summarizing the main points: 1. Real time From an emergent perspective, it is necessary to consider syntax in real time. While producing their syntactic projects on-line, speakers constantly monitor the other participants’ expectations and projections. 2. Sedimentation These projections rely on expectations fueled by more or less sedimented routines. Despite a scepticism regarding the categorization of linguistic experiences and longue durée sedimentation in emergent grammar, we suggest that both are necessary and indeed the basis of on-line syntax. 3. Gestalt These speakers’ categorizations, however, are not always captured well in linguists’ analyses of language structure. The categorizations seem to rely significantly more on gestalt-psychological similarity than on logic-semantic category systems. 4. Constant reanalysis In interaction, recipients (re-)analyse the on-line sound chain, which sometimes leads to a re-bracketing of the units produced by the speaker. 5. Improvisation There is no need to exclude routines from an emergentist approach to spoken syntax; rather, this approach presupposes categorized linguistic knowledge. An uncategorized set of previously heard utterances does not explain how improvising speakers play with expectations. 6. Mixed approach The opposition of emergent and emerging constructions can be overcome. Emergent structures are the basis of emerging constructions.

Constructions: Emergent or emerging ?

19

References Aikhenvald, A. & R. Dixon (eds.) 2006 Serial verb constructions, Oxford. Auer, P. 2006 Construction grammar meets conversation: Einige Überlegungen am Beispiel von ‘so’-Konstruktionen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 291–314. 2009 Projection and minimalistic syntax in interaction. Disc. Proc. 46, 2, 180–205. Auer, P. & S. Günthner 2005 Die Entstehung von Diskursmarkern im Deutschen – ein Fall von Grammatikalisierung? In: T. Leuschner, T. Mortelmans & S. de Groodt (eds.), Grammatikalisierung im Deutschen (Linguistik – Impulse und Tendenzen, 9), Berlin, 335–362. Bakhtin, M. 1986 Speech genres and other late essays, transl. by V. W. McGee, ed. by C. Emerson, Austin. Beckermann, A., H. Flohr & Kim, J. (eds.) 1992 Emergence or Reduction? Essays on the Prospects of Nonreductive Physicalism, Berlin. Birkner, K. 2008 Relativ(satz)konstruktionen im gesprochenen Deutsch: syntaktische, prosodische, semantische und pragmatische Aspekte, Berlin. Blitz, D. 1992 Emergent Evolution: Qualitative Novelty and the Levels of Reality, Dordrecht. Clifford, J. 1986 Partial Truths, in: J. Clifford & G. E. Marcus (eds.), Writing Culture. The Poetics and Politics of Ethnography, Berkeley, 1–29. Croft, W. & D. A. Cruse 2004 Cognitive Linguistics, Cambridge. Crowley, T. 2002 Serial Verbs in Oceanic: a Descriptive Typology, Oxford. Ehmer, O. 2011 Imagination und Animation. Die Herstellung mentaler Räume durch animierte Rede, Berlin/New York. Ellis, N. C. 1998 Emergentism, connectionism and language learning. Language Learning 48, 4, 631–664. Fox, B. A. & S. A. Thompson 2007 Relative clauses in English conversation: relativizers, frequency and the notion of construction. Studies in Language 31, 293–326. Giddens, A. 1984 The Constitution of Society. Outline of the Theory of Structuration, Cambridge. Gregg, K. R. 2003 The state of emergentism in second language acquisition. Second Language Research 19, 2, 95–128. Günthner, S. 2006 ‘Was ihn trieb, war vor allem Wanderlust’ (Hesse: Narziss und Goldmund). Pseudocleft-Konstruktionen im Deutschen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 59–90.

20

Peter Auer and Stefan Pfänder

2007 Brauchen wir eine Theorie der gesprochenen Sprache? Und: wie kann sie aussehen? Ein Plädoyer für eine praxisorientierte Grammatiktheorie. = gidi Arbeitspapierreihe (http://noam.uni-muenster.de/gidi/ 6, 1–22). 2008 Projektorkonstruktionen im Gespräch: Pseudoclefts, die Sache ist-Konstruktionen und Extrapositionen mit es. Gesprächsforschung – Online-Zeitschrift zur verbalen Interaktion (www.gespraechsforschung-ozs.de). Hartmann, P. 1959 Offene Form, leere Form und Struktur. In: H. Gipper (ed.), Sprache – Schlüssel zur Welt. Festschrift für Leo Weisgerber, Düsseldorf, 146–157. Hopper, P. 1987 Emergent grammar. Berkeley Linguistic Society 13, 139–157. 1992 Emergence of grammar. In: W. Bright (ed.), International Encyclopedia of Linguistics. Vol. I, Oxford, 364–367. 1998 Emergent grammar. In: M.Tomasello (ed.), The New Psychology of Language, Mahwah, NJ, 155–175. 2001 Grammatical constructions and their discourse origins: Prototype or family resemblance? In: M. Pütz, S. Niemeier & R. Dirven (eds.), Applied Cognitive Linguistics I: Theory and Language Acquisition, Berlin, 109–129. 2004 The openness of grammatical constructions. In: 40th Annual Meeting of the Chicago Linguistic Society, April 15th, 2004. Chicago, 153–175. 2005 Bi-clausal constructions and emergent grammar. Talk presented at the German Department of the Westfälischen Wilhelms-University Münster, April 2005. Hopper, P. & S. Thompson 2003 Grammaticalization, Cambridge. Langacker R. W. 1987 Foundations of cognitive grammar, Vol. 1: Theoretical prerequisites, Stanford, Calif. Linell, P. 1998 Approaching dialogue: talk, interaction and contexts in dialogical perspectives, Amsterdam. 2005 The written Language Bias in Linguistics, Abingdon, New York. Mielke, J. 2008 The emergence of distinctive features, Oxford. Mill, J. S. 1843 Of the Four Methods of Experimental Inquiry. In: A System of Logic, Raciocinative, and Inductive. Book 3. Reprint. Toronto: University of Toronto Press, 1974. Norén, N. 2007 Apokoinou in Swedish talk-in-interaction, Linköping. Oesterreicher, W. 2001 Historizität – Sprachvariation, Sprachverschiedenheit, Sprachwandel. In: M. Haspelmath, E. König, W. Oesterreicher & W. Raible (eds.), Language Typology and Language Universals, Berlin/New York, 1554–1595. Piattelli-Palmarini, M. 1989 Evolution, selection and cognition: from “learning” to parameter setting in biology and in the study of Language. Cognition 31, 1–44. Pierrehumbert, J. B. 2001 Exemplar dynamics: Word frequency, lenition, and contrast. In: J. Bybee &

Constructions: Emergent or emerging ?

21

P. Hopper (eds.), Frequency Effects and the Emergence of Lexical Structure, Amsterdam, 137–157. Racine, E. & J. Illes 2009 Emergentism at the crossroads of philosophy, neurotechnology, and the enhancement debate. In: J. Bickle (ed.), The Oxford Handbook of Philosophy and Neuroscience, Oxford et al., 431–453. Sacks, H. 1995 Lectures on conversation, Oxford. Scheutz, H. 1992 Apokoinukonstruktionen. Gegenwartssprachliche Erscheinungsformen und Aspekte ihrer historischen Entwicklung. In: A. Weiss (ed.), Dialekte im Wandel, Göppingen, 242–264. 2005 Pivot constructions in spoken German. In: A. Hakulinen & M.Selting (eds.), Syntax and Lexis in Conversation, Amsterdam/Philadelphia, 101–129. Stephan, A. 1999 Emergenz. Von der Unvorhersagbarkeit zur Selbstorganisation, Dresden. Thompson, S. & Mulac, A. 1991 A quantitative perspective on the grammaticization of epistemic parentheticals in English. In: E. Traugott & B. Heine (eds.), Approaches to Grammaticalization, Vol 2, Types of Grammatical Markers, Amsterdam, 313–327. Wiklund, A. 2009 The Syntax of Surprise: Unexpected event readings in complex predication, Working Papers in Scandinavian Syntax 84, 181–224.

22

Paul J. Hopper

Paul J. Hopper

Emergent Grammar and Temporality in Interactional Linguistics*

Doch auch die geringste Veränderung des Usus ist bereits ein komplizierter Prozess, den wir nicht begreifen ohne Berücksichtigung der individuellen Modifikation des Usus. Da, wo die gewöhnliche Grammatik zu sondern und Grenzlinien zu ziehen pflegt, müssen wir uns bemühen alle möglichen Zwischenstufen und Vermittelungen aufzufinden. (Hermann Paul, Prinzipien der Sprachgeschichte § 17)1

1

Introduction

In his essay Laokoon, oder, Über die Grenzen der Malerei und Poesie (1766), Gotthold Ephraim Lessing opened a question of central concern to students of language. Some kinds of art, Lessing points out, are permanently there for us to examine. Sculpture and painting are the prime examples. Others occupy a time span. These are music, drama, poetry, and other forms of literature. Asked which of these two varieties of experience language should be matched with, linguists could be expected to reply the latter. Speech is not a set of marks on a page, and is not to be confused with its written representation. Yet the task of describing language has always presupposed a reduction of a temporal medium to a fixed, stable, and timeless one, articulated in a terminology of parts and units. Only recently have a small number of linguists begun to “take temporality seriously,” as Auer (2000) has put it. * This article was completed during a stay as Senior Fellow at the Freiburg Institute for Advanced Study (FRIAS) during May 2008. I am grateful to Dr. Peter Auer and Dr. Werner Frick, the directors of FRIAS, for their support of this stay, and to Stefan Pfänder for much in-depth discussion of the issues. Thanks also to Peter Auer, Susanne Günthner, and Sandra Thompson for comments on the paper; they are in no way responsible for its contents. 1 “But even the tiniest change in usage is already a complex process that we cannot understand without taking into account the modification of usage by individuals. Where conventional grammar is wont to separate and draw boundary lines, we must strive to discover all the possible intervening stages and intermediate forms.” (Tr. PJH)

Emergent Grammar and Temporality in Interactional Linguistics

23

Placing temporality in the front rank has important implications for the practice of linguistics. Speakers do not possess a bird’s eye view of an utterance, but rather move forward in time through it (Franck 1985, Lerner 1991, Auer 1992, Auer 2000). The current speaker is always already situated at the leading edge of an unfolding interaction, and can only create new structure by a process of incrementation, adding bits and pieces to the utterance. In this paper, I will illustrate this process through examples from conversational transcripts, and comment on the notion of emergence that is the theme of this volume.

2

Fixed expressions and formulaicity

The possibility of an on-line perspective on grammar suggests an alternative to the standard lexical-item-and-rules model of linguistic description. In place of a uniformly available language system, we should reckon with a massively large pool of linguistic resources, one that is not uniformly available, and a few simple linear techniques, a “toolkit”,2 for combining them into structured utterances. This pool is a collection that consists of many different kinds of items, among which are: – constructions, that is, formulaic pieces with open slots; – fixed expressions of many kinds, which may or may not correspond to syntactic phrases in a traditional sense; – fragments, that is, partial utterances; – discourse markers, including conjunctions such as and, when; – intensifiers such as really, even (not necessarily distinct from the discourse markers) – lexical items (to the extent that they are separate from the above). The degree to which ordinary discourse is indebted to the recycling of commonly used forms is often underestimated. In a recent paper, Newmeyer (2010) offers the following sample of dialogue from a corpus: (3) A: hi B: hi so did you hear what the topic is A: yes it’s about terrorism right B: yeah

2

Cf. Widdowson (1989), for whom communication is “… a matter of knowing a stock of partially pre-assembled patterns, formulaic frameworks, and a kit of rules, so to speak, and being able to apply the rules to make whatever adjustments are necessary according to contextual demands.”

24

Paul J. Hopper

B: um A: so what are your feelings on that [laughter B: i have [laughter] i personally can’t imagine anyone staying calm [laughter] A: yeah nor can i yeah B: um you would even i- though if you’re panicked i would assume you would try and B: keep your head clear enough to act to protect yourself but A: right A: yeah i don’t know if there was an explosion or something i don’t it it’s a shock so i don’t know that anybody can really think about it and control themselves B: um B: right even with all the um B: (( [sigh] the )) B: the publicity and media coverage you know that’s been on that topic A: (( [mn] right )) B: twenty months it’s still um B: is something that you wouldn’t be

Newmeyer comments: “There are certainly formulaic expressions here: hi, right, take in stride, I don’t think and possibly a few others. But in other respects the transcript reveals a sophisticated knowledge of syntax that defies any meaningful analysis in terms of ‘fragments’.” An approximate assessment of the degree of formulaicity in Newmeyer’s example of putative syntactic knowledge can be obtained by the simple process of typing identifiable sequences into a search engine, in this case Google.com. The results are edifying:

Emergent Grammar and Temporality in Interactional Linguistics

25

Table I: Formulaic Expressions in Conversation (3) (Google.com) Input String

Number of Hits in Google.com

did you hear what what the topic is it’s about terrorism what are your feelings your feelings on that can’t imagine anyone staying calm nor can I you’re panicked I would assume you would keep your head clear clear enough to enough to try and to act to to protect yourself there was an explosion or something it’s a shock i don’t know that anybody can really think about it control themselves even with all the publicity media coverage on that topic in the last twenty months something that you wouldn’t be able to take in stride

195,000 488,000 5,240 339,000 5,350 387,000 275,000 1,210,000 2,090 25,300 9720 422,000 71,900 250,000 2.86 million 488,000 74.6 million 98,400 17,900 1,680 406,000 460 295,000 190,000 1,070 6.999 million 7.160 million 15,400

On the basis of such data one can say that there is some syntactic knowledge, but its complexity is not as great as it might seem. The utterance: though if you’re panicked I would assume you would try and keep your head clear enough to act to protect yourself can be represented as the interleaving of fragmentary phrases, with at the most some overlap between adjacent formulas:

26

Paul J. Hopper

though if you’re panicked I would assume you would try and keep your head clear clear enough to to act to to protect yourself

It does not seem that much more than this is needed. Informal dialogue from an on-line perspective consists not of sentences generated by rules, but of the linear on-line assembly of familiar fragments. Grammar is emergent and epiphenomenal to the ongoing creation of new combinations of forms in interactive encounters. As Roberts (1944: 299) put it, “Yesterday’s discourse is today’s language; today’s discourse is tomorrow’s language.”

3

Emerging and emergent

The term ‘emergent’ and its nominalization ‘emergence’ refer to the fact that a grammatical structure is always temporary and ephemeral (Hopper 1987, 1998). The provisionality of grammar follows directly from the decision to study spoken language in its natural setting, which is informal conversation. The term ‘structuration’ used by sociologists appears to be very close in meaning to ‘emergence’. It is associated with the work of Anthony Giddens, who in an early stage of his work noted that “to enquire into the structuration of social practices is to seek to explain how it comes about that structures are constituted through action and reciprocally how action is constituted structurally” (1976).3 Citing the work of a sociologist rather than a cognitive psychologist at this point underlines the social perspective on language that is at issue here. Ian Craib (1992) identifies two aspects of Giddens’ work. One is analogous to the construction toy Lego, static with interchangeable parts, but restricted. The other moment, structuration, gives a freedom to the moving parts, emphasizing the transformative aspects of human action (…). The moving parts are not simply rotating, albeit in different ways, but constantly making and remaking the model. None of the things I have discussed – structures, systems, and their elements, institutions, the aspects of interaction – exist in and for themselves, but only in and through social action and social practices. All these things are done they are not. The difficulty is that they have to be stopped and sorted out in order to make sense of them. The danger is that in sorting them out they take on aspects of objects that really exist in the

3

Cited in Bryant and Jary 1991, 7.

Emergent Grammar and Temporality in Interactional Linguistics

27

world like other objects. We lose … the advantage of structuration theory: the way in which it remains open to the openness of praxis and of historical change. (Craib 1992: 72)

3.1 Uses of the term ‘emergence’ in recent linguistics A number of linguists (e.g., Dahl 2004, Keller 1994, Traugott 2008, the papers in MacWhinney, ed., 1999) and others have also used the term ‘emergence’, but have understood it in the following way. A language has a set of fixed constructions. The historian’s and child language investigator’s task is to show how the presence of these constructions is due to a convergence of changes, changes that themselves may not be interconnected but whose effect is to result in useful linguistic forms. This is the “invisible hand” theory of Keller (1994) – the metaphor is taken from Adam Smith’s economic theory. It is a kind of historical conspiracy. Emergence is here treated as a question of where constructions come from. On this account, once they have arrived, they enter the synchronic grammatical system of the language. While this may give an account of the mechanism of change, Giddens’ warning is of great relevance. There is a danger that the emphasis in this understanding of emergence will be on the sources of a present-day, i.e., synchronic, state of affairs. The Invisible Hand theory, since it addresses only the historical origins of constructions, is thus a useful strategy for preserving the view of a language as a synchronic system où tout se tient. Its adherents acknowledge the principle that language is always changing, but do not carry through the consequences of this idea into current interactions. Moreover, the methodological privileging of long-range historical accounts, with their necessary reliance on written forms, risks a continuation of the scriptist bias into the present-day, and appears to legitimize what Love (1999) has called the Fixed Code theory of a language, according to which the role of grammar is precisely to ensure uniformity of usage among speakers. Speakers, it is believed, communicate with one another by virtue of a shared grammar that is present and similar for all speakers. I suggest that the very idea of a uniformly shared grammar stems from the “Written Language Bias” that Per Linell describes (see Linell 2005: 101–12), for only writing presents the model of a stable, uniform and complete system presupposed in the Fixed Code postulate. As a noun, ‘emergence’ conceals an ambiguity that is not present in the two adjectives that it conflates, emerging and emergent. By ‘emerging’ we are entitled to understand the development of a form out of its surroundings, its epigenesis. The term ‘emerging’ is thus appropriate for the view of grammar as

28

Paul J. Hopper

a stable system of rules and structures, which may ‘emerge’ (i.e., come into existence) out of a less uniform mix. The work of grammatical analysis along traditional lines is important, indeed indispensable, in providing a starting point for a more discourse-oriented perspective on structure. However, when the study of language is directed toward spoken conversational interactions, the relevant results of traditional linguistics are soon exhausted. It is understood that categories don’t exist in advance of the communicative setting. Instead, they are constantly being elaborated in and by communication itself. They are unfinished and indeterminate. It is in this sense that the term ‘emergent’ is used.4 Emergent Grammar focuses on the boundaries of categories rather than their prototypes, exploring the leading edges and the territory around them as they move (see Pekarek Doehler [this volume] for an empirical exploration of this position.) 3.2 Structuration The term ‘structuration’ used by Giddens and his French predecessors seems very close to ‘emergent’ as I am using it in the present paper (and also in Hopper 1987, 1988, 1998), and has some advantages over it, the chief being that the ‘emergent’ meaning of structuration, being warranted by its decades-long use among sociologists with this same meaning, is protected from loss or distortion. ‘Structuration’ is not susceptible of misinterpretation in the same way as ‘emergence’, which allows a confusion of ‘emergent’ with ‘emerging’. Also, ‘structuration’ retains a relationship with the word ‘structure,’ and is thus less likely to be seen as beyond the pale of linguistic projects. Moreover, the suffix -ation has long been associated with linguistic ideas about becoming, such as ‘grammaticalization,’ ‘morphologization,’ and so on. On the other hand, ‘structuration’ lacks a convenient adjectival form corresponding to ‘emergent’. For this reason it is convenient to preserve the term ‘emergent’ for some contexts. 4

It should be conceded that ‘emergent’ is often used in the sense of ‘emerging’, for example by some social and cultural theorists, as in ‘emergent working class’, ‘emergent literacy’, ‘emergent democracy,’ and so on. But this sense conveys the idea that the head noun will eventually appear as a recognisable entity or property, and is therefore indistinguishable from ‘emerging’. ‘Emergent’ when applied to linguistic structure (‘Emergent Grammar’) takes up the by now standard use of this term by cultural theorists such as E. P. Thompson, Stuart Hall, and Raymond Williams for whom culture “is not an object to be described, neither is it a unified corpus of symbols and meanings that can be definitively interpreted. Culture is contested, temporal, and emergent.” (Clifford and Marcus 1986: 19)

Emergent Grammar and Temporality in Interactional Linguistics

29

Obviously the transportation of such a rich term from one discipline to another will involve important differences in both detail and scope. A useful restatement of Giddens’ position in regard to linguistic structuration is supplied by Linell: … linguistic structures, cultural routines, norms, etc. do exist prior to interactions (but only in and through the interactants’ being acquainted with them). At the same time, however, these structures, routines, and norms are interactionally generated, traded down, and reconstructed. That is, they exist prior to individual interactions, yet would not exist without a living historical continuity of interactions. Social structures are (re)created, tried out, tested, negotiated and modified every time they are instantiated or drawn upon. Habit is modified by accommodation, while accommodation is enabled and constrained by habit (Linell 1998: 59–60).

Interactions are thus “dialogically related to a continuity of praxis” (Linell 1999: 60, emphasis as in original). ‘Continuity’ is crucial here. Structuration, like emergence, implies that there is no natural terminus, no complete current set of linguistic facts and no synchronically bounded entity. Rather, structures are constantly being modified and negotiated during use.5

4

Projection

A relevant question to ask is: what linguistic problems or puzzles might we expect our research to illuminate? Among these would be: – Fluency – The ability to rapidly assemble and deliver coherent utterances (Pawley and Syder 1983, 2000). – Irreversability – The temporal unidirectionality of the stream of speech. (Auer 2000) – Ellipsis – The ability to accurately recover omitted or incompletely heard words and phrases, and the related ability, – Redundancy – The ability to understand incomplete messages. – The ‘Darmok’ Effect – ‘understanding the words’ but unable to interpret the message.6

5

6

Thus the matching of actual utterances to previously available schemas is at best achieved only approximately. Schemas are in fact changed each time they are used. For some interesting discussion, see Barlow 2000. The term is taken from the Star Trek episode Darmok, in which the crew of the Enterprise must communicate with a race of beings speaking a language whose words, while recognizably English, have meanings that derive from their own mythological past.

30

Paul J. Hopper

– Synchronicity – The mutual timing of utterances in harmony with interlocutors (Auer 2000). A key concept for understanding these central features of spoken language is projection. The term has its origins in Conversation Analysis and referred quite narrowly to the anticipation of the end of a Turn Constructional Unit. More recently, however, it has been used in the more extensive sense of any kind of signaling of “more to come”. Auer (2005) defines projection as follows: “By projection I mean the fact that an individual action or part of it foreshadows another. In order to understand what is projected, interactants need some kind of knowledge about how actions (or action components) are typically (i.e., qua types) sequenced, i.e. how they follow each other in time.” Auer goes on later to say: “Human interaction rests on the possibility of projection; the grammars of human languages provide interlocutors with sedimentated and shared ways of organising projection in interaction.” I would suggest that the reverse formulation: the sedimented and shared ways of organizing projection and interaction comprise [a large part of] the grammar of a language provides a useful alternative way of viewing the same relationship of linguistic structure to action. In the present paper I elaborate on this observation from the perspective of English casual conversation.

5

The such a/an construction

The sequence such a(n) is pivotal in several distinct constructions, including the following two, which are here illustrated from the corpus: (1) (a) (H) .. we had such a nice day. (b) (H) And he was such a hit there, that Macy’s gave him his very own charge card.

Typically, though not always, the (a) type, consisting of a single turn construction unit, imposes an evaluative interpretation on some segment of the turn. In the Santa Barbara Corpus all examples of the (a) such a/an construction terminate in the full stop intonation symbolized in the transcription by a period, which licenses the interlocutor to begin a turn if she wishes to (a transition-relevance place). The (b) type of the such a/an construction is biclausal. Here, the sequence such a/an anticipates (projects) a second clause which usually begins with that. The meaning is resultative: the second clause follows as a result of the first clause. The first clause of the resultative construction has in all instances comma intonation, a signal that the turn is incomplete.

Emergent Grammar and Temporality in Interactional Linguistics

31

In the following excerpt,7 Doris and Angela are discussing a third person, a woman named Sam, who has previously taken no interest in retirement seminars: (2) DORIS:

ANGELA: DORIS:

... Sam has been, .. has taken such an interest in this retirement bit. .. (H) ... That it– .. it really surprises me. .. Well she’s begun to listen. .. Yes she has.

The “Retirement” dialogue serves to illustrate a common feature of the grammar of conversational discourse: speakers do not create utterances by matching them in advance of the utterance to an a priori schema, but rather improvise at each point as the discourse unfolds. This process of unfolding frequently takes place through a process of adding increments to an already realized utterance part (for recent discussion and analysis see Ono and Couper-Kuhlen 2007a, 2007b; Ford, Fox and Thompson 2002; Auer 1992, 2007). The dialogue between Angela and Doris displays an interaction between these two constructions that cannot be explained on the assumption that the speaker has a ready made construction to hand that she is implementing at an appropriate point in the interaction. It requires instead a perspective in which utterances unfold in time and are subject to renegotiation at any point. Doris’s statement: DORIS:

... Sam has been, .. has taken such an interest in this retirement bit.

is identified through its prosody (the full-stop intonation) and its syntactic closure as a type (a) such a/an construction, in which interest in this retirement bit is marked diacritically (through extra high pitch on the word INterest) as an evaluative utterance. Its falling intonation qualifies it to be taken by the interlocutor, Angela, as having reached a transition relevance place and thus inviting some response such as uh-uh or a comment. But Angela makes no immediate response at the transition relevance place; that is, she does not accept the implicit invitation to come in with her turn. This absence of uptake is reflected in the long pauses in lines 2 and 3 and the restart that it – it. It has been shown (Ford, Fox and Thompson 2002) to be a typical environment

7

The data for this article are taken from the Santa Barbara Corpus of Spoken American English (2000), University of California Santa Barbara Center for the Study of Discourse, directed by John W. Du Bois.

32

Paul J. Hopper

in which the speaker adds an increment to a completed turn. Doris reacts to Angela’s silence by converting her single-clause such an (the (a) construction) into a double clause construction such an … that (the (b) construction), exploiting the double potential in the sequence such an. It seems as though the projective value of such an manifested in Doris’s utterance was not planned at the time the utterance was initiated.8 Rather, viewed from an on-line perspective, such an had the value of an intensifier, and has taken such an interest in this retirement bit was first intended as a complete utterance. An a priori grammar analysis that edited out the evidence of on-line production and started out with a complete biclausal entity like Sam has taken such an interest in this retirement bit that it really surprises me would miss an important point: Doris is not following an a priori grammatical rule, but is rather creating her grammar as she goes. Speakers are free to reanalyse their utterances on the fly by adapting their discourse to the linguistic resources they are provided with. The leading edge of an utterance is always open to renegotiation through increments which under some circumstances may work to recreate an original grammatical project and (seen retroactively) transform it into a new construction.9 Moreover, this transformation takes place in dialogical circumstances. Doris’ continuation is done with the complicity of Angela, who by suspending her uptake and foregoing her turn at talk forces Doris to extend her turn. While Doris’ renegotiation of her construction is not verbally co-constructed, Angela’s physical and social presence is a prerequisite for it.

6

Structuration in Conversation

Grammar at the conversational level is not inscribed in a permanent form, but is the result of spontaneously creating new combinations of forms. These forms are fragments of previously heard expressions assembled on the fly (Widdowson 1989). They are improvised in much the same way that a jazz performance is improvised: publicly and jointly. The jazz musician’s per8

9

Auer 2005 describes an almost identical construction in German involving the use of so. See also Goodwin and Goodwin 1987, who similarly describe the momentby-moment unfolding of the construal of “assessments” in the course of an interaction. In the taxonomy of increments proposed in Ono and Couper-Kuhlen 2007b, Doris’s continuation would seem to be an “add-on”, since it is separated from its host by a prosodic break. It is not entirely clear whether instances like this one in which the increment retroactively creates a new construction is an uncomplicated case of an add-on, however, or a special case.

Emergent Grammar and Temporality in Interactional Linguistics

33

formance is based on extemporized variations of familiar themes, which may have been learned as part of the jazz musician’s training or culled from public music: Sinatra, the Blues, Beethoven. These variations are linked and synchronized with the performances of the other musicians in the group. They don’t preexist the occasion of their performance, but neither are they novel. Like the Singer of Tales described by Albert Lord (2000), speakers weave together previously heard utterances. And like the Singer, speakers need a reactive audience, one that ratifies their utterances, responds to them, and gives them permission to continue. In Bakhtin’s terms: When we select words in the process of constructing an utterance, we by no means always take them from the system of the language in their neutral, dictionary form. We usually take them from other utterances, and mainly from utterances that are kindred to ours in genre, that is, in theme, composition, and style. Consequently, we choose words according to their generic specifications (Bakhtin 1986: 87).

The “words” to which Bakhtin refers are phrases, expressions, fragments and so on that have accompanied past actions. These actions are “kindred to ours in genre,” that is, they were useful in contexts analogous to the present one. Consider the following example. Someone was recently overheard to say: “He just got thrown in with the wrong sort of people.” Structurally this short utterance consists of two formulas: get [etc.] thrown in with and the wrong sort of people. Several items in the utterance index textual genres. The typical context for it would be that the subject, he, otherwise from a good family, has been involved in some kind of unsavory, perhaps illegal, activities. But in putting it this way, the speaker simultaneously diminishes the scandal and distances herself from responsibility for it. The word just mitigates the subject’s character flaw, suggesting that his behavior has not been normal for him, but that his wrong-doing was more or less an isolated accident, and not really his fault. This sense is reinforced by the use of the passive, with its decreased emphasis on personal accountability for an event, and specifically the getpassive, which often connotes a negative outcome. Got thrown in with, which collocates strongly with a negative complement, elaborates on the idea that the subject is a victim rather than an agent, as if he was unaware of what was happening. The wrong sort of people implies that the subject’s companions were involved in activities that the speaker disapproved of. Again, the subject’s culpability is being attenuated, since “the wrong sort of people” could cover a wide range of such activities, from a criminal gang to a different political party. The example suggests how types of utterances are indexed to previous occasions of use, and are employed to classify and evoke current occasions by relating them to previous experiences.

34

Paul J. Hopper

Some, indeed many single utterances consist of nothing but holistic items, or formulas. In the following example, from the Santa Barbara Corpus, Patty is telling her friend Gail that she is being firm with her daughter regarding her studies: (4) GAIL: PATTY: GAIL: PATTY:

Well that’s good, just keep doing it baby, she’s learning. Right? just keep do- -She’s learning, I mean it. I mean business.

The short utterances are everyday English ones: that’s good, just keep doing it baby, she’s learning, I mean it, I mean business. Many exchanges are of this nature, that is, simple formulas with little or no internal complexity. Often, however, spoken discourse is built up of blends of these simple formulas. The next example concerns a construction that has been much discussed in linguistic publications, but which looks different when considered from the perspective of on-line, temporal, emergent grammar. The name given to this construction by J. R. Ross (1969) was “sluicing”. A sluice is a channel that diverts a small amount of the water of a larger river for forming a lock or an irrigation canal, so the term “sluicing” refers to the idea that a minor echo of a sentence has been created out of a longer sentence. A couple of constructed examples will illustrate it: (5) (a) Burglars forced open the door, but the police don’t know how (b) My ex-girlfriend has gone off to Cancun, and I don’t care who with. We can easily recognise here a biclausal construction, one that consists of two closely linked clauses. The second clause in this case, the “sluiced” one, is held to be an elliptical contraction of a longer sentence, as in: (6) (a) Burglars forced open the door, but the police don’t know how burglars forced open the door (b) My ex-girlfriend has gone off to Cancun, and I don’t care who my ex-girlfriend has gone off to Cancun with Let us consider the testimony of a conversational corpus. In the 46 conversations available to me I found several examples that would be relevant to a corpus study of sluicing. I should emphasize at the outset that almost none of the corpus examples

Emergent Grammar and Temporality in Interactional Linguistics

35

actually matches the templates offered by syntactic treatments of sluicing, which work with a monological decontextualized sentence of the kind that is very rare in spoken discourse. A fairly clear example is the following, (7)

.. she said okay, I’ll stay another year. .. But, .. I don’t .. know why.

(the two first-person pronouns have the same referent). In addition to this one, the following are the only exemplars in the corpus that conform more or less to the pattern required by the syntacticians’ account: (8)

he comes over there, and is talking with that woman. .. I don’t know about what,

(9)

(H) American democracy is dying, ... and I want you to try to think .. of why.

(10) and their ... dog’s dripping blood and they’re wondering why. (11) ... (TSK) we knew we were losing oil, we didn’t know where.

Obviously even here some generous allowances have to be made for the differences between spoken data like this and the sort of highly normativized and edited data represented in the syntactic literature, but these examples will suffice to show that ‘sluicing’ actually occurs in conversational English. In the syntactic literature, such sluicing sentences are generally accounted for by some variants of an ellipsis with raising of a wh-word to a complement position. The full versions of these utterances would be: (12) (a) He is talking to that woman, but I don’t know what he is talking to that woman about (b) Their dog is dripping blood and they are wondering why their dog is dripping blood (c) We knew we were losing oil, but we didn’t know where we were losing oil Explanations that rely on the restoration of missing abstract elements are especially problematic in examples like the following: (13) KRISTIN: It does make it easier, if you do eat less carbo. Cause you don’t have to take these big (H) amounts, and then wonder what they’re gonna do and when.

36

Paul J. Hopper

If we are to have a syntactic analysis of when in terms of sluicing here it would have to include: (14) you wonder [they are going to do SOMETHING] and [they are going to do (something?) AT SOME TIME] There are two reasons why we might be encouraged to look elsewhere for a better account. One is that we should be suspicious of all accounts that invoke ellipsis because the rhetorical function of an expanded ellipsis is quite different from that of the allegedly identical elliptical sentence. Consequently the material in the expanded ellipsis is meaningful and cannot be conceived of as deleted to form a reduced version. It could in fact be argued that in many cases the usually assumed relationship between an elliptical and a full version should be reversed – it is the elliptical utterance that is basic, and the supposed fuller version has a special pragmatic function. In this next example, which takes place in a veterinary clinic, the repetition of we could send her home tonight has a special pragmatic force that would be lacking if the speaker had merely “sluiced”: (15)

LINDSEY:

I= think, If we– .. she woke up, we could .. w- send her home tonight, ... I wouldn’t see why we couldn’t send her home tonight, would you? If she wakes up, ... fine.

The speaker might have said I wouldn’t see why we couldn’t or perhaps I wouldn’t see why not. Instead she repeats her remark, thereby converting it into an actual proposal presented to the other person for approval. The second reason for wanting a different account of sluicing than that offered by abstract syntacticians is a more general one. By making fewer assumptions about the nature of linguistic data we arrive at greater realism and economy. Syntactic accounts rest on the notion of a priori grammar, something that interactive accounts reject in favor of the emergent grammar perspective. When we move from intuitive data to conversational data, it becomes quite obvious that the ‘sluicing construction’ is a rare sedimentation of a basic feature of talking, that is, the incremental addition of ready-made expressions, in this case a wh-expression. Alina’s .. oh I forgot which one in:

Emergent Grammar and Temporality in Interactional Linguistics (16) ALINA:

37

They’re right off of Pico and, .. (H) I know. ... (Hx) . But it’s in- -Right next door is !Ted !Rich, who’s uh=, .. (H) one of the biggies at MTM, .. (TSK) or, Lorimar or MGM, .. oh I forgot which one.

is not to be regarded as an integrated part of the “sentence” Right next door is Ted Rich, who is one of the biggies at MTM, or Lorimar, or MGM, I forgot which one. Rather it is simply a formulaic extension, an increment. Lenore has commented on the lower class address of their acquaintance (“south of Sunset Boulevard”). In response Alina adopts a mock-elite posture, indexed by a supposed upper-class voice, observing that while it is held to be a low status address, their next door neighbor is a top executive at one of the big movie companies. Lenore follows Alina’s I forgot which one with a loud peal of laughter, and Alina joins in. The humorous implication is that Alina knows so many top executives of movie companies that she can’t remember one from the other. Her whole segment is imbued with a tone of heavy irony. It is clear that her oh I forgot which one is a spontaneous addendum serving to amplify the comic tone (and thereby mollify any implied criticism of Lenore’s snobbishness), rather than a part of any fully integrated biclausal sluicing construction. A similar analysis of a sluice as an increment applies to the following example, which has already been mentioned: (17) .. she said okay, I’ll stay another year. .. But, .. I don’t .. know why.

It seems reasonable here to suggest that why is elliptical for why I’ll stay another year. But when heard on line, there is a clear sense in which but, .. I don’t .. know why is an increment. The utterance I’ll stay another year is a distinctly separate TCU, with syntactic, prosodic, and pragmatic closure. The pauses separating but, together with the hesitant delivery of I don’t .. know why point to the “sluiced” clause as an added-on increment to I’ll stay another year. In the next example we see the sluiced expression functioning as an organizer of the discourse:

38

Paul J. Hopper

(18) TAMMY: BRAD: TAMMY:

.. I’m one of these weird people that has trouble with auto reverse. [(TSK) oh yeah]? [I don’t] know why. The first time– I mean, this is the first time I’ve had it, on my b=oombox. .. (H) And for some reason, ... maybe put the tape in, I never know which direction it’s gonna go.

Tammy, who is buying a new tape recorder, is explaining to Brad, the salesman, her problem with auto reverse. Once again, there is final intonation in the first clause (I’m one of these weird people that has trouble with auto reverse.), so Tammy’s I don’t know why is to be interpreted as an add-on. However, her I don’t know why does more than wind up her statement I’m one of these people … In fact, speakers often use a sluice here as a way of launching into a new discourse segment. I don’t know why is not a lame admission of ignorance, but a preface to an explanation, in this case of Tammy’s difficulty with auto-reverse. It is a familiar problem: one can never be sure which direction the tape is going, so Tammy is doing nothing more than associate herself (“I’m one of those people …”) with a general confusion about auto-reverse. Notice that Brad doesn’t come in after the TRP in I don’t know why and that Tammy doesn’t expect him to. I attribute the absence of uptake here to an expectation raised in this kind of context by I don’t know (wh-), which in some contexts conventionally licenses the speaker to continue. It would in fact be rude of Brad not to allow Tammy a chance to mitigate her profession of ignorance; it could even be heard as an interruption. I would therefore argue that in some circumstances I don’t know (wh-) can function as a projector of more-to-come. Such an analysis reveals that sluicing is not a finished completed construction but an emergent sequence, the second part of which – the “sluice” – consists of some such expression as I don’t know, which I’m provisionally calling an epistemic token, followed by a wh-word, which may occur solo or, as in the next example, with a ‘pied-piped’ preposition: (19) he comes over there, and is talking with that woman. .. I don’t know about what,

I’ll call the wh-word and its possible preposition the wh-piece. The sluice itself, then, is also bipartite. Standard linguistic analyses present the wh-piece as a reduced complement clause that is structurally dependent on an antecedent

Emergent Grammar and Temporality in Interactional Linguistics

39

clause containing the epistemic verb phrase. This epistemic clause is itself coordinated with a previous sentence that supplies the implicit content of the wh-piece. The evidence from interactional data points more toward a double formula in the “sluice”, whose component parts are themselves formulaic. They include: (1) the epistemic token, which in the corpus is overwhelmingly I don’t know. There’s also one instance of they’re wondering, and one of (I want you to) try to think of, and (2) the wh-piece, which includes which, about what, why (4 instances), when, how long, where, who. I will state my claim as follows: 1. The so-called ‘sluicing construction’ is a construction, but from a different point of view from that usually adopted. The construction is fluid and emergent and secondary to the on-line and incremental process of turnbuilding in which an utterance is followed by a further composite utterance consisting of an “epistemic token” and a “wh-piece.” There is no consistent structural dependency between the epistemic token and the wh-piece. Rather, their cooccurrence is conditioned solely by the implied relevance of the wh-piece to some segment of previous talk produced by any of the participants. In some cases the relevance will be close; for example, the main sentence and the sluice run together and can be thought of as a biclausal “sluice” construction. But whether or not this happens depends on the actions these turns are carrying out. The corpus data suggest that the sedimenting of a sluicing construction is a random and indeed rare event. 2. The clause and the wh-piece are formulaic. Overwhelmingly the verb in the epistemic token is ___don’t know. While no doubt a much larger corpus would yield further kinds, it is likely that in casual spoken English the possibilities are restricted enough to be listed. (They would include not sure, not certain, not figure out, and others.) The formula I don’t know itself occurs 237 times in the corpus, whose total length is 350,000 words. Both the epistemic token clause and the wh-piece are possible stand-alone utterances. Both occur so frequently in this way as scarcely to need exemplification: (20) I don’t know (a) MILES: ... What what– .. Who is this. JAMIE: ... I don’t know, it’s from the radio station years ago when I was ... in Santa Cruz. (b)

LENORE:

uh, who’s gonna take Hector’s place.

40

Paul J. Hopper ALINA:

.. It- -.. (Hx) I don’t know. They’ll have to see.

(21) wh-word expression (a) SHANE: ... You’re gonna pour it in the meat? JULIA: ... (COUGH) (COUGH) .. Y=eah. SHANE: ... What for. JULIA: ... So it won’t be so, s=o dry, (b)

MARIE: LISA: KEVIN:

(H) See I’m not supposed to give him juice in a bottle. ... He’s supposed to drink it out of a cup. .. Why not. ... Oh cause all the air?

Moreover, there are instances where the wh-words occur as syntactic appendages but where no analysis as sluicing seems necessary or appropriate: (22) PHIL:

BRAD:

[But .. to] also ... put into perspective, and see, w- [that was one] point of view why, [Yeah=].

It is clearly part of the grammar of wh-words that they can occur in numerous contexts where they can serve as loose increments or extenders. There is naturally a pragmatic dependency on an antecedent phrase: point of view implicates reason, explanation, etc. But this is generally true of words and expressions. If we regarded every such case as an ellipsis, we would have to postulate a virtually unlimited expansion of every utterance and utterance-part in a discourse. It follows that sluicing is achieved through on-line asssembly, often over several TCUs. Each of the two parts of the sluice can occur alone, and the sluice itself is only loosely attached to an antecedent utterance. This comes over especially clearly when we see the main clause and the sluice being coconstructed by different speakers, as in the following example: (23) LYNNE:

DORIS:

(H)Oh= yeah=. She= .. t- -.. (H) ... well they– she never did take shots. [Did she]. [No, I don’t] know why.

And the next one is a more complicated situation:

Emergent Grammar and Temporality in Interactional Linguistics (24) ALINA:

LENORE: ALINA:

41

And she’s like VOX>, (H) he said, I would have been here, but , she’s going, (H) . .. [I don’t know why=. [@@@@@@@@ @@]@@@ @@]

Here Alina is satirizing a young couple at a party. She does so by adopting the different voices – the woman being portrayed as girlish and self-centered and dependent on the older man. So this time the main clause is in the girlish voice and the comment, in this case the sluice, is in Alina’s own voice. It is, one might say, a case of co-construction by a single speaker. Dialogic examples like this underline the functional separability of the main clause and the sluice.

7

Conclusions

Sluicing was first named and analysed by Haj Ross in 1969. Since then it has continued to fascinate syntacticians. Yet there do not appear to be any corpus-based studies of the construction. The only linguists who have ventured to go beyond a single sentence and talk about a possible explanation for sluicing in terms of dyadic discourse are Culicover and Jackendoff (2005), who however (1) use only constructed examples, and (2) adhere to a variant of the “ellipsis” explanation, refining it only by arguing that the propositional content of the sluice comes from the antecedent sentence by way of what they call indirect licensing. But the student of spoken language will want to argue that all interaction proceeds naturally on the basis of “indirect licensing”. Without “indirect licensing”, coherent communication would be impossible. The relationship between any utterance and an “antecedent” utterance will therefore always be characterizable as “indirect licensing”, and so the value of this concept as an explicans for sluicing is zero. In this paper I have tried to show how a puzzling grammatical construction, sluicing, which has a 40-year history of analysis by syntacticians, is illuminated by a close examination of the interactional conditions for its use. Such an examination leads to the rejection of two premises of syntactic analysis. One is that we can conduct analyses on sentences as static, contextfree objects existing only in the mind of the individual. The other premise

42

Paul J. Hopper

is a corollary of the first, namely that sentences are timeless entities that are fully present and complete, whose beginning, middle and end can be surveyed simultaneously. Instead, we must view utterances as a form of behaviour that unfolds in time, produced by a speaker in reference to listeners whose ongoing ratification of the utterance as it develops is inseparable from the act of production. As utterances unfold, they structurate according to 1. the hierarchization introduced by turn-taking, and 2. prior texts. They do not conform to fixed prior grammatical forms. Instead they conform to norms, which means they more or less conform to recognizable patterns. As Pekarek Doehler has put it in another context, they are “the emergent products of partially routinized interactional practices” (Pekarek Doehler 2007: 6). Twentieth century linguistics and its current manifestations have been a story of the effacement of time, the representation of language in a deactivated form. It is no exaggeration to say that Interactional Linguistics prepares the way for a new paradigm in linguistics, one in which temporality is reinstated to its prior, privileged position.

References Auer, P. 1992 The neverending sentence: Rightward expansion in spoken language. In: M. Kontra & T. Váradi, (eds.), Studies in Spoken Language: English, German, Finno-Ugric, Budapest, 41–59. 2000 On-line Syntax – oder was es bedeuten könnte, die Zeitlichkeit der mündlichen Sprache ernst zu nehmen. Sprache und Literatur 85, 43–56. 2005 Projection in interaction and projection in grammar. Text 25, 1, 7–36. 2007 Why are increments such elusive objects? An afterthought. Pragmatics 17, 4, 647–658. 2006 Construction Grammar meets Conversation: Einige Überlegungen am Beispiel von so-Konstruktionen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 291–314. 2009 On-line syntax: thoughts on the temporality of spoken language. Language Sciences 31, 1–13 (English translation of Auer 2000). Bakhtin, M. M. 1986 The problem of speech genres. In: C. Emerson & M. Holquist (eds.), Speech Genres and Other Late Essays, Austin, 60–102. Barlow, M. 2000 Usage, blends, and grammar. In: M. Barlow & S. Kemmer (eds.), Usage Based Models of Language, Cambridge, 315–46. Bryant, C. G. A & D. Jary 1991 Introduction: Coming to Terms with Anthony Giddens. In: C. G. A. Bryant & D. Jary (eds.), Giddens’ Theory of Structuration: A Critical Appreciation, London, 1–31.

Emergent Grammar and Temporality in Interactional Linguistics

43

Clifford, J. & G. E. Marcus (eds.) 1986 Writing Culture: The Poetics and Politics of Ethnography, Berkeley. Craib, I. 1992 Anthony Giddens, London. Culicover, P. & R. Jackendoff 2005 Simpler Syntax, Oxford. Dahl, Ö. 2004 The Growth and Maintenance of Linguistic Complexity, Amsterdam/Philadelphia. Ford, C. A., B. Fox & S.A. Thompson 2002 Constituency and the grammar of increments. In: C. Ford, B. Fox, & S. Thompson (eds.), The Language of Turn and Sequence, Oxford, 14–38. Franck, D. 1985 Sentences and turns. In: H. Cuyckens (ed.), Dialogue: An Interdisciplinary Approach, Amsterdam, 233–45. Giddens, A. 1976 New Rules of Sociological Method: A Positive Critique of Interpretation, New York. Goodwin, C. & M. H. Goodwin 1987 Concurrent Operations on Talk: Notes on the Interactive Organization of Assessments. IPrA Papers in Pragmatics 1, 1, 1–52. Hopper, P. J. 1987 Emergent Grammar. Berkeley Linguistics Society 13, 139–157. 1988 Emergent Grammar and the A Priori Grammar Postulate. In: D. Tannen (ed.), Linguistics in Context, Georgetown University, 117–134. 1998 Emergent Grammar. In: M. Tomasello (ed.), The New Psychology of Language: Cognitive and Functional Approaches to Linguistic Structure, Englewood Cliffs, 155–175. 2008 Die Bedeutsamkeit der mündlichen Interaktion für die Linguistik: Die Pseudocleft-Konstruktion im Englischen. In: A. Stefanowitsch & K. Fischer (eds.) Von der Konstruktion zur Grammatik, Tübingen, 179–88. Keller, R. 2004 On Language Change: The Invisible Hand in Language, London. Lerner, G. 1991 On the syntax of sentence-in-progress. Language in Society 20, 3, 441–458. Lessing, G. E. 1766 Laokoon, oder Über die Grenzen der Malerei und Poesie. Mit beiläufigen Erläuterungen verschiedener Punkte der alten Kunstgeschichte, Berlin (Reprint München 1994). Linell, P. 1998 Approaching Dialogue: Talk, Interaction, and Contexts in Dialogical Perspectives, Amsterdam. 2005 The Written Language Bias in Linguistics, London. Lord, A. B. 2000 The Singer of Tales, 2nd edition, Cambridge. Love, N. 1999 The fixed code. In: G. Wolf & R. Harris (eds.), Integrational Linguistics: A First Reader Oxford, 49–64. MacWhinney, B. (ed.) 1999 The Emergence of Language, Mahwah, NJ.

44

Paul J. Hopper

Newmeyer, F. J. 2010 What conversational English tells us about the nature of grammar: A critique of Thompson’s analysis of object complements. In: K. Boye & E. Engberg-Pedersen (eds.), Usage and Structure: A Festschrift for Peter Harder. Berlin, 3–43. Ono, T. & E. Couper-Kuhlen 2007a Increments in cross-linguistic perspective: Introductory remarks. Pragmatics 17, 4, 505–512. 2007b “Incrementing” in conversation: A comparison of practices in English, German and Japanese. In: Pragmatics 17, 4, 513–552. Pawley, A. & F. H. Syder 1983 Two puzzles for linguistic theory: nativelike selection and nativelike fluency. In: J. C. Richards & R. W. Schmidt (eds.), Language and Communication, New York, 191–227. 2000 The one-clause-at-a-time hypothesis. In: H. Riggenbach (ed.), Perspectives on Fluency, Ann Arbor, 163–199. Pekarek Doehler, S. (this volume) Emergent grammar for all practical purposes: the on-line formating of left- and right dislocations in French conversation. Unpublished MS (2007). Subordination in grammar – subordination in action: projection sequences in French conversation, 46–88. Roberts, M. H. 1944 The science of idiom: A method of inquiry into the cognitive design of language. Publications of the Modern Language Association 59, 291–306. Ross, J. R. 1969 Guess who? In: R.I. Binnick, A. Davison, G. M. Green, & J. L. Morgan (eds.), Papers from the Fifth Regional Meeting of the Chicago Linguistic Society, Chicago, 252–286. Traugott, E. C. 2008 Grammatikalisierung und emergente Konstruktionen. In: A. Stefanowitsch & K. Fischer (eds.), Von der Konstruktion zur Grammatik, Tübingen, 5–32. (translated by A. Zeschel). MS (2008) “All that he endeavoured to prove was …”: On the emergence of grammatical constructions in dialogual and dialogic contexts. Widdowson, H.G. 1989 Knowledge of language and ability for use. Applied Linguistics 10, 128–37.

Emergent grammar for all practical purposes

45

Simona Pekarek Doehler

Emergent grammar for all practical purposes: the on-line formatting of left and right dislocations in French conversation*

“An utterance is a piece of behavior that unfolds in time” (Paul Hopper, FRIAS workshop, 2008). “Time in the form of sequential organization is a pervasive intrinsic component of both talk and action” (Charles Goodwin 2002: 24).

1

The temporality of language and the temporality of action

One of the fundamental properties of both social action (including talk-ininteraction) and language is that they unfold across time. Conversational openings and closings, repair, and disagreement, for instance, are configured on a moment-by-moment basis as talk evolves, so that in the course of their unfolding, their organization can be reoriented, and a sequence can be reopened, expanded or closed down. The same is true for the syntactic trajectory of utterances. This is most clearly manifest in the expandability of units of talk and their syntactic shapes (cf. Auer 1996). This expandability is configured incrementally in real time, allowing participants to prolong or revise syntactic trajectories, and thereby to accomplish various social actions (e.g. Couper-Kuhlen and Ono 2007; Ford, Fox, and Thompson 2002; Schegloff 1996). In this way, the temporality of language is indissociably linked to the temporality of action. One central underpinning of this inextricable embeddedness of language and action – of the moment-by-moment deployment of language along the moment-by-moment configuration of action – is that the structures of language are used as a resource for organizing and coordinating actions and are in turn shaped in response to this organization: They are made and put to work to accommodate local interactional needs. Goodwin (2002) empirically documents how “orientation toward diverse forms of time organization is built into the units and tools used to construct human action” (34), such as * I thank Peter Auer, Elwys De Stefani, and Anne-Sylvie Horlacher for their insightful and inspiring comments on a previous version of this paper.

46

Simona Pekarek Doehler

language, gesture and gaze. In an argument more centrally concerned with the nature of language, Auer (2005, 2007, 2009) defines the grammar of spoken language as an on-line grammar: Inscribed in the temporal unfolding of talk-in-interaction and the synchronization of mutual actions, syntax is a process in which constructions are configured in real time. In an earlier statement, Hopper (1992) points out the thoroughly temporal character of grammar, suggesting that “language owes the way it is to its temporal unfolding through […] spoken interaction” (236). This temporal character implies two empirically validated and theoretically consequential properties of linguistic constructions: projection and emergence. 1.1 Projection How language configures the temporal sequential unfolding of actions has been persuasively documented in empirical conversation analytic and interactional linguistic research on the notion of projection. In his papers on online grammar, Auer (2005, 2007, 2009) argues that projection is at the heart of the inscription of language in the temporal unfolding of actions (for a related argument see Goodwin 2002). Projection refers to the property of one segment of discourse (an action or part of an action, or a grammatical structure) to prefigure possible trajectories of the next (actional or grammatical) segment (cf. Auer 2005; Goodwin 2002; Schegloff 1996). In their seminal paper on the turn-taking machinery, Sacks, Schegloff, and Jefferson (1974) document the role of syntax for projecting transition relevance places (TRPs). The participants’ ability to recognize the syntactic (but also prosodic and pragmatic) trajectory of a turn represents the sine qua non of turn-taking: Minimization of gap and overlap is possible only due to the fact that participants can anticipate turn-ends before they actually occur – and they do this on the basis of projections emanating from the grammatical (syntactic, prosodic) and actional (pragmatic) dimensions of talk-in-interaction (Sacks, Schegloff, and Jefferson 1974). This is perhaps the most classic example of grammar serving as a resource for the organization of talk-ininteraction. The notion of projection has since attracted the attention of many scholars in conversation analysis and interactional linguistics as a basic organizational principle of talk-in-interaction. Action projection relates to the sequential organization of actions (a question, for instance, projects an answer as a relevant next), while grammatical projection relates to the sequential moment-by-moment deployment of linguistic units (a determiner, for instance, projects a noun as a relevant next; an if-clause projects a then-

Emergent grammar for all practical purposes

47

clause as a relevant next; level pitch projects more to come in the same turn construction unit [TCU]). Projection does not determine what follows, but foreshadows a range of possible upcoming trajectories. Hence, it offers the possibility of utterance co-construction (Lerner 1991, 1996; cf. section 3) and it is one – if not the central – resource for floor-holding and for the construction of complex, multi-unit turns (Hopper and Thompson 2008; Pekarek Doehler 2011). Also, projection implies that participants have some sense of how both actions and linguistic structures are organized sequentially (cf. Auer 2005) – that is, how they are deployed on a moment-bymoment basis in real time. Most importantly for our purpose here, some grammatical formats can be interpreted as emergent products of partially routinized interactional projection practices, as has been argued in recent work on “projector constructions” (Günthner 2006; Hopper and Thompson 2008; Pekarek Doehler 2011). The inextricability of the temporal unfolding of language and action condenses in these properties of projection. As Auer puts it, “the notion of on-line processing of grammar suggests that syntax is a formal(ized) way of human language to make projection in time possible” (Auer 2005: 14). 1.2 Emergent grammar A second consequence of the thoroughly temporal character of language and action (or more precisely: of language-as-inscribed-in-action) bears on the very nature of linguistic constructions.1 As language is a central tool for the coordination of the temporal and sequential unfolding of actions, its structures cannot but be continually adapted to the contingencies of social (inter)actions. This point has been documented in Goodwin’s (1979) classic analysis of how the construction of a single sentence is formatted in real time in response to local contingencies such as recipient actions or the absence thereof. It has also been demonstrated in Ford and Thompson’s (1996) and Selting’s (2001) analyses of TCUs as emergent entities. This work empirically documents the fundamental contingency of both linguistic and interactional units. Linguistic constructions, then, are not merely prefabricated, static resources for actions; rather, they are continually adapted in the very course 1

I will use the term “construction” in line with usage-based approaches to grammar to refer to patterns of language use of various size (e.g. NP, clause, clausecombination) comprising multiple linguistic items and whose meaning or function cannot be derived from the sum of their constituents. Constructions are (more or less) sedimented patterns for accomplishing communicative functions/ actions.

48

Simona Pekarek Doehler

of their production in response to locally emergent interactional needs. The grammatical constructions that participants use for the collaborative organization of talk in real time are hence adaptative, flexible, and contingent. They involve (partially) sedimented constructional schemata, but these are also subject to in-time emergence (cf. Ono and Thompson 1995). This point is at the core of Hopper’s notion of emergent grammar (e.g. 1987, 1992, 2001, 2004). Grammar, Hopper argues, is not a fixed code, a static set of structures and combination rules enclosed as abstract representations in the individual’s mind. Rather, it is the ever evolving inventory of constructions for discourse that are (partially) sedimented through repeated use: “‘Grammar’ is an epiphenomenon of frequent combinations of constructions. Because grammar is a result of interactions rather than a prerequisite to them, it is not a fixed code but is caught up in a continual process of local adaptation (emergence)” (Hopper 2004: 153). A central part of this emergent nature of grammar, as Hopper demonstrates (e.g. Hopper 2001, 2004), is the “openness” of the forms (or: constructions) that make up grammar: Any grammatical construction is in principle “open”, i.e. it materializes different yet related constructional schemata that may be only partially sedimented. This point has empirically been corroborated by an important body of research investigating how grammar is deployed in talk-in-interaction. Several studies have questioned classic conceptions of canonical grammatical patterns, such as pseudoclefts (for English, see Hopper 2001, 2004; Hopper and Thompson 2008; for French, see Müller 2006) or extrapositions (Couper-Kuhlen and Thompson 2006, for English). Others have destabilized well-established categorical boundaries between constructions that have traditionally been treated as strictly distinct (see e.g. Pekarek Doehler and Müller 2006 for pseudoclefts and left-dislocations in French). Yet others have deconstructed widespread conceptions of basic patterns of clause-combining, revising established notions of subordination (Thompson 2002; Matthiessen and Thompson 1988) or hypotaxis and parataxis (Auer 1998). The quoted body of research provides empirically robust counter-evidence to classic fixed-code conceptions of grammar. The emergent character of grammar is a crucial correlate of the inscription of language in action; it is a fundamental trait of grammar’s nature as “a result of interactions rather than a prerequisite to them” (Hopper 2004: 153). The openness of grammatical constructions results from the fact that participants use these structures as local solutions to interactional contingencies (cf. Ford 2004) that emerge in real time. Mutatis mutandis, it is due to this plasticity that language can and does serve as a “shared matrix” (Ford 2004: 31) for the management of interactional contingencies.

Emergent grammar for all practical purposes

49

1.3 Object and purpose of this paper In this paper, I wish to explore what this conception of grammar implies for our understanding of a given construction, and what type of empirical evidence is apt to corroborate such a conception. I will examine participants’ use of so-called left and right “dislocated” constructions in French conversation. My aim is to demonstrate how participants use left and right dislocations as partially sedimented constructional schemata in a contingent, adaptative way, so that a given grammatical format, once initiated, can be reconfigured moment-by-moment to yield another format as a practical solution to some locally occasioned interactional need. The analysis will concentrate on a series of phenomena that have been documented in research as testifying to the processual and contingent nature of grammar, such as collaborative utterance constructions, increments, and pivots. But rather than analyzing how completions, increments, and pivots materialize in different grammatical shapes, I will examine them through the lens of two constructions: left and right dislocation. This focus is designed to highlight the fact that even “classic” constructions, i.e. constructions that are considered to be highly grammaticized, are molded in real time to accommodate locally emergent interactional needs. This is an important point for my purpose here. Rather than tracking the occurrence of new construction formats, this paper sets out to explore how, once a construction is initiated, its concrete trajectory is shaped moment-tomoment for all practical purposes, to a point such that it can be expanded or revised to yield another construction. The analysis will show that this process of recalibration, whether it yields a “canonical” or “less canonical” grammatical format, represents a practical solution for dealing with recurrent interactional contingencies. The findings suggest that the “continual process of local adaptation” (Hopper 2004: 153) is in no sense limited to the “newness” of grammatical shape, but reflects the omnipresence of the dynamic, processual features of grammar. This, I hope, will contribute to a larger body of current empirical work documenting what the established objects of an a priori grammar become as part of an emergent, on-line grammar (cf. section 1.2). In the following, I will first specify the grammatical constructions under analysis and propose a critical comment on terminology (section 2). After a brief presentation of the data (section 3), I will analyze the emergent character of dislocated constructions. The analysis will show how the syntactic trajectories of dislocated constructions are configured on-line, how they are distributed across speakers and spread out across several interactional mo-

50

Simona Pekarek Doehler

ments (section 4), how they are expanded (section 5) or revised (section 6) in the very course of their production. The analysis will also demonstrate that these on-line adaptations are grammatical practices by means of which speakers get some interactional business done. On the basis of this empirical evidence, I conclude that the patterns documented in this paper represent recurrent grammatical formats2 that respond to locally occasioned yet recurrent interactional contingencies (section 7). As such, both left and right dislocations are part of an emergent grammar for all practical purposes.

2

Left and right dislocation

2.1 A critical word on terminology Before turning to a detailed presentation of the constructions under analysis, a critical word on terminology is in order. In this paper, I use the terms left dislocation (LD) and right dislocation (RD) for the sake of clarity, in accordance with the dominant literature concerned with these constructions. I believe, however, that the terms themselves are utterly misleading. They have their roots in a generativist tradition, which understands dislocated constructions as resulting from transformations applied to a basic clause structure, namely SVO for languages such as French or English (e.g. Ross 1967). This epistemological embeddedness of the notions of LD and RD has had profound repercussions on the way these constructions have been conceptualized in the literature – even far beyond the generativist tradition: LD and RD are typically understood as “marked constructions”, measured against the so-called canonical word order. I argue that such a view is both pragmatically and cognitively implausible. Most importantly for our purpose here, it disregards the fundamental moment-by-moment temporal unfolding of talk, and hence its sequential character. In this regard, it is noteworthy that LD and RD have typically been treated in the literature as being tightly related, both formally and functionally (cf. section 2.2). However, LD and RD differ drastically in how they shape utterances on-line – that is, how they configure the temporal grammatical unfolding of talk, including the projections emanating from such talk. LDs (along with topicalizations and hanging topics) are resources that allow participants to display TCU-beginnings in specific ways, while RDs 2

In recent work in interactional linguistics, formats are defined as “instruments for contingently building turns at talk and implementing actions” (Thompson and Couper-Kuhlen 2005: 483; see also Thompson 2002).

Emergent grammar for all practical purposes

51

do the same to TCU-ends (Pekarek Doehler, De Stefani, and Horlacher 2011). This is significant insofar as TCU-beginnings and TCU-ends are interactionally sensitive places for doing very different things: The former are particularly relevant places for configuring projections (Auer 2005; Lerner 1991, 1996; Schelgoff 1996), dealing with turn-taking issues (Sacks, Schegloff, and Jefferson 1974), displaying on-topic talk (Jefferson 1978), or managing the preference for agreement (Pomerantz 1984). TCU-ends, on the other hand, are particularly sensitive places for marking transition relevance places (Sack, Schegloff, and Jefferson 1974), dealing with issues of recipiency, and they can lead into different kinds of turn-extension (CouperKuhlen and Ono 2007; Ford, Fox, and Thompson 2002; Schegloff 1996). In the course of this paper, we will see that these properties of TCU-beginnings and TCU-ends are crucially relevant to how participants treat LDs and RDs on-line and what interactional tasks these constructions accomplish. 2.2 Forms and functions of left and right dislocation in French LD and RD are much more frequent in French than in English.3 Hence, they often do not translate into English. Excerpt (1) provides a first illustration of dislocated constructions in French. Line 1 shows an LD of the NP ma mère, and line 4 shows an RD of the NP votre mère: (1) FNRS C, l. 159–163 “ma mère” 1 Jul 2

ma mère elle arrive pas à me parler en allemand. my motheri shei succeeds not PREP to me speak in German ‘my mother (she) can’t manage to speak in German to me’

3 Mar

mhm ‘mhm’

4 Int

elle est germanophone votre mère? shei is German-speaking your motheri ‘is your mother German-speaking’

3

Lambrecht (1987) argues that SVO is far from being the basic word order in spoken French. Rather, “the vast majority of nouns appear neither in object nor in subject position but in prepositional and adverbial phrases, in extra-clausal topic phrases and in phrases that have no syntactic connection with the proposition at all” (219). He thereby pinpoints the comparatively high frequency not only of dislocations or topicalizations, but also of clefts, presentatives, and hanging topics in French. Lambrecht relates this to typological reasons: French uses these constructions, while other languages use word-order variations or accentuation, for marking e.g. topics of foci.

52

Simona Pekarek Doehler

In his detailed discussion of dislocated constructions across several languages, including French, Lambrecht (2001) provides the following definition: “A dislocation construction (also called detachment construction) is a sentence structure in which a referential constituent which could function as an argument or adjunct within a predicate-argument structure occurs instead outside the boundaries of the clause containing the predicate, either on its left (left dislocations) or on its right (right dislocations)” (1050). This resonates with the common understanding of a dislocated construction as a sentence structure in which a referential element (most often an NP, in example 1: ma mère, votre mère) is located to the left or right of a matrix clause containing a pronoun (elle, in example 1) that is co-referential with that element (for French, see e.g. Barnes 1985 and Blasco-Dulbecco 1999). In the above quote, Lambrecht (2001) specifies that the pronoun does not necessarily need to be co-referential, but can be co-indexical (e.g. in the case of associative anaphora/cataphora). In French, the pronoun is a clitic, while in English it is a free morpheme (cf. Givón 1983). The extra-clausal element can cover a range of grammatical functions and syntactic categories (see Lambrecht 2001 for a detailed discussion). By far the most recurrent cases documented in the literature (cf. Ashby 1988) as well as in the data analyzed here are detached pronominal or lexical subjects (as illustrated in example 1, line 1), followed by objects – direct (example 1, line 3) or indirect. The NPs are typically referentially definite (that is, they are definite NPs, but can be indefinite NPs in the case of generic reference, cf. Givón 1983). The data suggest that the prosodic properties of LD and RD are highly sensitive to their sequential environment (but see Barnes 1985, for French; Selting 2005, for German; Geluykens 1992, for English – who identify typical prosodic profiles, but whose results contradict each other). This has been exemplified by De Stefani (2007) who documents that LDs in specific sequential locations, where they are involved in the closing of episodes or topics, show rhythmic profiles that enhance their closing effect. As to their discourse functions, both LD and RD are said in the dominant discourse-functionalist literature to be used for topic promotion: They serve to promote an accessible yet non-active referent (i.e. assumed by the speaker not to be in the current cognitive state of attention of the interlocutor) to the status of topic (cf. Ashby 1988; Chafe 1976; Givón 1983; Lambrecht 1987). RD, however, is generally considered to presuppose a higher state of activation of the topic element than LD (cf. Givón 1983). LD is also used for establishing contrast (Geluykens 1992), as is RD (Ashby 1988). Finally, RD is often associated with a repair function (“afterthought”, Chafe 1976; see also Geluykens 1994).

Emergent grammar for all practical purposes

53

While much of the work on dislocations is based on the study of monologic data, topic-promotion functions have been attested for conversational data as well (Ashby 1988; Geluykens 1992; Horlacher and Müller 2005). However, there are a small number of studies of talk-in-interaction which document across several languages that dislocations do much more than organize information structure: They are used by participants as a resource for organizing actions, and for making that organization mutually recognizable. LDs serve as turn-entry devices (Duranti and Ochs 1979; Mondada 1995; Pekarek Doehler 2001, 2004); they are a frequent format for definition requests (De Stefani 2005) and the construction of lists (Geluykens 1992; Pekarek Doehler and Müller 2006; see also Barnes’ 1985 classic study). Also, they participate in the sequential organization of actions (including preference organization) and in the mutual positioning of participants (Pekarek Doehler 2001, 2004). RDs, in turn, present a privileged format for evaluative statements (Horlacher and Müller 2005) and are used to deal with issues of recipiency, most typically calling for a display of co-participants’ agreement (Horlacher 2007; cf. section 5). We will return to some of these interactional dimensions of LD and RD in the analysis section.

3

Data and analysis

The present study is part of a larger research project4 investigating the interactional functioning of what have traditionally been called “topic” and “focus” constructions – dislocations, clefts, presentatives, etc. (cf. Pekarek Doehler, De Stefani, and Horlacher, in preparation). The database for the project consists of children’s interactions in French (their first language), interactions with language-impaired children and, most centrally, approximatively 15 hours of French conversational interviews among adult native speakers. Most of the data was collected in the French-speaking part of Switzerland. This paper draws from the 15 hours of conversational interviews, along with some data from everyday conversations, classroom interactions, and media debates. The data were transcribed following the Jeffersonian 4

The project, entitled “Topic and focus constructions as interactional resources. A grammar-in-interaction account”, has been generously supported by the Swiss National Science Foundation for the periods 2003–2007 (no. PP001–68685) and 2007–2009 (no. FN 100012–117938/1) http://www2.unine.ch/cms/pid/10617. html. The reflections presented in this paper have greatly profited from many discussions with the members of the research team: Elwys De Stefani, Anne-Sylvie Horlacher, Stéphane Jullien, and Gabriele Müller, who also contributed substantially to establishing the transcriptions of the main body of data used for this study.

54

Simona Pekarek Doehler

transcription conventions (see annex). Prosody has been noted intuitively (i.e. through listening) and, where necessary, checked against a prosodic interpretation generated through Praat. The dislocated constructions are highlighted in bold in the quoted excerpts. For the sake of clarity, and due to their frequency, the analysis will concentrate on the “dislocation” of lexical and pronominal NPs. Also, the focus will explicitly be on cases that highlight the on-line deployment of the constructions, without addressing the more classic occurrences of dislocations in the data.

4

Distributed syntax

Within the conceptual framework adopted in this paper (cf. section 1), constructions are not seen as the mere product of the exteriorization of representations stored in a single speaker’s mind. Rather, constructions are shared adaptative resources for action. An empirically strong case in point for this view is the fact that constructions can be distributed between speakers and spread out across two or three interactional moments. For instance, they can be collaboratively established on the basis of utterance co-construction (section 4.1). Also, in their course of production, speakers can design the initial part of a construction so as to invite recipient reaction before proffering the subsequent part(s) (section 4.2). Finally, constructions can be adapted on-line as part of the recipient design of complex turns (section 4.3). These recurrent features of dislocated constructions provide evidence for what we might call “distributed syntax”. 4.1 Co-constructed left dislocations Due to the property of projection and its recognizability for co-participants, an emerging utterance opens up the possibility of completion by another speaker. This has been demonstrated in Lerner’s (1991) classic discussion of the “syntax of sentence-in-progress” (see also Lerner 1996), showing how, in compound TCUs such as if-then sentences, the preliminary component (“if X”) projects the format for the final component (“then Y”), and hence enables the possibility of a second speaker producing the final component. As projections can emanate from any level of grammar, co-construction too can occur at different levels (Lerner 1991 provides an example of a co-constructed spelling of a name). In our data, LDs appear as one recurrent object of utterance co-construction. Here, the left-peripheral element provides a preliminary component, projecting the occurrence of a final component. This is almost exclusively the case with lexical NPs, which are most typically

55

Emergent grammar for all practical purposes

taken up by a co-indexical subject clitic in the subsequent clause. An initial illustration is provided in example (2). (2) CODI sec II SPD 22 “le diable” 1 A

(alors que) le (.) diable (.) eh: c’[est: iti’s’ ‘whereas the devili

2 B

[c’est ‘iti’s

3

tout ce qui est mauvais. all that is bad’

4 A

oui ‘yes’

The first speaker, A, produces a simple lexical NP and then hesitates; at that very moment the second speaker, B, produces a sequence of talk that can be interpreted (and indeed is interpreted by A, line 3) as completing the initial component to yield a co-constructed utterance. Although the first speaker has already started to pronounce c’est ‘it’s’ (line 1), the second speaker’s onset is almost simultaneous; as a consequence, it appears highly unlikely that the second speaker’s clitic c’ (‘it’ line 2) is mapped onto the first speaker’s c’ (line 1). Rather, the second speaker’s turn can be read as displaying a finetuned fitting of a final component (c’est tout ce qui est mauvais ‘it’s all that’s bad’) onto a preliminary component (le diable ‘the devil’) produced by the first speaker. The excerpt bears the typical traits of utterance co-construction as defined by Lerner (1991, 1996): – syntactic break-off or hesitations by speaker A (the latter being the case here, line 1); – anticipatory completion by speaker B (line 2), which is designed as a syntactic continuation of speaker A’s utterance; – speaker A’s display of acceptance or refusal (the former being the case here, line 3). The result is a collaborative establishment of a dislocated construction, where the preliminary component consists of a lexical NP produced by speaker A, and the final component consists of a matrix clause, produced by speaker B, comprising a clitic pronoun that is co-indexical with the NP. Another illustration is provided in example (3), showing a complex extraclausal constituent: (3) FNRS A, 1942 “le seul mot” 1 Ral

le- le seul mot que je ‘the the only word that I

comprends pas do not understand’

56 2 3 Bri 4 Ral

Simona Pekarek Doehler eh:[:h

>il ‘it

est is

il]= it’

[il est en allemand?] ‘it is in German?’ =est< (mais oui.) il est germanique. ‘is well yes it is Germanic’

In line 1, Ralph initiates a turn by means of le seul mot ‘the only word’ plus a restrictive relative clause, followed by slight hesitation. The initial component here again shows a [def. NP + hesitation]-pattern. This time it consists of a complex NP, which projects not only a limited range of possible syntactic follow-ups (cf. infra), but also the possible pragmatic-praxeological nature of the follow-up: What is expected next is a specification of the initial complex NP ‘the only word that I do not understand’. This property brings this LD interestingly near to the pragmatics of pseudo-cleft constructions (see Pekarek Doehler and Müller 2006 for a discussion of the fuzzy boundaries between LDs and pseudo-clefts in French). The specification is then provided by Brigitte (line 3) by means of a completion that is both syntactically and pragmatically fitted to the initial complex component. Brigitte takes up the initial complex NP by means of the subject clitic il (here: ‘it’) just a micro-second earlier than Ralph himself. Her completion can be read as a guess, providing one possible specification of what le seul mot que [Ralph] comprend[s] pas (‘the only word that [Ralph] cannot understand’) is. This interpretation is corroborated by the rising intonation on the end of Brigitte’s turn, suggesting that she is calling for confirmation. The confirmation is then provided by Ralph’s mais oui. il est germanique (‘well yes it is Germanic’, line 4). Note that Ralph’s own turn is competing with Brigitte’s completion, as evidenced by his accelerated tempo and the repetitions in lines 2–4, clearly showing his attempt to maintain the floor. The fine-tuned syntactic and temporal fitting of the final component onto the TCU in progress shows how participants monitor the on-line unfolding of syntactic patterns. As Lerner (1991) puts it: “Because a second speaker can produce an instance of the final component and initiate it at a place it could be due, it suggests the sequential availability of these features from an inspection of the utterance-in-progress” (445). Interestingly, the projection emanating from an NP, whether simple or complex, does not fully specify the possible format of the subsequent component. While an ifclause, for instance, normatively projects a then-clause to follow (although the “then” may be omitted), a TCU initial-simple or complex NP normatively projects several possible follow-ups (at least in languages such as English and French): a VP (yielding an SVO), a full clause (yielding an LD

57

Emergent grammar for all practical purposes

or hanging topic) or a S+V+Ø combination (yielding a topicalization construction).5 This is significant as it clearly shows that grammatical projection and its interpretation are highly context sensitive, depending heavily on the local sequential environment. In examples (2) and (3) the participants display joint orientation to the relevance of a dislocated construction within this specific moment of talk-in-interaction, despite the fact that other follow-ups would be grammatically possible. Even when the speaker has formulated his or her utterance beyond what can be interpreted as the preliminary component, other-completion is possible. This provides further evidence that participants analyze syntactic structures during their course of production. In the course of such lateplaced anticipatory completion, the other-completion typically recycles the subsequent component from its start (Lerner 1996). Examples are provided in (4) and (5). As opposed to examples (2) and (3), the second speaker’s completion is initiated here at a sequential moment and within a time span that clearly allows him to map the onset of his completion onto the first speaker’s start of the final component. In particular, it allows the second speaker to “copy” the subject clitic from the first speaker’s turn and use it as a starting point for his own completion. (4) FNRS, A, 330 “la notion de perfection” 1 Mar

donc ‘so

pour moi la notion de [per]fection ça:: (.) ((smack)) for me the notion of perfection it’

2 Int

[°eh°] °e-° (.)

3 Mar 4 Int

ça [veut rien dire. ‘it doesn’t mean anything’ [V:IA col ventito.= ‘gone with the wind’

=(ah) si ((laughter)) ‘oh yes’

(5) FNRS, D, 2265 “les gens” [talking about plurilingualism and code-switching] 1 Hel

quand on parle une autre lan- .hh ‘when you speak another lan-’

2

là on: (1.2) on- on mélange tout. on fait un- (.) et‘there you you you mix up everything you make aand’

5

Of course, other follow-ups are possible and, depending on the sequential environment, can even be expected. For instance, depending on its concrete sequential embeddedness, the NP alone can accomplish an action, such as providing an answer. In such a case, no grammatical follow-up is projected.

58

Simona Pekarek Doehler

3

et les gens ils ‘and DET people they

4

(1.0)

s’y- (r-) enfinwell’

5 Ren ils ont plus de [grammaire. ‘they have no more PARTIT grammar’ 6 Hel

7

[il me semble qu’il y a ‘it seems to me that there plus d’identité. is no more identity’

In both excerpts, the first speaker produces a lexical NP plus a clitic pronoun (and more, example 5) that can be read as co-indexical with the preceding NP. The construction-in-progress, which is broken off in both cases by speaker 1, can thus be unambiguously read as an LD construction – and is in fact read as such by speaker 2 in both excerpts. However, the second speaker fits his contribution not to the very end of the first speaker’s unfinished utterance, but rather recycles the initial elements of the matrix clause following the left-peripheral constituent. By means of these co-constructions of the left-dislocated grammatical format, a range of social actions are being implemented. For instance, in example (2) the second speaker visibly collaborates in explaining the meaning of le diable. The completions in (3), (4) and (5), in turn, can be read as anticipatory displays of understanding or as guessing what the first speaker is about to say. This interpretation is supported by the fact that, in (4), the first speaker is proffering some kind of conclusive comment on what he has just elaborated on through a lengthy stretch of talk, which is then confirmed by the second speaker’s completion. In (5), by contrast, the completion (line 5) appears to proffer the second speaker’s own interpretation of what people do when they switch from one language to the other, which is then refused by the first speaker, who proposes an alternative interpretation (line 6). In sum, the quoted excerpts show three points: First, LDs represent a recurrent object of utterance co-construction, possibly due to syntactic and praxeological projections emanating from the TCU-initial constituent, together with its sequential embeddedness. In these cases, the left-dislocated format itself results from a co-construction process that is spread across two adjacent speaker contributions. The resulting syntactic construction, which we call LD, is hence distributed across two speakers; it is a joint product. Interestingly, there is no occurrence of co-constructed RD in our data. This may be related to the different ways in which LD and RD shape utterances along the temporal unfolding of talk (cf. section 2.1): The left-peripheral constituent in LD incorporates syntactic (and other) projection, which is not

59

Emergent grammar for all practical purposes

the case for the right-peripheral constituent in RD, nor for the preceding matrix-clause taken as a whole. Second, this co-construction is based on a minute synchronization of mutual actions. As Auer (2005: 14) argues in his paper on syntax as process, such synchronization is only possible because participants closely monitor emergent grammatical structures; they jointly orient toward a syntactic trajectory that is configured moment-by-moment, across the temporal unfolding of talk. In the cases quoted here, the second speaker in particular exhibits his or her orientation to the specific compositional scheme of a left-dislocated construction, treating the left-peripheral element as a preliminary component, and the following matrix clause as a final component. The recurrence of these features suggests that participants orient in several regards similarly to LDs as they do to the compound TCUs discussed by Lerner (1991, 1996). Third, a range of interactional business is being accomplished by means of the final component of the co-constructed LDs: providing help, displaying knowledge or involvement, enacting alignment or disalignment, etc. This shows one use of grammar as a resource for organizing action. 4.2 The left-periphery as a try-marker A second type of evidence for LD as emanating from distributed syntax can be seen in examples (6) and (7), where the left peripheral constituent is marked as a try. (6) FNRS F, 912 “l’acqua” 1 Xav

mais (..) on connaît jamais un mot (..) qu’est-ce qu’ ‘but we never know a word what’

2

il veut dire vraiment. ‘it means really’

3 (0.6) 4

du genre^euh: (.) l’acqua? ‘like l’acqua’

5 Mar

ouais ‘yeah’

6 Xav

euh: enfin, (.) >on ‘well one

7 Mar

ouais ‘yeah’

sait que c’est knows that it’s

de l’ eauet ça ça
0.2 sec.) Precedent: addressee’s turn

25,8 % (8/31) 41,9 % (13/31) 25,8 % (8/31)

Relation to following talk: Turn-continuation Use as a turn-preface (prosodically integrated with ensuing TCU) Parenthetical insertion into ongoing turn Next event: pause (> 0.2 Sek.) Next event: response

93,5 % (29/31) 70 % (21/30) 6,5 % (2/31) 12,9 % (4/31) 25,8 % (8/31)

a speaker’s multi-unit turn (see Houtkoop and Mazeland 1986). More precisely, it seems to have a projective function regarding a topically coherent continuation of the ongoing turn. 4.3 Conversation analytic findings The figures in table 2 show how verstehst du? is used, but they do not explain how these uses are motivated. Detailed sequential analysis of single cases is needed to determine precisely what the sequential conditions for using verstehst du? are, how it is coordinated with the partner’s activities, which turntypes precede and which follow verstehst du?, and what function it has for the production of a multiunit turn. From a conversation analysic perspective as well as from a constructionist point of view, we want to know whether the uses of verstehst du? are really just one coherent practice. This can only be the case if we can show that the deviant cases, which do not match the most frequent distributional pattern, are nevertheless produced according to the same general orientation, but employed in a context-sensitive manner. From a distributional point of view, deviant cases need to be accounted for as cases which can be explained by the need to adapt talk to specific sequential affordances in order to comply with the general function of verstehst du?.

Constructions vs. lexical items as sources of complex meanings

97

4.3.1 Verstehst du? as an index of relevance and insistence A primary use of verstehst du? is this: The speaker signals that s/he has made a relevant point which needs to be acknowledged and accepted by the addressee, but from the speaker’s point of view, the addressee has not yet displayed his/her understanding and acceptance of this point clearly enough.10 Lack of uptake is most evidently at issue, when verstehst du? is used turninitially in an insisting response following a partner’s turn. This is the case in the following extract from a seminar at a film school in which two professors and four students discuss a script for a film. The professor had criticized the opinion of one student, who claimed a pickpocket would not be prestigious enough as a good protagonist. The professor argues that a good plot does not depend on a glamorous hero, but rather that there is a lot at stake for the protagonist. When one of the students confirms this, the professor insistently repeats his objection. 1 Pitching: Taschendieb_00:28:01–00:29:09 01 HA: 02 03 04 05 06 07

08 09 10 KA: 11 12 HA:

10

es kommt NUR darauf AN, ((clears throat)) welche beDEUtung; ‘It only matters ((clears throat)) which importance’ (1.2) das ZIEL, (--) für den hat. ‘the goal has for the hero’ (3.6) und es kommt NUR darauf an d; (-) dass:- (--) ‘and it only matters that’ SIE den HELden so BAUen-= ‘you build the hero in a way’ =dass der ZUschauer ‘that the spectator can identify with the hero’ (---) kann um einen teller ERB ‘can be about a bowl of pea soup’ (1.0) oder um den norMAlen Lebensunterhalt; ‘or about normal living’ [des is eben so] ‘that’s it’ [ aber äh verst] ‘yes but PRT underst do you understand I think you are making’

In this way, verstehst du? is what Jefferson (1981) terms a “post-response pursuit of response token”. However, in contrast to other tokens that do this, verstehst du? projects a reformulation of the speaker’s point which was not responded to adequately from his/her point of view.

98

Arnulf Deppermann

13

einen RICHtigen- (-) also einen RICHtich äh; ‘a real PRT a real PRT’ mh dramaTURgischen denkfehler den wir FJETZT, ‘PRT dramaturgical mistake of reasoning which we now’ (---) am beGINN unserer ARbeit miteinander AUSräumen müssen. ‘at the beginning of our work together need to get rid of’ (2.0) es kommt NUR darauf an was das ziel für DIE:sen‘It is only important what the goal for this’ (--) DIEses indiVIduum dass sie jetzt SCHAFfen MÜSsen, ‘this individual that you must create now’ (---) beDEUtet. ‘means’

14 15 16 17 18 19

The extract starts with a professor’s instruction (lines 01–08) on how to construct a protagonist. The student KA confirms this instruction by reformulating the upshot (lines 10–11.). Obviously, the professor does not accept her turn as a sufficient display of understanding. In particular, the rather vague reference to norMAlen lebensunterhalt (‘normal living’, line 10) runs counter to a maxim which the professor advocates, namely that a character needs to be created with reference to concrete actions. The professor responds to the student’s reformulation with an adversative turn: “ja aber verst verstehn sie” (line 12). The turn-beginning is produced in overlap with the student’s reaction; it projects that the student does not understand the instruction well enough from the professor’s perspective. The recycled turn-beginning and its prosodical integration with the upcoming turn shows that the professor is not using verstehen sie to elicit a response, but that he is projecting a reformulation. This reformulation is not delivered immediately, however, and the professor first gives an account why the point he is insisting on is extremely important for the students: The professor criticizes the students’ conception as a dramaturgical mistake of reasoning, (dramaTURgischen denkfehler, line 14), which needs to be abolished (JETZT […] AUSräumen müssen, line 15). Then follows the reformulation, which is iconically indexed by the repetition of the formula es kommt NUR darauf an (‘it only matters’, line 17), which serves to mark the upshot. The turn which is prefaced by verstehen sie? thus consists of an upgraded reformulation of the prior criticism.. Verstehst du? is a discourse marker construction, which is used here in a canonical sequential pattern, which runs like this: 1. 2. 3. 4. 5.

A: Main point B: Inadequate uptake (from A’s point of view) A: Verstehst du? Reformulation of main point B: Repair of uptake

Constructions vs. lexical items as sources of complex meanings

99

In this way, verstehst du? is used as a display of an inadequate uptake. “Inadequacy” from A’s point of view may relate to various matters: The response might have been minimal, disaffiliative, or, as in extract 1, misaligned in terms of being premature or displaying some misunderstanding of A’s prior turn. In this way, B’s turn is taken to index a deficit in the observable achievement of intersubjectivity. This deficit may have different sources and may concern different levels of meaning. Accordingly, the more specific local functions of verstehst du? in its sequential context vary. Verstehst du? works both retrospectively and prospectively: Retrospectively, it upgrades the relevance of the point the speaker has made, but at the same time, it projects a reformulation and perhaps an expansion of this point as the topic of the upcoming turn. Its projective force is furthermore made clear by prosodic integration with the turn-continuation. The verstehst du?-construction has a double temporal scope, i.e. it belongs to both the previous and the present turns of the speaker. Its temporal-indexical interpretive properties, which emerge out of its routine use in canonical sequences, are thus:

}

A: Main point retrospective interpretation B: Inadequate uptake (from A’s point of view); of prior interaction A: Verstehst du? Reformulation of main point Projection of turn-continuation B: Adequate uptake and preferred reaction

}

As far as the topic is concerned, the construction indexes only one topic, i.e. the speaker’s position as expressed by the speaker’s prior turn, which will now be elaborated. This temporally double and topically single scope makes the construction a marker of relevance and insistence. In such cases, verstehst du? is not used as a tag, which prompts ratification, but rather as a call for enhanced attention and more profound and observable cognitive processing, which is prompted by the speaker’s diagnosis that the addressee still has not achieved sufficient understanding. Although grammaticalized, the original lexical semantics of the verb verstehen is still present in the construction, albeit in the negative: Adequate understanding on the part of the listener has not yet been reached from the speaker’s point of view. The professor allocates the students a task of understanding, which is cognitive (denkfehler, ‘mistake of reasoning’, line 14), but also practical, because it calls for a practical accomplishment (schaffen müssen, ‘must create’, line 18). So, not only the meaning of “grasping intellectually” is at issue, but also the meaning of “acceptance” and “confirmation of knowledge”.

100

Arnulf Deppermann

Verstehst du?-constructions are used when the achievement of intersubjectivity becomes problematic. They occur in the context of repeated initiatives and repeated accounts of positions which B reacts to only minimally or not in the way A expects. In the data, such problems arise when – B overtly rejects A’s position, – B produces reactions which make it obvious for A that the B lacks relevant knowledge and misunderstands A’s prior turns, – B does not respond at several TRPs, – B departs in his behaviour from essential normative expectations, so that A starts to doubt whether B can be regarded as a competent member. All four problems are present in extract 2. The extract is from a mediation session. Mrs. Heuler (B) raises accusations which the other parties, including the mediator (C), consider to be ridiculous. The mediator and her opponent’s advocate try to persuade her to withdraw her accusations, but she does not react at most TRPs. When she does respond, however, she defends her accusations with absurd arguments. The following extract starts in the midst of an extensive multi-unit turn in which the mediator tries to convince Mrs. Heuler once again to drop her charge and abstain from appealing to court. 2 Mediation 3001.22 B_00. 03. 25–00.03.41 (Kartoffelklau) 01 C: zum geRICHT gehen frau (.) heuler? ‘if you go now to court Mrs. Heuler’ 02 (--) dann KRIEgen sie allein ‘then possibly you get PRT already’ 03

‘do you understand?’ 05 B: hört mal zu:. ‘listen-PL PRT’ 06 C: eine strafe möglicherwei’ in‘there you will PRT possibly a fine of’ 07 (-) wenn FÜNFhundert (REIchen). ‘if five-hundred are enough’ 08 (-) hörn sie wenn sie verURteilt werden,= ‘listen-SG if you are convicted’ 09 =woran ich eigentlich ich mein ich bin: hier nicht das geRICHT, ‘what I actually I mean I am not the court here’ 10 (-) aber nach der beWEISlage‘but according to the evidence available’

The mediator adds “verSTEHN sie” with high-rise “question-intonation” (line 04) as a tag to her turn, in which she insists that Mrs. Heuler should not appeal to the court. Mrs. Heuler does not respond to the mediator’s pre-

Constructions vs. lexical items as sources of complex meanings

101

vious turns. “VerSTEHN sie” is used here as a tag, serving as an understanding-check to solicit a hitherto lacking response. Because the tag is added to the prior TCU without a delay, which would call for a turn-transition, the use of verSTEHN sie? does not seem to be motivated (only) locally, but it points to a more global problem of accomplishing intersubjectivity. This problem is not solved, but it is confirmed in the extract: Instead of responding to the mediator’s argument, Mrs. Heuler asks her opponents to listen, as if they, not she, were the primary target of the mediator’s turns (line 05). Obviously, the mediator is so embarrassed by this unexpected response that she produces an incoherent, anacoluthic turn (lines 06–10). This turn reformulates her prior argument (line 06), partially recycling her prior construction (da kriegen sie […] schon eine strafe, line 03–05 vs. da werden sie schon eine strafe, line 06). It follows another discourse marker construction hörn sie (‘listen’, line 08), which is similar to verstehn sie and which also employs a mental verb (line 09). The mediator thus takes up the lexeme hören from Mrs. Heuler’s command “hört (.) mal (.) ZU” (line 05) and redirects it against Mrs. Heuler herself. “Hörn sie” is a discourse marker, which like verstehen sie? indexes a problem of understanding and acceptance, but in addition to the latter, it makes the recipient’s lack of attention an issue. Verstehen sie? and hören sie? can also both be used as a reproaching device in order to ask the addressee to comply with the basic requirements of interactional participation. Extract 2 thus differs from extract 1, because verstehen sie? here is used as a turn-final tag and not as a turn-initial projector. However, it also deals with inadequate uptake of the speaker’s position, and the speaker also reformulates her position after the recipient does not respond to verstehst du? with an aligning response. Thus, despite the differences in the sequential organisation of verstehst du? regarding turn-design and turn-taking, the more general function of dealing with a problem of a lack of intersubjectivity and projecting the need to recycle the speaker’s main point due to lingering inadequate uptake is identical. Problems with achieving intersubjectivity, however, can be much more local. In Extract 3, the professor HA objects to an argumentation produced by the student Cornelius (CO). When the student does not respond, the professor uses verstehen sie? to elicit a response.

102

Arnulf Deppermann

3 Pitching: Bernd2_00:10:07–00:10:19 01 CO: [(…) ] 02 HA: [cornelius] verZEIHung; ‘Cornelius excuse me’ 03 (-) das is nun einfach dahe (.) daHIN geredet; ‘now this is simply just chatter’ 04 (--) ’äh er kann doch nun was FTU:N um die frau zu überZEUgen oder? ‘PRT he can PRT do something in order to convince the woman PRT?’ 05 (---) er kann doch [RAUS]gehen und den RA ‘he can PRT go out and mow the lawn.’ 06 CO: [ja? ] ‘yes?’ 07 HA: (--) also äh’ (-) äh ver[stehn] sie (.) ‘PRT PRT PRT do you understand? PRT’ 08 CO: [okay.] ‘okay’ 09 RA: ‘but he has PRT’ 11 CO: [aber (--) ich meine ‘but I mean’

Having produced his objection, the professor shows by using the tag oder? with high-rise “question-intonation” (line 04) and leaving a pause after it that he expects the student to produce an agreement. The student, however, does not react, and the professor carries on with his argumentation, adding an example (line 05), which again makes turn-transition relevant, because the turn ends reaching the speaker’s lowest pitch register and is followed by a pause. At this TRP, the student again does not show whether he aligns with the professor’s position. Resuming his turn with “also äh verstehn sie” (line 07), the professor now projects an insisting reformulation of his position. Verstehen sie? here does not seem to be simply a tag which allocates the turn to the student. Both the discourse markers – “also äh’ (-) äh” before verstehn sie and ähm (line 07) after it – are turn-holding devices which project an expansion of his argument. The student now shows his sensitivity to the fact that his uptake has repeatedly been made relevant by producing an agreement token, which, however, is only used to mitigate a following disagreement (see aber ‘but’, line 11). Meanwhile, the second professor, RA, claims the floor (lines 09–10). Verstehen sie? is again used in an environment where B does not give in to A’s standpoint. It is not used as a means to elicit some reaction, but it projects a self-reformulation of the upshot of the speaker’s position. Recipients can anticipate the projected turn-continuation and preempt it by directly reacting

Constructions vs. lexical items as sources of complex meanings

103

to verstehst du?, even if no TRP is reached. In the face of the partner taking the turn, the producer of verstehst du? can forgo the reformulation and wait until the partner produces a satisfactory response to his position: So, recipients sometimes use the temporal-indexical interpretive properties of the verstehst du?-construction, which retrospectively indexes a prior inadequate uptake on the recipient’s part, while prospectively projecting a reformulation of the speaker’s point, which calls for a repair of the prior response. The recipient thus can cut the projected sequence short by immediately delivering a repair of his/her uptake which preempts the reformulation of the speaker’s main point. 1. 2. 3. 4.

A: B: A: B:

Main point Inadequate uptake (from A’s point of view) Verstehst du? Repair of uptake

This is what happens in extract 3: After the intervention of the second professor RA, the student CO starts to produce an elaborate response to the professor’s argumentation (line 11). In such contexts, verstehst du? actually becomes a tag which serves to elicit a response from B. This is paradoxical, because it projects a turn-continuation of the speaker, which, however, is framed as a subsidiary activity, and which becomes necessary because of a lack of uptake. But precisely because of this subsidiary property, the projected turn-continuation can be preempted by a repaired, upgraded response from the recipient, who may understand that the speaker may abort his/ her turn if the recipient starts to deliver a more adequate response from the speaker’s point of view. In sum, one use of verstehst du? occurs in a context in which the accomplishment of intersubjectivity becomes problematic from the speaker’s point of view, because the recipient does not display an uptake which is sufficient for the speaker. In some of the cases in the corpus, verstehst du? is used as a tag calling directly for a repaired and enhanced response. It most cases, however, it projects a reformulation and establishes a conditional relevance for a repaired uptake of the reformulation. The bridge between these two sequential patterns may consist both of preemptive cases as in Extract 3, where B anticipates the reformulation, and of varying expectations of the producer of verstehst du?: If A assumes that B is able to produce a repaired response immediately, A may produce verstehst du? as a tag which allocates the turn to B to elicit a direct response; if A assumes that B is not yet able to do so, A integrates verstehst du? with a following reformulation, leaving no room for an immediate response, but implying that it is due after the reformulation.

104

Arnulf Deppermann

4.3.2 Verstehst du? indicating problems of formulation Problems of achieving intersubjectivity may arise from a speaker’s problems of formulation, i.e. of conceptualization or encoding. In these cases, it is neither a lack of uptake nor a lack of understanding or acceptance on the part of the listener which causes failure, but is rather the speaker’s fault. In these cases, verstehst du?-constructions index that the speaker him/herself assumes that s/he has not yet managed to produce a formulation which should provide for an adequate understanding on the part of the listener. This use of verstehst du? co-occurs with other indices of problems of formulations, such as hesitation phenomena, cut-offs, self-repairs, reformulations, accounts, etc. A case in point is extract 4, in which the professor repeatedly tries to convey to the students what is essential for a good story. 4 Pitchings: Journalistin3_00. 43. 05–00.43.25 01 HA: DENKT bitte:; (-) an DIEse;(-) an diese WARnung- und DENKT dran‘please mind this this warning and mind that’ 02 ??: (-) [.hh, ] 03 HA: [dass-] ‘that’ 04 (-) ’äh ’äh‘PRT on the other side PRT’ 05 (-) versteht ihr; ‘do you understand’ 06 (--) es MUSS: n geWISser; ‘there must be a certain’ 07 (--) ä::h äh äh SCHAUwert; ‘PRT PRT PRT PRT show value’ 08 (-) in der geschichte DRIN sein‘in the story’ 09 (--) sonst is sie wieder nich: GLAUBw?ürdig‘otherwise it is not really credible’ 10 RA: ja‘yes’ 11 HA: also (--) verSTEHT ihr es MUSS diese; ‘PRT do you understand it must be this’ 12 (--) ((schluckt)) ich ha hab MACHT‘((swallows)) I have no idea how you do it’

Before the professor manages to find a formulation of a criterion for a good story in line 07 (schauwert ‘show value’), he produces cut-offs (line 04) and hesitation markers (äh, lines 04 and 07) and intra-turn pauses, and he projects with versteht ihr (line 05) that an important point is to follow. Then he projects another reformulation, again using versteht ihr (line 11), which is again aban-

Constructions vs. lexical items as sources of complex meanings

105

doned. Finally (line 12), he concedes that his attempt at finding a solution for the formulation problem has failed. Versteht ihr here projects the formulation of the relevant criterion, which is started twice by es muss (lines 06 and 11). The professor makes it clear that he assumes that the main point has not yet been understood, because he has not managed to find a proper formulation. Verstehen sie is a turn-holding device and projects a new attempt at formulating the relevant point. More specifically, it calls for enhanced efforts to understand on the part of the listeners, i.e. it indexes that the listener should cooperate in the constitution of a comprehensible message, which the speaker himself does not manage to produce alone. In the end, the speaker admits failure and the task is handed over to the listeners (ich hab keine ahnung wie ihr das macht ‘I have no idea how you do it’, line 12). 4.3.3 Verstehst du? used for refocusing When a speaker produces a thematic digression, verstehst du? can be used to announce a return to the prior focus of the interaction, i.e. that which was operative before the digression. It may be brought about by a self-referential comment or an explanation which the speaker produced in an immediately preceding account. Verstehst du? is then used to reorient the listener to the main line of talk and to project its continuation; that is, it turns the comment into a parenthesis which deserves no further attention. An example is extract 5 from a talk show. The famous entertainer Karl Moik (KM) tells how he became an anchorman for folk music shows on TV. 5 Talkshow 4050.08C_00:16:00–00:16:21 01 KM: und da sin ma mal FURCHbar BECHern gegangen,= ‘and then we once PRT went carousing’ 02 =und – (---) äh in für FA:LSCH, (‘I PRT consider this PRT in a certain way to be wrong’; lines 08–9). Now, its revised semantics – “not to accept a position” – is displayed by the self-repair. The student goes on not only to formulate the position he does not accept. In addition, he gives an account why he does not accept Saussure’s notion of the “symbol” (lines 10–14). Da verSTEH ich nicht thus refers to something which can be grasped semantically, but which is argumentatively flawed from the student’s point of view. In the next TCUs of his multiunit turn, however, the student makes clear that his argument is designed to clarify with the professor’s help whether there are reasons in favour of Saussure’s position. The pragmatics of da verSTEH ich nich, thus, is not a plain rejection, but a conditional one, which is presented as being negotiable in the light of additional justification for the disputed view. Da verSTEH ich nicht projects the search for such a justification. The meaning “to see/accept the reasons” carries an argumentative

Constructions vs. lexical items as sources of complex meanings

111

connotation, i.e. it questions the availability of “good reasons” for some statement or action. This use of verstehen is a conversational equivalent to the concept of semantics in normative-argumentative theories of meaning, which are advocated e.g. by Habermas (1992) and Brandom (1994). According to these authors, participants treat the comprehensibility of actions and formulations as being dependent on the intelligibility and acceptability of the reasons which can be recovered for them. Because of this inextricable link between semantic and argumentative connotations, the negative construction is used to refer to both semantic and argumentative problems.12 5.2.1 The negative construction as a pre-disagreement In extract 6, the negative construction was used for formulating a complex argument which dealt with conditional disagreement in search of a deeper understanding of a problematic position. Nicht verstehen, however, can also be used to foreshadow disagreement in a more obvious and clear-cut way. In the following extract, disagreement is displayed by the use of nicht verstehen as a matrix sentence for objections. The extract is from the same mediation session as extract 3. Mrs. Heuler (B) claims that her neighbours stole potatoes from her cellar. The mediator (C) counters this claim with objections: 7 Schlichtung 3001.22 Kartoffelklau 01 C: also ich ä:h‘PRT I PRT’ 02 (1.5) 03 versteh eigentlich net frau SO:HN ( ) ] ‘although your son’ 06 B: [= ‘the leprechauns didn’t take them’ 07 C: =bitte? ‘pardon?’ 08 B: die heinzel[männchen waren nicht da-] ‘the leprechauns were not there’ 09 C: [ah ja vielleicht ] vielleicht hat ihr so:hn ab und ‘well maybe maybe your son sometimes’ 12

See Deppermann (2005a) on the reflexivity of semantics and argumentation in conversation.

112

Arnulf Deppermann

10

zu ‘just if you say’ 11 [ihr sohn trägt se ho:ch? (-) der trägt se aber ho:ch-] ‘your son carries them up but he carries them up’ 12 B: [ ‘I do PRT that’ 14 (-) der trägt bloß die flaschen [runter. ] ‘he only takes the bottles down’ 15 C: [ ‘I imagine when you they are piled up’

The mediator makes clear that her lack of understanding refers to the fact that she does not see sufficient reasons for Mrs. Heuler’s claim that there was a theft. She does so by producing a series of objections, which define the local semantics and pragmatics of the negative construction. The mediator does not explicitly reject Mrs. Heuler’s claim; rather, the negative construction ich versteh eigentlich net (‘I actually do not understand’, lines 01–03), which opens up her opposing turn, projects a disagreement (Jacobs and Jackson 1989).13 The pre-disagreement provides the recipient with the opportunity to self-repair or withdraw her claim without having to face overt disagreement and conflict. It thus also serves to avoid an explicit reproach of being unreasonable or non-credible. Because of these projective properties as a pre-disagreement, the negative construction projects a threat to the status of a rational co-participant, but at the same time, it also gives him/her the chance to defend and maintain it (cf. Deppermann 2005b: 204–209). 5.2.2 The negative construction as a reproach Extracts 8 and 9 show how the interpretation of the negative construction as a reproach is locally made relevant and negotiated by the participants. The process of negotiation makes obvious that the negative construction has a particular rhetorical potential. Extract 8 is also from a mediation session. Both opponents (A1 and B), who are in their fifties, had pressed charges against each other. The mediator (C) comments on this fact: 13

In fact, the mediator produces an overt disagreement a little later, after B insists on her position despite the mediator’s attempts to show that the defense of B’s position amounts to subscribing to absurd arguments (see Deppermann 2005b: 204–209).

Constructions vs. lexical items as sources of complex meanings

113

8 Schlichtung 3001.20_00. 04. 16–00.04.32 (Hitler) 01 C: 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18

wem mer in dem ALder is,= ‘when one is this age’ =also das verSTEH isch nit wie‘PRT I don’t understand how’ (.) wie mer sich da gegenseidisch des LE: ‘how one give each other a hard time’ A1: ja, (-) SO [is es.] ‘yes that’s right’ B: [(…)-][WER hat] die (schuld) ‘who is to blame?’ C: [also:; ] ‘PRT’ A1: ja [das KOMMT doch von] UNne RUFF, ‘PRT that comes PRT up from below14’ C: [JA: isch mein des;] ‘PRT that’s what I mean’ B: [moMENT- ] ‘wait a moment’ C: [des WIRD,] ‘this will’ des WIRD [hier noch;] ‘this will be here’ B: [er HAT, ] ‘he lived’ B: [mir gegenÜwwer geWOHNT,] ‘on the other side of the street’ C: [wird ja wohl uff ] BEIde ‘will PRT PRT on both’ B: [und da hat er ETlische male] [die:;] ‘and several times he got’ C: [SEIde wird ja des nit ganz ] oKAY: [sein;] ‘sides this doesn’t PRT seem to be thoroughly okay’ B: ‘to put it plainly because he behaved in a way’

In this extract, the meaning of the matrix sentence versteh isch nit (‘I don’t understand’, line 02) is clarified as “I cannot see/accept the reasons” by the negative assessment in the dependent sentence (wie mer sich da gegenseidisch des LE:we schwer macht; ‘how one gives each other a hard time’, line 03). Here, the lexical instantiation of the complement sentence provides the local meaning of the negative construction. But what kind of action does it perform? By his rhetorical question WER hat die schuld? (‘who is to blame?’, line 05), which B 14

The opponent lived in an apartment which was on the other side of the street, one floor below the speaker’s.

114

Arnulf Deppermann

directs towards A1, B displays that he understands C’s turn as a reproach. A1 rejects this and blames B instead (das KOMMT doch von UNne RUFF, ‘that comes PRT up from below’, line 07). A sequence of reciprocal reproaches emerges. C’s initial turn could also have been understood as a rejection of the behaviour in question (pressing charges), without reproaching anyone. Indeed, mediators use such negative assessments in order to appeal to the opponents’ common sense, persuading them to act according to higher-order values and bring the conflict to a closure. However, this pragmatic meaning is taken up neither by the opponents, who instead treat the mediator’s turn as an opening of a blame-negotiation, nor by the mediator, who does not manage to calm down both sides. As already mentioned in Section 5.1, in eleven instances, the negative construction is used to introduce an objection which refers to a contradiction and thereby makes the opponent’s argument look flawed. The contradiction is formulated in an adversative turn-format. In another mediation session, A1 protests against her being fired because of being absent from work without a valid excuse. As evidence, her employer (B1) describes how she went to a coffee shop while allegedly missing work for a doctor’s visit. 9 Schlichtung 3003.119.3_00. 12. 46–00. 13. 2101 (kaffee) ich versteh auch andereseits FNI:CHT? wenn die‘on the other hand I don’t understand when the’ 02 (-) klägerin zum arzt nach BRUCHtal (.) MUSS JA:? ‘plaintiff (=A1) needs to go to the doctor in Bruchtal’ 03 (--) dass sie dann (-) in de innenSTADT im maxim‘that she then goes to Maxim’s in the city’ 04 (-) KAFfee trinkt; ‘to have a cup of coffee’ 05 (1.8) 06 A1: ((sniffs)) ‘((sniffs)) why is that forbidden?’ 07 B1: (---) ‘only said that I don’t understand it’ 09 (1.7) 10 A1: isch hatt daheim keine ZEIT mehr d kaffee zu trinke,= ‘I didn’t have time to have my coffee at home’ 11 =weil isch konnt mir auch keiner mache weil mir der STROM abgestellt ‘because I couldn’t prepare any either because the electricity was’ 12 worre is‘turned off’

B1:

Constructions vs. lexical items as sources of complex meanings 13 14 15

16 17 18

115

AUCH durch den die firma säuberle; ‘also by the Säuberle company’ (--) weil isch s net ZAHle konnt? ‘because I couldn’t pay for it’ (--) also irgendwo !MUSS! isch jo mol ä tass warme kaffee trinke od, ‘well I need to have PRT a cup of hot coffee somewhere don’t I’ (2.4) ?: [ja- ] ‘yes’ A1: [des-](-) des kann mir KÄNner verBIEte; ‘nobody can forbid me to do that’

By contrasting the necessity of going to the doctor with going to a coffee shop (lines 01–04), B1 makes clear that ich verstehe ((…)) NICHT (line 01) locally means “I cannot see/accept the reasons” for A1’s action. A1 treats this as a reproach: wieso is das verBOte? (‘why is that forbidden?’, line 06). This response presupposes that B1 meant that going to a coffee shop is forbidden when one has to go to the doctor’s. B1, however, rejects this interpretation of his prior turn: ich hab jetz (-) nur gsagt ich verSTEH s net (‘I now only said that I don’t understand it’, lines 07–08). B1 insists on a “literal” interpretation. Instead of categorizing his first-positioned turn as a reproach, he (re)categorizes it as a repair-initiation, referring to a problem of understanding. The negative construction is thus framed as a request for justification. A1 delivers this justification in her next turn (lines 10–15), and when she does not receive an uptake (cf. the 2.4 second pause in line 16), she explicitly rejects the reproach (des kann mir KÄNner verBIEte; ‘nobody can forbid me to do that’, line 18). Extract 9 shows the systematic ambiguity of the negative construction. It can be understood as a conventional way of producing a reproach, but this interpretation is always defeasible by reference to the lexical semantics of verstehen. Participants themselves can thus operate either with an idiomatic, non-compositional meaning of the negative construction as such, or with a compositional, lexically-based meaning of verstehen, and participants can argue about which meaning was intended. Similar to why-questions (see Günthner 2000), the negative construction is a rhetorical resource which can be used for conventionally conveying a reproach without needing to defend this interpretation, because the speaker can always (re)interpret his/her turn as a next-turn repair-initiator displaying a lack of understanding and calling for some justification. In the interactional sequence, turns with the negative construction therefore often remain ambiguous. This also the case in extract 9. The justification provided by A1 can either be interpreted as an answer to

116

Arnulf Deppermann

a request for justification, or a defending statement against a reproach. In turn, the producer of the negative construction often displays his/her negative moral assessment by other actions (here: B1 fired A because of her absenteeism), thus preserving the interactional relevance of the interpretation as a reproach even if the speaker denies it, as in extract 9, line 08. Sequential placement and the prosody of reproaches and disagreements with the negative construction suggest that this construction is stylistically marked. It appears in the context of arguments which are framed as rational disputes, and it is never realized with a high-involvement prosody contextualizing excitement and indignation. This is in contrast to other formats for constructing reproaches (cf. Günthner 2000). The negative construction thus belongs to the register of stylizing a dispute as a rational argument.

6

Conclusions

This section discusses the results of the analyses of the two verstehen-constructions with respect to the question of the lexical vs. constructional basis of meaning. I then point out some questions regarding the notion of “construction”. 6.1 Lexical items vs. phrasal constructions as bases of meaning In the data analysed, verstehen can have five meanings: – to be able to perceive speech – to identify a referent – to grasp some dimension of meaning (intension, intention, inferences, allusions) – to see reasons and motives for a position – to accept a position These meanings are related to each other by psychological and pragmatic motivation. The semantic spectrum is characterized by family-resemblances (cf. Wittgenstein 1953). The same applies to the pragmatic functions and the interactive uses of the various constructions with verstehen. For example, grasping meaning is often the prerequisite for recovering motives and reasons, and behaviours are interpreted as actions by imputing reasons and intentions. Understanding the meaning and the motives of a turn may be the prerequisite for its acceptance.15 15

See Clark (1996: 221–252), who describes this conditional relation as “upward completion” of joint actions.

Constructions vs. lexical items as sources of complex meanings

117

In both the verstehst du?-constructions and the negative verstehen-constructions, the meanings “to grasp the meaning”, “to see/accept reasons and motives”, and “to accept a position” were found. Although the verstehst du?-construction is fairly grammaticalized, its lexical basis is still present in most of its occurrences. There is only one subtype of each construction which has a semantics that seems to be unique and which is also distinct from other uses of the same formal pattern. The first is the semantically light use of verstehst du? as a marker of refocusing (see 4.3.3); the second is the use of the negative construction as a reproach (see 5.2.2), which is a moralizing and personalizing extension of the meaning “(not) to accept a position”. This interpretation, however, is hard to pinpoint; the negative construction is often vague and ambiguous, and its interpretation is defeasible, which makes it useful as a rhetorical resource. In sum, for both constructions, there is a considerable intra-constructional variance of possible meanings, while the spectra of meaning shared by both constructions overlap considerably. Moreover, this intra-constructional variance of meaning is organized in a similar way: – The different meanings are often hard to tell apart; the constructs are often vague and ambiguous, but in most cases, this does not become an interactional issue (however, see extract 9). – The meaning of the construct is specified in the interaction sequence and not determined by the construction itself. For the negative construction, the lexical instantiation of the object-NP or the COMP-sentence constrains the semantics of nicht verstehen (cf. extract 8, line 03: wie mer sich da gegenseidisch des LE:we schwer macht;, “how one gives each other a hard time”). Apart from such specifications by instantiation, there are other practices external to the construction which are used for clarifying its local meaning. These are e.g. reformulation (see extract 6, lines 08–09: halt ich ((…)) für falsch, “I consider this to be wrong”) and argumentative practices, such as objection (see extact 7) and pointing out contradictions (see extract 8). The meaning of a single construct can often only be constrained within the wider sequential context or even with reference to cultural norms and knowledge about the participants’ attitudes. While the semantic spectra of both constructions overlap considerably, their syntax is completely different. Consequently, syntactic motivations for similarities in meaning can be ruled out. What is more, the same spectrum of meaning for verstehen is also present in other constructions. These findings corroborate the view that the semantic similarities of both constructions rely on a shared lexicosemantic meaning potential (see Norén and Linell 2007; Linell 2009: Ch.15) of verstehen, which seems to operate more or less indepen-

118

Arnulf Deppermann

dently of the specific construction. This context-free potential is specified locally in the interactional sequence, but most often independently of the construction. Local meanings are not simply compositional: Verstehen as such is polysemous, and its meaning is sometimes locally disambiguated, while in other cases it remains vague and ambiguous. Regarding pragmatic and interactional properties, there are also similarities between both constructions. They both occur in the context of argumentations and assessments (cf. Imo 2007b: 291–292). However, their sequential, projective, and action-related properties are different: While verstehst du? projects a reformulation of the speaker’s position which received insufficient uptake and is mainly tied to instructional contexts, the negative construction projects (and, indeed, is part of) the constitution of a pre-disagreement or a reproach and is used in argument sequences which are performed as being “rational debates”. While these interactional properties are clearly different, both constructions converge in that they are used to achieve intersubjectivity and mutual alignment, and they index that the partner’s turn is judged to be linguistically, semantically, or pragmatically flawed. What conclusions can we draw from these findings for a model of the local constitution of meaning? Four systematic sources of meaning could be shown to be relevant for the local interpretation of verstehen: a) the lexical meaning potential b) the meaning of the construction c) interactive practices and specification of meaning in the sequential context d) background knowledge a) Lexical meaning potential In most cases, the lemma verstehen constrains the semantics of the constructs independently of the phrasal constructions in which it occurs. The semantics of verstehen itself, however, is polysemous or underspecified. When talking about “ambiguity” and “polysemy”, we must be clear whether we are referring to the representation of lexical items in the (mental) lexicon or to their situated uses (see Deppermann 2000). If “underspecification” and “polysemy” refer to different representations of context-free items in the mental lexicon, a conversation analytic study cannot answer this question, because the difference cannot be linked directly to observable verbal action in talkin-interaction. Conversation analysis can only deal with the local semantics of instances of use as it is displayed by participants’ uses of constructions and their reactions. Turning to this, we see that participants make rhetorical use of the systematic ambiguity of the negative construction between the

Constructions vs. lexical items as sources of complex meanings

119

two meanings “to see reasons and motives for a position” and “to accept a position” (cf. extract 9): While the latter meaning is used for producing a reproach, the former can be appealed to if the reproach is countered by the recipient. Thus, ambiguity is never formulated by a participant as being a problem, but it is used as a resource for rhetorical concerns of local (re)interpretation. Underspecification might be relevant insofar as the most general, underspecified meaning of verstehen can be paraphrased as “successful cognitive processing of some symbolic object”. This meaning is encompassed in all the polysemous meanings. The precise nature of the criteria for success (e.g. identifying referents, uncovering reasons) and the objects (e.g. sound patterns, intentions, actions) then define how the meaning of verstehen is specified. b) Meaning of the construction Only for the subtypes “refocusing” and “reproach” is the meaning specific to the construction. This statement, however, is partially misleading, because although these meanings are conventional and construction-specific, they are not context-free. The construction itself can acquire different meanings depending on its instantiation and the local context. Thus, the construction does not have a determinate meaning, but meaning-potentials (see Norén and Linell 2007; Linell 2009: Ch. 15), which are realized only with respect to types of context and clarifying pre- or post-positioned sequential activities. c) Interactive practices and specification of meaning in the sequential context The specification of the meaning of verstehen in the constructs is mainly based on its local sequential context. The meaning of verstehst du? seems to be determined mainly by the preceding context (i.e. the position taken by the speaker and the minimal uptake of this position), its timing, and its prosodic integration with prior and following TCUs. For the negative construction, the progression of the interaction seems more important: Self-repairs and reformulations, argumentation, and recipients’ reactions determine its meaning in cooperation with more general preferences for the interpretation provided for by the activity type in operation (e.g. dispute). For the negative construction, the instantiation of the complement plays a major role in determining the meaning of verstehen, as it provides coercion phenomena (cf. Michaelis 2005). d) Background knowledge The formulation of the constructs as well as preceding and following TCUs contextualize frames of background knowledge which contribute to the local interpretation of verstehen. For example, the idiomatic formula sich gegen-

120

Arnulf Deppermann

seitig das Leben schwer machen (‘to give each other a hard time’, cf. extract 8), indexes a scenario which does not comply with criteria of a good interpersonal relationship. Arguing with contradictions (like in extract 9) relies on background knowledge, which is often needed not only to understand the precise nature of the contradiction, but also to discover the contradiction in the first place, because the contradiction often is not marked by connectives or other lexical devices. These four sources of interpretation overlap with those proposed by Fischer (2006) for the functional interpretation of discourse particles (“invariant meaning aspects”, “constructions”, and “communicative background frames”). Her model needs to be expanded to include the specification of meaning by the sequential context and the practices the interactants use, because these contribute to and constrain substantial aspects of meaning, which are not provided by the other three sources. For instance, while verstehst du? as projecting a reformulation or marking relevance can still be regarded as a sequentially-based elaboration of the basic lexical meaning of verstehen, the use of verstehst du? for refocusing seems to be based on sequential grounds only, since it is very remote from the lexical semantics of verstehen. The sequential practices embody the local pragmatic works of the participants to specify and negotiate meaning according to their practical situated interactional business (cf. Deppermann 2005a, 2007; Schegloff 1984). Moreover, the sequential context is needed to access relevant background knowledge which is not directly contextualized by the lexical item verstehen and the phrasal constructions. This study has shown that the meaning of constructs in context derives neither from context-free constructional meanings nor from context-free lexical meanings. Although both of these sources of meaning can be seen to offer a scope of routine ways of interpretation, the precise local meaning emerges from a complex interplay between the conversational history, the vocal and linguistic adaptation of constructions to the contextual moment of their production, interactional negotiation, and relevant background knowledge. In this way, not only the formal side of grammatical constructions is emergent (see Hopper 1998, 2004) – the same also applies to their meanings in situated interaction. 6.2 Problems with the notion of a “construction” From the above argumentation, several questions concerning the status of verstehst du? and the negative constructions as constructions arise. If the semantics of the constructs is not determined by the construction itself, and

Constructions vs. lexical items as sources of complex meanings

121

if it is not even specific to several of its many uses, then we may wonder whether they are constructions at all. If we stick to the definition of constructions as being form-meaning pairings (cf. Goldberg 1995; Croft 2001), then only the subtypes “marker of refocusing” of the verstehst du?-construction and the “reproach” meaning of the negative cases are real constructions. Only these two have a definite non-compositional function which hinges on the construction itself. If we view constructions from a usage-based approach (see Langacker 2000; Goldberg 2006; Bybee 2006), the picture is different. From this point of view, criteria for constructions are psycholinguistic entrenchment as a linguistic unit and the property of being a normatively expectable or even required way of encoding some conceptualization or some communicative function (cf. Feilke 1996). These usage-based conditions are fulfilled for the two constructions studied. Both of them are “encoding idioms” (cf. Fillmore, Kay and O’Connor 1988). Verstehst du? is idiomatic, while begreifst du? or erkennst du? would be comprehensible, but unidiomatic and pragmatically anomalous. The same applies to the negative construction, which cannot be replaced by semantically equivalent but unidiomatic expressions, such as keine Einsicht gewinnen or nicht erkennen, although nicht nachvollziehen können would work. In addition, the verstehst du?-construction definitely has a unitstatus because of its formal (phonetic and syntactic) reduction and its grammaticalization (cf. Bybee 2006). The requirement of a one-to-one mapping of form and function cannot be satisfied in the data. There are several potential meanings that are locally selected according to contextual parameters. It would be misleading to project facets of meaning and function which are solely provided by the context of different constructions, just as if these were self-contained entities. It seems that it is not necessary for a construction to have a determinate meaning or function by itself. Rather, we can view constructions as tools for the situated construction of context-sensitive turns: schemata which can be flexibly adapted to interactional contingencies (cf. Pekarek Doehler and Müller 2006). “The grammar of context” (Kay 1997) is just one aspect of the contextual determination of meaning. It contributes only one, sometimes small part to the local constitution of meaning.

122

Arnulf Deppermann

Transcription conventions GAT (Selting, Auer, Barden, Bergmann, CouperKuhlen, Günthner, Meier, Quasthoff, Schlobinski and Uhmann 1998) Sequential structure [ [

overlap and simultaneous talk

] ]

latching

= Pauses (.) (-), (--), (---) (2.85)

micropause (shorter than 0.2 sec) brief, mid, longer pauses of 0.2–0.5, 0.5–0.7, 0.7–1.0 sec. measured pause (more than one second)

Other segmental conventions

assimilations within units segmental lengthening, according to duration hesitation signals, so-called ‘filled pauses’ cut-off with glottal closure

und=äh :, ::, ::: äh, öh, etc. ’ Laughter so(h)o haha hehe hihi ((lacht))

laugh particles within talk laugh syllables description of laughter

Accentuation

strong, primary stress extra strong stress weaker, secondary stress

akZENT ak!ZENT! akzEnt

Pitch at the end of units

rising to high rising to mid level falling to mid falling to low

? , ; .

Conspicuous pitch jumps

to higher pitch to lower pitch

F G Changed register



low register high register

Constructions vs. lexical items as sources of complex meanings

123

Changes in loudness and speech rate









=forte, loud =fortissimo, very loud =piano, soft =pianissimo, very soft =allegro, fast =lento, slow =crescendo, continuously louder =diminuendo, continuously softer =accelerando, continuously faster =rallentando, continuously slower





Breathing

inbreath, according to duration outbreath, according to duration

.h, .hh, .hhh h, hh, hhh Other conventions ((coughs))

para- und extralinguistic activities and events commentaries regarding voice qualities with scope unintelligible according to duration uncertain transcription uncertain sounds or syllables possible alternatives omissions in the transcript

References Auer, P. 2005 Projection in interaction and projection in grammar. Text 25, 1, 7–36. 2006 Construction grammar meets conversation: Einige Überlegungen am Beispiel von “so”-Konstruktionen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 291–314. 2007 Syntax als Prozess. In: H. Hausendorf (ed.), Gespräch als Prozess. Linguistische Aspekte der Zeitlichkeit verbaler Interaktion, Tübingen, 95–142. Auer, P. & S. Günthner 2005 Die Entstehung von Diskursmarkern im Deutschen – ein Fall von Grammatikalisierung? In: T. Leuschner, T. Mortelmans & S. de Groodt (eds.), Grammatikalisierung im Deutschen, Berlin, 335–362. Barth-Weingarten, D. 2006 Fuzzy boundaries – Überlegungen zu einer Grammatik der gesprochenen Sprache nach konversationsanalytischen Kriterien. In: A. Deppermann, R. Fiehler & T. Spranz-Fogasy (eds.), Grammatik und Interaktion, Radolfzell, 69–94. www.verlag-gespraechsforschung.de/2006/deppermann.htm (25 February, 2009).

124

Arnulf Deppermann

Blakemore, D. 2002 Relevance and Linguistic Meaning: The Semantics and Pragmatics of Discourse Marker, Cambridge. Blutner, R. 2004 Pragmatics and the lexicon. In: Horn, L. & G. Ward, (eds.), Handbook of Pragmatics, Oxford, 488–514. Brandom, R. 1994 Making it Explicit, Cambridge. Bublitz, W. 1988 Supportive Fellow Speakers and Cooperative Conversations, Amsterdam. Bybee, J. 2006 From usage to grammar: The mind’s response to repetition. Language 82, 4, 711–733. Carston, R. 2002 Thoughts and Utterances: The Pragmatics of Explicit Communication, Oxford. Clark, H. 1996 Using Language, Cambridge. Croft, W. 2001 Radical Construction Grammar, Oxford. Croft, W. & D. A. Cruse 2004 Cognitive Linguistic, Cambridge. Deppermann, A. 1999 Gespräche analysieren, Opladen. 2000 Semantic shifts in argumentative processes: A step beyond the fallacy of equivocation. Argumentation 14, 17–30. 2005a Conversational interpretation of lexical items and conversational contrasting. In: Hakulinen, A. & M. Selting (eds.), Syntax and Lexis in Conversation, Amsterdam, 289–317. 2005b Glaubwürdigkeit im Konflikt. Rhetorische Techniken in Streitgesprächen, Radolfzell. http://www.verlag-gespraechsforschung.de/2005/deppermann.htm (24 February, 2009). 2007 Grammatik und Semantik aus gesprächsanalytischer Sicht, Berlin/New York. Fauconnier, G. 2004 Pragmatics and cognitive linguistics. In: Horn, L. & G. Ward (eds.), The Handbook of Pragmatics, Oxford, 657–674. Feilke, H. 1996 Sprache als soziale Gestalt, Frankfurt am Main. Fillmore, C., P. Kay & M. K. O’Connor 1988 Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64, 3, 501–538. Fischer, K. 2006 Frames, constructions, and invariant meanings: The functional polysemy of discourse particles. In: Fischer, K. (ed.), Approaches to Discourse Particles, Amsterdam, 427–447. Ford, C., B. Fox & S. Thompson 2002 Social interaction and grammar. In: M. Tomasello (ed.), The New Psychology of Language, 2, Mahwah, NJ, 119–143.

Constructions vs. lexical items as sources of complex meanings

125

Fraser, B. 2006 Towards a theory of discourse markers. In: K. Fischer (ed.), Discourse Particles, Amsterdam, 189–204. Goldberg, A. 1995 Constructions. A Construction Grammar Approach to Argument Structure, Chicago. 2006 Constructions at Work, Oxford. Günthner, S. 2000 Vorwurfsaktivitäten in der Alltagsinteraktion, Tübingen. Habermas, J. 1992 Toward a critique of the theory of meaning. In: Habermas, J. (ed.), Post-metaphysical Thinking, Cambridge MA, 57–86. Hopper, P. J. 1998 Emergent grammar. In: M. Tomasello (ed.), The New Psychology of Language, Mahwah, NJ, 155–175. 2004 The openness of grammatical constructions. Chicago Linguistic Society 40, 153–175. Houtkoop, H. & H. Mazeland 1985 Turns and discourse units in everyday conversation. Journal of Pragmatics 9, 595–619. Imo, W. 2007a Probleme der Anwendung der Construction Grammar bei der Analyse gesprochener Sprache. Gesprächsforschung, 8, 24–47. http://www.gespraechsfor schung-ozs.de/heft2007/ga-imo.pdf (24 February, 2009). 2007b Construction Grammar und Gesprochene-Sprache-Forschung, Tübingen. Jackendoff, R. 1997 The Architecture of the Language Faculty, Cambridge, MA. 2002 Foundations of Language (Brain, Meaning, Grammar, Evolution), Oxford/New York. Jacobs, S. & S. Jackson 1989 Building a model of conversational argument. In: B. Dervin, L. Grossberg, B. O’Keefe & E. Wartella (eds.), Rethinking Communication, 2, Newbury Park, CA, 153–171. Jefferson, G. 1981 The abominable ne? An exploration of post-response pursuit of response. In: P. Schröder & H. Steger (eds.), Dialogforschung, Düsseldorf, 53–88. Kallmeyer, W. 1978 Fokuswechsel und Fokussierungen als Aktivitäten der Gesprächskonstitution. In: R. Meyer-Hermann (ed.), Sprechen – Handeln – Interaktion, Tübingen, 179–250. Kay, P. 1997 Words and the Grammar of Context, Stanford. 2004 Pragmatic aspects of grammatical constructions. In: Horn, L. & G. Ward, (eds.), Handbook of Pragmatics, Oxford, 675–700. Langacker, R. 2000 A dynamic usage-based model. In: Barlow, M. & S. Kemmer (eds.), Usage-based Models of Language, Stanford, 1–60. 2001 Discourse in cognitive grammar. Cognitive Linguistics 12, 2, 143–188. Levinson, S. C. 2000 Presumptive Meanings. The Theory of Generalized Conversational Implicature, Cambridge, MA.

126

Arnulf Deppermann

Linell, P. 2009 Rethinking Language, Mind, and World Dialogically, Charlotte, NC. Lyons, J. 1995 Linguistic Semantics, Cambridge, MA. Mazeland, H. & M. Huiskes 2001 Dutch ‘but’ as a sequential conjunction. Its use as a resumption marker. In: Selting, M. & E. Couper-Kuhlen (eds.), Studies in Interactional Linguistics, Amsterdam, 141–169. Michaelis, L. A. 2005 Entity and event coercion in a symbolic theory of syntax. In: Oestman, J.-O. & M. Fried, (eds.), Construction Grammar(s): Cognitive Grounding and Theoretical Extensions, Amsterdam, 45–87. Norén, K. & P. Linell 2007. Meaning potentials and the interaction between lexis and contexts: some empirical observations. Pragmatics 17, 387–416. Pekarek Doehler, S. & G. Müller 2006 Linksherausstellungen im Handlungsvollzug der Auflistung: Probleme formaler und funktionaler Abgrenzung im Französischen. In: A. Deppermann, R. Fiehler & T. Spranz-Fogasy, (eds.), Grammatik und Interaktion, 245–278. Radolfzell. http://www.verlag-gespraechsforschung.de/2006/deppermann. htm (24 February, 2009). Pustejovsky, J. 1995 The Generative Lexicon, Cambridge/London. Schegloff, E. A. 1984 On some questions and ambiguities in conversation. In: Atkinson, J. M. & J. Heritage (eds.), Structures of Social Action, Cambridge, 28–50. Schiffrin, D. 1988 Discourse Markers, Cambridge. 2007 Gesprochenes Deutsch, Berlin. Selting, M., Auer, P., Barden, B., Bergmann, J. Couper-Kuhlen, E., Günthner, S., Meier, C., Quasthoff, U., Schlobinski, P. & Uhmann, S. 1998 Gesprächsanalytisches Transkriptionssystem (GAT). Linguistische Berichte 173, 91–122. Stefanowitsch, A. & S. Th. Gries 2003 Collostructions: investigating the interaction of words and constructions. International Journal of Corpus Linguistics 8, 2, 209–243. Stein, S. 2003 Textgliederung, Berlin. ten Have, P. 2007 Doing Conversation Analysis, London. Thompson, S. A. 2002 ‘Object complements’ and conversation: towards a realistic account. Studies in Language 26, 1, 125–164. Wittgenstein, L. 1953 Philosophical Investigations, Oxford. Zifonun, G., L. Hoffmann & B. Strecker 1997 Grammatik der deutschen Sprache, 3, Berlin/New York.

Online changes in syntactic gestalts in spoken German

127

Wolfgang Imo

Online changes in syntactic gestalts in spoken German Or: do garden path sentences exist in everyday conversation?*

1

Garden path structures and spoken language

So-called garden path sentences have long been a favourite phenomenon of generative, experimental and cognitive linguistics. Most of the work on garden path structures has been done on the English language, which – due to its more or less rigid verb-second structure and lack of morphosyntactic markings – offers significantly fewer opportunities to project a syntactic gestalt and a concurrent semantic and pragmatic structure than, for example, German. As garden path sentences have usually been used as analytical tools (they are a “testing instrument in psycholinguistic research in the process of understanding of texts”; (Glück 2000: 229; my translation)) and have not been analysed as actual phenomena of spoken or written language, definitions of these structures vary considerably, depending on the aims the researchers have when they use garden path sentences as demonstration tools for syntactic or psycholinguistic theories. Pritchett (1988), for example, uses garden path sentences to determine which explanations in the context of a generative approach – namely, the application of theta criteria and rules – can best explain how these structures could be parsed by language users and why they cause problems; for this purpose, he only accepts a restricted set of locally ambiguous sentences as candidates. While he includes sentences such as “The boat floated down the river sank”, he excludes “I knew the man hated me passionately” on the grounds that the processing difficulties for the latter sentence are not as grave as for the first sentence. His explanation for the processing differences is the “Theta-Reanalysis-Constraint: Syntactic reanalysis which reinterprets a theta-marked constituent as outside of its current theta-domain is costly” (Pritchett 1988: 545). The garden path sentences in Pritchett’s analysis are used to illustrate the workings of * I wish to thank Peter Auer for his helpful comments on this paper and Elin Arbin for her corrections. All remaining errors are mine.

128

Wolfgang Imo

generative theory and his analysis is thus restricted to those structures which are useful for such a purpose. Ferreira et al. (2001) have a different research agenda. They work experimentally to find out how garden path sentences are understood, i.e. whether a full or partial reanalysis of the sentence meaning takes place (Ferreira et al. 2001: 5). Because the question here is not one of testing syntactic explanations, the range of phenomena that count as garden path sentences is larger and includes any kind of sentence with ambiguities that are resolved over time.1 Auer (2007: 104; my translation) also opts for a wider perspective by treating garden path structures merely as a special case of temporally unresolved ambiguity: “Problems are caused by those structures which stay ambiguous for a long time and do not offer any predictions about their further syntactic development. The above-mentioned garden path sentences are an extreme case of this”. This broader view is also endorsed by Glück (2000: 229; my translation), where a garden path sentence is defined as a “clause with a lexical or structural ambiguity that first is disambiguated into one reading, then into another. For example, Der Fahrer trat auf die Bremse, weil sie ihn gestochen hatte” (‘The driver stepped on the brake/horsefly because it stung him’). The garden path effect is caused by the double meaning of Bremse (‘brake’ and ‘horsefly’). This is an extremely open approach, as the problems in processing this sentence are not caused by syntactic ambiguity but by the fact that the word Bremse has two unrelated meanings which create a more or less comic effect. A similarly absurd English example is “He stepped on the fly because he had put on his trousers the wrong way up”. Such a sentence can only work as a garden path sentence out of context. The definite articles in both sentences imply the existence of a given context and, due to the lack of further information, a most likely scenario is selected: Auf die Bremse treten (‘to step on the brake’) can almost be called a complex verb on its own (Schumacher et al. 2004: 736; list this phrase as the first entry of the verb treten auf ‘to step on’). Stepping on the fly of one’s trousers is something that does not happen often – and if it does happen it is usually not relevant enough to be verbalized, let alone noticed (presumably, one would report that one stepped on one’s trousers, not on the fly of one’s trousers). Hopper (1997: 245) sarcastically proposes a special communicative genre for invented sentences used in linguistic works, suggesting the following features of this genre: Yet the alternative seems to be to regard the isolated sentence elicited in experiments and presented in grammatical work as a genre in itself. The characteristics 1

See also Gunnar (2005: 8) about garden path sentences: “The characteristic feature of this kind of sentence is temporal ambiguity.”

Online changes in syntactic gestalts in spoken German

129

of this genre would be an absence of involvement and embellishment, intrinsic newness of topics, and a discourse context that had the immanent property of always being absent. It would be a bizarre genre, and the idea that it was fundamental to all other genres would be even stranger.

Agreeing with Hopper’s view on the absurdity of isolated sentences, my aim in this paper is to look at garden path structures in spoken everyday language, where context is usually given. Ambiguities of the type illustrated by Glück almost never occur in natural conversation: When a German person is ordered by his or her co-driver to step on the brake (Bremse), he or she is not puzzled and will not look for a horsefly on the floor of the car. Lexical ambiguities that become a problem in situated interaction are rare, and if they occur, they are often produced intentionally for a comic effect (compare example 1 below). Instead, if there is ambiguity, it is usually of a syntactic kind, for example, when an intransitive verb is reanalysed as a transitive one after an object has been added during the production of an utterance. Not only is lexical ambiguity a rare phenomenon in spoken German,2 structural ambiguity also occurs very rarely. While “context” is a variable which greatly reduces the garden path effects of both lexical and structural ambiguity, structural ambiguity is further decreased by prosody, which helps to group phrases together and thus shows us what “belongs together.” This results in the dilemma that while garden path sentences are a very interesting phenomenon, the “nicest” and most illustrative examples are restricted to context-free settings that only occur in linguists’ imagination and in experimental settings.

2

A new epistemology for linguistics

If garden path sentences are so peripheral, then why do I make them the topic of a paper? The reason is that they are one of the most obvious and extreme examples of patterns that are the result of the process-related character of language. For a long time, in the wake of structuralism, sentences were treated as hierarchical phrase-structures that were viewed as complete 2

I am referring here to unintended fundamental ambiguity that results in a word having several meanings and causing on-line processing problems, not to the fact that many (if not most) words used in talk-in-interaction have fuzzy meanings that need to be fine-tuned or negotiated in what Deppermann (2007: 304; my translation) calls the interactional accomplishment of meaning: “Local constitution of meaning thus emerges as a process that is connected to routines of using an expression as well as to habitualized meaning potentials and that modifies and tailors those routines and potentials in regard to the locally relevant factual, referential, pragmatic and evaluative contexts”.

130

Wolfgang Imo

units. Only after research in spoken language had become established did temporal linearity come to the fore. The simplification that language structure consists of neatly demarcated categories and, consequently, that language is completely analysable, can no longer be upheld. Von Foerster (1991: 71) highlights the problem of “non-triviality” which researchers in social sciences as well as in the humanities have to cope with: When asked, all my friends consider themselves to be like non-trivial machines, and some of them think likewise of others. These friends and all the others who populate the world create the most fundamental epistemological problem, because the world, seen as a large non-trivial machine, is thus history dependent, analytically indeterminable, and unpredictable. How shall we go about it? I can see three strategies that are currently applied to alleviate this situation: ignore the problem; trivialize the world; develop an epistemology of non-triviality.

Since the 1960s, when conversation analysts started to show that spoken language is highly structured, the quest for such a new epistemology has been going on, resulting in some closely related approaches which all share the aim of highlighting the production aspect of language. This means that previously neglected aspects of language – such as co-production and contextualization (prosody, facial expressions, gestures, world-knowledge, frames etc.) – and, most important for aspects of grammar: time – have moved into the centre of attention.3 A universal, overarching, epistemological concept that captures all of these aspects and tries to find a way out of the “crisis of representation” (Hopper 1992) in linguistics is Linell’s (1998, 2008) approach of dialogism. While acknowledging the linguists’ need to “freeze” their objects of study, Linell nevertheless calls for new expressions, conceptualizations and frameworks that can cope with the fact that interactants are always “engaged in building situated discourse” (Linell 1998: 84): “To describe talk-in-interaction one needs a vocabulary based on dynamic rather than static concepts. […] At the same time, it is not an easy task for analysts to be consistent in applying a dynamic perspective; we often slip back into using static terms and concepts” (Linell 1998: 85). In his overview of the aspects of spoken (or, rather, dialogically situated) language, he therefore lists a wide selection of relevant aspects that must somehow be taken into account for developing non-static terms and concepts. 3

Hopper (2005: 40) even goes as far as to state that “the rules and units devised by linguists are in the end simply what is left of language when the time dimension has been removed.” See also Hopper (1992). Fanselow (2002: 250), too, states that local ambiguities (lokale Mehrdeutigkeiten) which are resolved while the production of syntactic structure is going on, play an important role for any theory of human language processing.

Online changes in syntactic gestalts in spoken German

131

Another attempt at building a coherent research programme and a theory for talk-in-interaction analysis is Interactional Linguistics as proposed by Selting and Couper-Kuhlen (2000). One of the central tenets of Interactional Linguistics is that linguistic structures are tailored to fit the organization of interaction. As language is analysed “in use,” aspects such as sequentiality, flexibility of syntactic units, possibilities of co-production and the emergence of syntax have moved to the centre of attention. While approaches that can be grouped under the heading of Interactional Linguistics are interested in any aspect of language-in-interaction, there is also a specialized concept that is geared primarily toward syntax-in-interaction: the online syntax (Auer 2000, 2006, 2007b, 2008). Auer stresses the process-related (in his terms, the incremental or online) character of any syntax that is used in real-time talk.4 Such a syntax cannot be conceived of in a static way (as structuralist syntactic trees suggest), but instead must be tailored to the temporality of its production: Incremental syntax describes the ongoing projections of the further proceeding of the emergent syntactic structure which allow the listeners to process the emerging utterance without delay. Syntactic projections rely on syntactic “gestalts” which, as soon as they are identified, must be closed according to the gestalt-psychological principle of “good continuation” by the production of a more or less precisely predictable closing structure (Auer 2007b: 97, my translation).

This approach to syntax explains why garden path structures are a possible feature of spoken language: The recipients of an utterance recognize a possible gestalt (or construction) more or less early during its production. When it turns out that instead of the expected construction another one is produced, the feeling of “being led up the garden path” occurs. When one takes an “online” perspective, garden path structures should be expected quite often, as syntax evolves (emerges) in a linear fashion and rightward expansion is thus a natural means of coupling syntax to the requirements of turn-taking or, rather, turn-keeping (Auer 1992). This naturally incrementing structure of syntax may explain why although structural and semantic ambiguity do occur in spoken German, there are not many instances where strong – or even absurd – garden path sentences can be found in everyday conversations. 4

Or, as Hausendorf puts it: “As analyses of spoken interactions in the end (also) aim at the reconstruction of (linguistic or linguistically constituted) structures, the question rises as to the quality of these structures, because they are obviously not the result of an abstraction from process-relatedness and temporality but, quite the contrary, process-relatedness is the constituting factor for these structures. Therefore, the task is to describe process itself as structure” (Hausendorf 2007: 14; my translation).

132

Wolfgang Imo

A further reason is that the relatively strong morphological markings in German disambiguate utterances quite early in their production and prevents major ambiguities: “Another indicator for that is that so-called garden path sentences […] whose syntactic processing is especially difficult, are more common in English than in German” (Auer 2007b: 99; my translation). The result is that when garden path sentences are illustrated on the basis of German, artificial and often implausible examples (like the one taken from Glück 2000 discussed above) are used: “The German psycholinguistic textbooks therefore usually rely on fairly artificial examples […] to illustrate the phenomenon” (Auer 2007b: 99; my translation).

3

A “nice” case of a garden path sentence

My aim is to look at garden path structures in everyday spoken language, i.e. at structures that emerge “online” and show an ambiguity at first, which is later resolved. The search for such structures in a corpus of over 30 hours of spoken German taken from transcripts of the first season of Big Brother, radio phone-in shows, and informal conversations between friends and family members did not yield many examples. Obvious garden path sentences, i.e. those that contain a strong ambiguity whose resolution over time is noticeable, seem to be restricted to the field of comedy. Accordingly, my first example is taken not from everyday conversation but from a well-known German comedy film: Otto – der Film. The main character, Otto, has pecuniary troubles when he accidentally saves the daughter of a rich family. He is invited to the family estate to receive a present and arrives there in high hopes of getting money or at least something valuable so that he can pay his debts. When he gets the present, it turns out to be an old, dusty bottle of wine in a basket. Otto is shocked, as he does not know that old wine can be very expensive, and after searching the basket for the “real” present he picks up the bottle and says:5 Example (1) Otto: I can’t accept that 1 2 3

5

Otto

das kann ich doch nicht ANnehmen; ‘I can’t accept that’ ich kann doch nicht ANnehmen; ‘I can’t accept’ dass das schon ALles sein soll. ‘that that is all.’

All data were transcribed according to the standards of the conversation analytic transcription system GAT (Selting et al. 1997).

Online changes in syntactic gestalts in spoken German

133

The comic effect is due to the fact that annehmen (like its English counterpart ‘accept’) can have two readings depending on the content of the direct object. If – on a semantic level – it refers to some object that can be received, annehmen roughly means ‘to take.’ If the direct object refers to something that is a concept or idea, the verb is translated as ‘to believe’. Example 1 is of course a very special case, as it can only work in certain situations and only if one uses a correlate (e.g. das/‘that’) in the object position that does not reveal its semantic content. The first reading (‘I can’t accept the bottle of wine’) occurs because the viewers of the film infer from the context that Otto is referring to the bottle of wine by ‘that’ (line 1); in the second reading the subordinate clause (‘that that is all’, line 2) supplies the semantic content of ‘that’ and changes the meaning of annehmen (‘accept’). Examples taken from comedy are problematic, of course, because they do not usually occur in informal everyday talk – in other words, they are consciously created and therefore artificial. Nevertheless, such examples should not be ignored entirely, as they can provide direction as to where to look for likely candidates of garden path structures, and how they work: As regards unexpected or bizarre uses of language, for example […] in certain artistic genres, these should not be immediately dismissed as deviations from the norms (and hence as linguistically uninteresting). Since, after all, people do indeed often make some sense of unusual language use, this must be due to the exploitation and expression of some aspects of the potentialities of the system, in combination with various contextual factors (Linell 2005: 58).

Can we find similar structures in ordinary, everyday situations, too? Indeed we can, but with an important qualification: Naturally occurring garden path structures almost never cause any interactional processing troubles, i.e. they are neither funny nor strange. The “problems” in these cases lie only in the eyes of the linguist, i.e. in the eyes of someone trying to classify utterances into categories. Reasons for this and possible consequences will be discussed after the analysis of a range of naturally occurring garden path structures in German. Interestingly, all of the garden path structures in German everyday conversation are related to turn continuation or incrementation. Therefore, some work concerning these areas of research will be discussed.

4

Increments – a special case of garden path structures in spoken language?

Increments – the term was coined by Schegloff (1996: 59) to describe a certain type of turn continuation structures – have received much attention over the past years. Only recently, a special issue of Pragmatics, edited by Couper-

134

Wolfgang Imo

Kuhlen and Ono (2007), was dedicated to the topic of “turn continuation.” Turn continuation is of course not to be equated with garden path structures. Quite often, additional material is merely added to a turn construction unit without coercing any syntactic or semantic changes to the utterance before. Sometimes, though, such a change can occur. Couper-Kuhlen and Ono (2007: 515) list the following cases of what is possible after a full turn constructional unit (TCU) has been delivered: One option is to start a new, independent unit (“new TCU”) that does not rely on the previous one at all. The second option would be a “free constituent” that consists of material which relies on the previous TCU semantically and pragmatically, but not syntactically. The third option, which is of interest for my analysis, is a “TCU continuation.” Here “material which is one way or another syntactically and semantically dependent on the prior unit” is used to continue “the prior action” (Couper-Kuhlen and Ono 2007: 515). TCU continuation can be subdivided into four different classes:6 – non-add-ons: A non-add-on is a TCU continuation that “follows a strongly marked syntactic closure, and is therefore ‘out of place’, but shows no prosodic break with its host” (Couper-Kuhlen and Ono 2007: 515). It appears to be integrated into the previous TCU. This type of continuation is often used to deliver additional information and corresponds to “‘right dislocations’ without a prosodic break” (Couper-Kuhlen and Ono 2007: 519). Garden path effects are not usually observed in the context of non-add-ons as they follow a “true” syntactic closure which is not reinterpreted afterwards. – replacements: Replacements are closely related to repairs in that they overwrite part of the previous utterance in order to replace it with new material. No garden path effect can occur here, either. – glue-ons: Both glue-ons and add-ons are preceded by “a prosodic break – be it in terms of pitch, loudness, tempo/rhythm or pause” (CouperKuhlen and Ono 2007: 514). Here, additional material is delivered which belongs to the previous TCU. Glue-ons actually fit the end of the previous TCU syntactically and thus furnish “new elements which are retrospectively seen to be part of the prior unit” (Couper-Kuhlen and Ono 2007: 515). In that case, garden path effects can occur. – insertables: Insertables are very similar to glue-ons in that they retrospectively change the syntactic gestalt of the previous unit. Unlike glue-ons, 6

See also Auer (1996) for the typological grouping of increments. In Couper-Kuhlen and Ono’s (2007: 515) terminology, glue-ons and insertables together are called “increments”, and increments and replacements together are called “add-ons”.

Online changes in syntactic gestalts in spoken German

135

however, they do not “fit the end of the prior unit but belong, canonically speaking, somewhere within it” (Couper-Kuhlen and Ono 2007: 515). Here, too, garden path effects can occur. For the further analysis, glue-ons and insertables will have to be taken into account as providing possible candidates for garden path structures. 4.1 Garden path structures involving lexical ambiguity In the following chapters, I will analyse examples of those structures in spoken German that come closest to garden path sentences. The first example is taken from a conversation between family members in Brandenburg.7 Example (2) Brandenburg: briefs 357

T

358 359 360 361 362

On

363 364 365 366

T

f f

aber der SCHLÜPFer passt mir nich. ‘but the briefs don’t fit me.’ sacht se der is so SCHICK, ‘she says it is so fashionable,’ ick würd dir den so jerne JEben; ‘I would like to give it to you;’ ob ick n ‘whether I want to ’ war so NIEDlich. ‘((that)) was so cute.’ wat hast du jeSACHT? ‘what did you say?’ na wir können drüber REden; ‘well we can talk about it;’ [(wat ZAHLST n;)] ‘[(what do you pay;)]’ [ich sach na ] mal SEhen, ‘[I say well let’s see,]’ ick hab gesacht na mal sEhen ob er MIR überhaupt passt. ‘I said well let’s see whether it even fits me.’

The daughter (T) tells her uncle (On) about a girlfriend of hers who offered her a pair of briefs as a present that she (her friend) cannot wear. She comments on the offer as being “cute” (line 361). Her uncle then asks her how she reacted (the offer of underwear touches a sensitive area in terms of face, as is indicated by the laughter and the expression “cute”), to which the daughter first answers that she replied with “let’s see” (line 365) and then “let’s see whether it even fits me” (line 366). The expression mal sehen can be translated either as “let’s see” or as “let’s wait and see” and is a lexicalized 7

An analysis of this excerpt from a construction grammar point of view can be found in Imo (2007b).

136

Wolfgang Imo

idiom: Schumacher et al. (2004: 657; my translation) list “Mal sehen!” as an idiomatic “request to wait for an uncertain result.” In line 366, on the other hand, mal SEhen is followed by a subordinate clause. This activates the core meaning of the verb sehen (‘to see’): The daughter can actually try on the briefs and see whether they fit or not. The second utterance is much less one of “wait and see” – implying a refusal of the offer – but rather “see if it fits” – implying an acceptance if the briefs really fit. Just as in example 1, the meaning of a verb is changed because of a change in its subcategorization frame. Following Couper-Kuhlen and Ono (2007), the phenomenon in line 366 is a mixture between a restructuring glue-on that changes the semantics of the verb sehen and a replacement. Nevertheless, the change in the verb’s semantics does not present any visible problems for the interactants. Only for the analyst does this example present a problem: Did the speaker first utter the idiomatic expression mal SEhen (‘let’s see’) and then a full sentence mal sEhen ob er MIR überhaupt passt (‘let’s see whether it even fits me’) or did she first utter a broken-off fragment which was then completed by the second utterance? As the meanings of both structures are closely related, this question cannot be answered definitively. Example 2 is not an intentionally produced garden path structure (as example 1), but results from the fact that speaking is a process, not a delivery of fixed units: Speakers do no speak out of their heads, on the basis of preplanned cognitive structures that exist prior to verbalization. […] Speakers talk not only in order to be understood by their interlocutors, but also in order to understand what they themselves say and think. The speaker is also a recipient of his own utterance. He will often only gradually realise new meanings, make new associations, and see novel aspects or additional problems connected to his topics, and this often happens in and through the very process of verbalization (Linell 1998: 94).

The reason for the interactionally unproblematic nature of example 2 is that language users accept ambiguous structures with overlapping interpretations as a matter of course. This is in accordance with the result of a series of experiments with invented garden path sentences in English by Ferreira et al. (2001: 16), who showed that “comprehenders are not disturbed to end up in [a] state of disequilibrum,” i.e. they are not disconcerted by the double assignment of meanings to an utterance. The function of the recycling of utterances, which is accompanied by changes of meaning, has received more and more attention in recent years. In his concept of an emergent grammar, Hopper (2005: 37) sees “previous utterances” as important sources for interlocutors that “serve as rough models for current utterances,” and Linell (2005: 72) states that the “re-use of

Online changes in syntactic gestalts in spoken German

137

others’ and one’s own words is a natural phenomenon in a conversation. […] All in all, ‘repetitions’ are not plain copyings; new utterances convey new meanings […]”. A gradual shift and stepwise production of meaning is therefore something that is typical for talk-in-interaction, which accounts for the fact that interactants show no signs of trouble in processing recyclings, providing the change is gradual enough. The linguistic description of such utterances, on the other hand, is problematic, as one must cope with structural ambiguity – something that is counter to traditional views of semantically and functionally well-defined categories and constructions. 4.2 Garden path structures involving syntactic ambiguity While examples 1 and 2 deal with verbs that have different meanings depending on the type of complements they take, the following examples are more concerned with ambiguous syntactic structures, i.e. with different possible “groupings” of elements (Example 3) or different constructions (Examples 4 and 5). Example (3) Domian: we can talk 006

Do

007 008 009

f

010

f

011

f

012

f

013

f

014 015 016 017

MEIne obligatorische frAge an EUch, ‘my obligatory question to you,’ über was habt IHR lUst (.) .hhh äh mIt mir- (.) ‘what do you want to (.) .hhh ehm with me- (.)’ EUch mit mir zu unterHALten, ‘to talk about with me,’ eGA:L- (.) ‘no matter- (.)’ über welches THEma, ‘which topic,’ wir können REden, ‘we can talk,’ wenn ihr WOLLT, ‘if you like,’ RUFT an, ‘call me,’ telefonnUmmer NULL acht hUndert (.) zwei zwei nUll, ‘telephone number o eight hundred (.) two two o,’ FÜNFzig FÜNFzig, (.) ‘fifty fifty, (.)’ oder FAXT unter null Achthundert, ‘or send a fax at o eight hundred,’ zwei zwei null fünfzig EINundfünfzig, ‘two two o fifty fifty-one,’

138

Wolfgang Imo

Example 3 is taken from the radio phone-in show Domian. Domian is a popular radio host who invites his audience to call in and talk about whatever they want to. In the example, the show has just started and Domian asks his audience to phone in. From lines 009 to 017, he offers the necessary information about the programme phrase by phrase in isolated, rising intonation contours. While eGAL (‘no matter’) in line 009 is still quite easy to process as a projecting element – it demands something that “doesn’t matter” as a complement, which is duly delivered by über welches THEma (‘which topic’) – lines 011 to 013 are syntactically ambiguous. The first interpretation which is activated over time combines lines 011 and 012: wir können REden, WENN ihr wOllt (‘we can talk if you like’), i.e. the conditional clause seems to refer to the statement wir können REden (‘we can talk’), which it qualifies. As Domian goes on, though, another possible phrase structure emerges: WENN ihr wOllt, RUFT an (‘if you like, call me’). In this interpretation, the wenn-clause is not post-positioned but pre-positioned and qualifies the invitation RUFT an (‘call me’ in line 013). The utterance wir können REden (‘we can talk’) has to be (at least partially) reprocessed as the free-standing motto of the radio talk-show. This example provides another good illustration of how spoken language works: “Spoken utterances are transient phenomena, distributed in real time, and produced and received over time in an incremental fashion” (Linell 2005: 55). The incremental fashion of production and reception – i.e. the “stitching together” (Hopper 1999) of fragments of speech – is particularly salient in example 3, because prosodic means are not used to disambiguate the phrases. Nevertheless, the piecemeal production of utterances does not cause any significant problems because the wenn-clause, for example, can actually have two possible scopes which are not mutually exclusive. As experiments of garden path sentence processing have shown, “the incorrect representation of the syntactic structure of a garden-path sentence is not fully inhibited and remains active to a certain extent even after analysis of the respective sentence is finished” (Jacob 2005: 39). A double activation of a syntactic structure can cause problems if it results in two very obviously divergent meanings that clash strongly. If – as in example 3 – both meanings are possible and plausible, there is no reason for the “naïve partners-in-talk” (Bühler 1982/1934: 102; my translation) to think twice about such a structure. It is only for a new epistemology of spoken language and for a reconceptualization of categories and sentence (or utterance) structures that these findings are important. The following two examples involve constructions with the particle (or adverb) so. Both examples are taken from the Big Brother transcripts.

Online changes in syntactic gestalts in spoken German

139

Example (4) Big Brother: depressed 452

Jhn

453 454 455 456 457 458

gestern abend ham wa so weit (.) ähm überSTANden? ‘so far we survived yesterday evening?’ also; hh .hh ‘well; hh .hh’ (1.0) f ick war dann doch nich SO:: nIEderjeschlagen; (.) ‘I wasn’t so/as depressed; (.)’ wie ick erst DACHte? ‘as I thought at first?’ also ick hab det janz jut so allet verKRAFtet, (0.5) ‘well I coped with all that quite well, (0.5)’ .hh dass ick jetzt noch HIER bin? ‘.hh that I am still here?’

Example (5) Big Brother: learning by heart 179 180 181

182 183

189 190

191 192

193 194 195 196

Adr

ich bin wirklich GANZ arg schlecht darin. ‘I’m really very bad at that.’ ich kann BESser- (.) ‘I can better- (.)’ gib mir n STICHwort und ich zAUber dir irgendwas daraus. ‘give me a catchword and I will make something out of it for you.’ aber lass mich nix (.) NICHTS AUswendig lernen; ‘but don’t have me learn anything by heart;’ ich KANN es nich. ‘I can’t do it.’ ((...)) Vero aber dann kannste es eigentlich MISCHen, ‘but then you should be able to mix it,’ f weil i- ich kann AUCH nich auswendig lernen dass ich alles LES und wortwörtlich wIEdergeb; ‘because I I also can’t learn by heart that I read everything and reproduce it word for word;’ ne? ‘can I?’ aber wenn du sagst du kannst aus einem wort viel MACHen, (.) ‘but if you say that you can make a lot out of one word, (.)’ denn (.) das mach ICH nämlich eigentlich. ‘because (.) that’s what I do.’ ich les den TEXT, ‘I read the text,’ Adr [(( was dabei))] ( ) Vero [und dann sag ich mal] SIEBzig prozent geb ich de‘and then let’s say seventy precent I re-’

140 197

198

Wolfgang Imo das WICHtige was sein muss wIEder, ‘the most important things that have to be there I reproduce,’ aber den rEst denk ich mir immer AUS. ‘but I invent the rest.’

In example 4, John is sitting facing the camera in a closed video cabin delivering his obligatory weekly statement. He talks about the upheaval within the Big Brother set the evening before, when one of the inhabitants was voted out of the house and reports that the other inhabitants have now come to terms with the new situation (lines 452f.). After the general statement about the feelings of all the Big Brother contestants, John goes on to talk about himself. He produces an utterance with a slowly falling intonation and a short pause. This utterance can stand alone and corresponds to a regular pattern in German, which Auer (2006: 312) calls “emphatic SO-construction.” This is a construction that consists of a noun phrase, a copula, the word so and an evaluating adjective. So is usually prosodically marked. In the example above, it is both stressed and lengthened. According to Auer (2006: 309–310, my translation) typical examples of the emphatic so-construction are s=war SO geil (‘it was SO great’), s=ist mir SO WURSCHT (‘it is SO all the same to me’), ich war SO WÜtend auf mich selbst (‘I was SO angry with myself ’) etc. In terms of its “outer syntax,”8 the emphatic so-construction usually requires contexts which include something that can be evaluated or assessed. These contexts are then referred to by a pronoun. In the case of example 4, such a context is given in that the Big Brother contestants had to cope with stress. So, until the end of line 455, the utterance could be interpreted as an unproblematic instance of an emphatic so-construction. As John speaks on, however, it becomes clear that another construction emerges: John delivers a second part, activating a pattern of so … wie (‘as … as’). This is a two-part construction which Weinrich (2003: 788; my translation) calls “Vergleichs-Junktion mit einem so/wie-Korrelat” (‘comparative junction with an as/as-correlate’). The construction is similar to the emphatic so-construction in that it, too, typically consists of a noun phrase, a copula, the word so and an adjective. The difference is that this structure only forms the first part of the complete con8

“In addition to the inner syntax of grammatical constructions […], there are many constructions that have a conventionalised outer syntax. The outer syntax specifies relations between the construction and what may precede and/or follow it in the sequence of talk or text. That is, constructions often require, presuppose or invite certain properties in their prior utterances and possible next utterances in the local sequences in which they occur. The boundary between the grammar and the ‘information structure’ of texts is partly fuzzy” (Linell 2005: 72).

Online changes in syntactic gestalts in spoken German

141

struction, which is then followed by wie (which together with so forms the so/wie-correlate construction) and a noun phrase or sentence expressing the correlate. If there is no context that allows an evaluation or assessment, this construction would likely be expected from the beginning. In the case of example 4, however, both constructions are possible and it is only the ongoing utterance that makes clear which construction is activated. Constructions involving so seem to be particularly prone to this kind of ambiguity. Auer (2006: 308) shows that the emphatic so-construction can be expanded into what he calls a “gradual-consecutive” so-construction, which is a third alternative candidate for online emergence involving so. Auer (2006: 308; my translation) notes: In order to decide whether an emergent construction orients towards one or another constructional pattern, the recipient has to take into account the conversational information about the context of the one construction, only to be forced to conclude under certain circumstances that it actually is the other construction which is emerging.

This also holds true for example 4. At first glance, example 5 seems to be different, as it does not contain the particle so. The Big Brother participants talk about the problems they have learning things by heart. One participant, Andrea, says that she is very good at improvising but hopeless at learning things by heart. To this, Verona answers that this is a good basis, as she can mix learning by heart and improvising (line 189). She then states that she, too, cannot learn by heart (line 190). This ties in exactly with what Andrea said before (aber lass mich nix (.) NICHTS AUswendig lernen; ich KANN es nich. ‘but don’t make me learn anything by heart; I can’t do it.’) and implies that learning by heart is something that they both cannot do at all. When Verona goes on, though, still in line 190, she transforms this absolute statement into a qualified one, activating a pattern one would expect to look like this: weil ich kann auch nicht so (i.e. auf diese Weise) auswendig lernen, dass ich alles les und wortwörtlich wiedergeb ‘because I also cannot learn by heart in such a fashion that I read everything and then reproduce it word for word’ Auer (2006: 307) calls this pattern a “modal-consecutive so-construction.” The so projects a modification of the verb which is delivered in the following clause. Thus, in example 5, again, two constructions compete. What makes this example interesting is the fact that the projecting so is not produced, leaving the recipients with no warning that the sentence goes on and changes

142

Wolfgang Imo

from an absolute statement (implying I can’t learn things by heart) to a qualified or modified one (implying I can’t learn things by heart in a word-for-word fashion but I can remember enough to get by). As soon as the subordinating conjunction dass is produced, the interactants must rework their interpretation to reconstruct the sentence as related to the structure of modal-consecutive so-constructions. Example 6 from Big Brother illustrates a case that is similar to that of examples 4 and 5, as the status of a correlate is unclear and is restructured by a following glue-on: Example (6) Big Brother: two worlds 440

Adr

441

Jhn

442 443 444

Adr

445

Jhn

446

Adr

447 448 449

Jhn

450 451 452 453 454

Adr

455

Jhn

456 457

und ER arbeitet AUch noch? ‘and he still works, too?’ er is SELBSTständig. ‘he has his own business.’ also der hat so ne (0.5) SICHerheitsfirma? ‘well he has a kind of (0.5) security firm?’ also die so sIcherheitstechnik EINbauen. ‘well which installs like security equipment.’ mhm? ‘mhm?’ die machen unter anderem jetzt gerade bei günther JAUCH det haus. ‘among other things, they are just working on Günther Jauch’s house.’ ach ECHT? ‘oh really?’ aber dann geht’s denen aber doch GANZ gut. ‘but then they are doing quite well.’ [oder?] ‘aren’t they?’ [mhm. ] (.) ‘mhm. (.)’ JAja. ‘yesyes.’ DEnen jehts gut. ‘they are doing well.’ die ham‘they have-’ det is ja det wat zum ANfang bei uns für problEme hatte. ‘that is what created problems for us at the beginning.’ mhm. ‘mhm.’ also denen jings schon IMmer jut. ‘well they have always done well.’ die hatten schon immer (0.5) .h JELD, ‘they have always had (0.5) money,’ f und (.) det WAret halt. ‘and (.) that was it.’

Online changes in syntactic gestalts in spoken German 458 459 460

Adr

461 462

Jhn

143

dass da diese zwee WELten ufenander jekommen sind.= ‘that there these two worlds collided.’ [=WEIßste?] ‘ [you know?]’ [mhm, ] ‘ [mhm, ]’ (1.0) det war det GRÖßte problem; ‘that was the biggest problem;’

Before the transcript starts, John is talking about his wife, who wanted to visit an expensive and famous hairdresser’s studio. He goes on to tell Andrea that his wife’s mother would certainly accompany his wife, because she loves to spend money. Andrea then asks whether John’s father-in-law still works (line 440). The question is motivated by the fact that Andrea knows that John does not have a lot of money – she therefore assumes that the family of John’s wife is not very rich, either. After John tells her that his father-in-law owns a security firm (line 442) and even works for well-known public persons (line 445; Günther Jauch is one of Germany’s most famous TV hosts), she replies with aber dann geht’s denen aber doch GANZ gut (‘but then they are doing quite well’). The adversative conjunction aber (‘but’) as well as the two modal particles aber and doch (both untranslatable in English) signal that the information is unexpected for Andrea. John confirms her interpretation (lines 450–452.) and goes on to a new aspect – the problems he has with his in-laws (or, rather, they with him). The pronoun det (‘that’) in line 453 first refers to the utterance DEnen jehts gut. (‘they are doing well’), so the problems John and his wife had with their in-laws are treated as the result of the in-laws’ wealth. In lines 455–456, John expands on his previous utterance by stating that his in-laws have always been well off. In line 457, he produces another utterance containing the demonstrative pronoun det (‘that’). Again, the pronoun seems to be used anaphorically, i.e. it seems to point backwards. Only after he goes on in line 458 and delivers the subordinate clause does a reinterpretation become necessary: If it were not for the falling prosody in line 457, the utterance in 458 could be seen as delivering the content of the pronoun, thus turning it into a cataphoric pronoun. Line 458 calls for a reinterpretation of the utterance und (.) det WARet halt (‘and that was it’), as the subordinate clause refers to the pronoun in line 457. The online emergence of this structure seems to be over at this point: Andrea gives a response signal (mhm in line 460), and there is a tag question and a long pause by John (lines 459 and 461). But then John utters an expanded reformulation of his utterance in 457: det war det GRÖßte problem (‘that was the biggest problem’; line 462). Again, the recipients have

144

Wolfgang Imo

to reorganize their interpretation of the structure: Now the subordinate clause in line 458 could be analysed as a left-dislocation with an uptake by the backwards-referring pronoun det (‘that’) in line 462. The subordinate clause in line 458 does not fit properly into either description: The strong prosodic break after und (.) det WARet halt (‘and (.) that was it’; line 457) argues against the interpretation as a post-positioned subordinate clause, and the tag question and the pause against the interpretation as a left-dislocation. Or, alternatively, both interpretations are equally possible. What we have here, then, is a structure (a subordinate clause) that refers both back and forwards and can be said to be embedded into two larger structures at the same time. Again, the differences in terms of meaning and structure are much too low-key for the interactants to become relevant. For a theory of language, though, such structures must be accounted for. The definition of such a pattern which is the result of the temporality of speech is that of the “pivot” or “apokoinou” construction as described by Auer (1992), Norén (2007), Sandig (1973), Scheutz (1995) and Poncin (2000). As the fundamental aspect of pivot constructions, Scheutz (1995: 104) singles out the fact that “during the production of utterances within a turn, changes in the syntactic structure can be made, but it is only in retrospect that these are recognized as changes and could therefore be categorized as two different syntagmatic structures.” Norén’s corpus-based study is a good example of an analysis of apokoinou constructions (which he prefers to call a “family of methods” because their constructional status is much too open for a single construction), since it is informed by the new epistemology for linguistics (dialogism, temporality in syntax). Temporality in particular plays a great role in the production and reception of apokoinou constructions. In my data, though, there are only few examples of these patterns, and most of them are based on verbs of meaning and saying which are used as quotatives,9 such as: Example (7) Swabia: beauty 606 607 608 609

9

T

da hat‘then did-’ f da hat se GSAGT, ‘then she said,’ ja SCHÖNheit muss leiden; ‘well beauty has to suffer;’ f hat die KUH zu mir gsagt. ‘this cow said to me.’

Looking at the examples Norén analyses, one gets the impression that in Swedish, too, quotatives (as well as copulas) provide most of the apokoinou constructions.

Online changes in syntactic gestalts in spoken German

145

Here, the utterance in line 608 serves as the basis for both the projective quotative in line 607 and the backwards-looking one in line 609.10 A static approach to syntax would view this as a faulty construction. An incremental approach to syntax, however, sees apokoinou constructions not just as examples of the step-by-step production of syntactic structure; they are also – according to Norén (2007: 362) – used regularly for specific functions, such as a means to achieve “semantic-pragmatic shifts and actions.” Such a shift also occurs in example 7: T is talking about having had her ears pierced, which hurt a lot. She then quotes the woman who did the piercing, first using the neutral quotative “then she said” and then changing to the evaluating quotative “this cow said to me”. So at the heart of apokoinou constructions, too, the double process of “projecting possible continuations of the utterance and retro-constructing the impact of previous actions,” which was seen in all of the cases discussed above, is at work. A further common aspect of garden path sentences, increments and apokoinou constructions is the fact that there are no strongly diverging semantic contents that are yoked together. Scheutz (1995: 111) states that in the conversational data he analysed there are “no instances of pivot constructions in which the semantics of the initial part is essentially different from the semantics of the final part.” Thus, strong garden path effects are blocked due to the semantic closeness of the first and last part of the apokoinou construction. My last two examples are illustrations of a type of increment that is only marginally responsible for garden path effects. Both cases involve the production of what is called an “insertable” in Couper-Kuhlen and Ono’s (2007) terminology. Example (8) Big Brother: bun 154

Adr

155

Vero

156 157 158 159

Sbr

160

Vero

10

und du hast heut morgen schon schön geFRÜHstückt? ‘and you already had a nice breakfast this morning?’ hähä (.), ‘hehe (.),’ ach ja n BRÖT‘oh yes a bun-’ (breaks off in the middle of the word) das war LECker; ‘that was tasty;’ VOLL. ‘very.’ so [WARM ne? ] ‘so [warm, wasn’t it?]’ [Echt ganz] WARM; ‘[really totally ] warm;’

The example is taken from Imo (2007b: 94).

146

Wolfgang Imo

Example (9) Big Brother: cameras 1585

1586 1587

Vero

jetzt kommen wieder die ganzen KAmeras weil ich mich hier Umziehen [muss. (.)] ‘now all the cameras are coming again because I have to change my clothes. (.)’ Sbr [a ja:: ] [(LOgisch). ] ‘ah yes that’s logical.’ Vero [angerROLLT.] ‘rolling in.’

In Example 8, Andrea asks Verona, who has come to visit the Big Brother contestants as a guest, whether she has already had breakfast. Verona answers that she has and that she ate n BRÖTchen (‘a bun’), adding the evaluation das war LECKer (‘that was tasty’). After that she utters the word voll (‘very’) which in this context is associated with informal youth slang (Spiekermann and Stoltenburg 2006) and belongs to the group of intensifiers (Helbig 1999: 99). In standard German such particles include sehr, höchst, außerordentlich, ziemlich and ungewöhnlich (very, extremely, exceptionally, quite, extraordinarily), among others. All of these particles could be used here, too, in principle. The canonical position of these words is in front of the adjective they modify. Helbig (1999: 99) goes as far as to state that the position of these intensifiers is fixed and that they must always be placed in front of the adjective. Therefore, the condition Couper-Kuhlen and Ono (2007: 515) list for insertables is met, namely, that they should be placed “canonically” within the utterance, not after it. But does this structure also qualify as a garden path sentence? The assessment das war LECKer (‘that was tasty’) is expanded over time11 and transformed from a positive statement to a very emphatically positive one. Therefore, a semantic and pragmatic change is triggered. Nevertheless, I do not count these cases as garden path structures but as a construction of their own. In spoken language – and even more so in youth language – post-positioned intensifiers are often used. The position after the syntagma itself serves to highlight the intensifier, making it salient. Therefore, three structures could be claimed for das war LECKer (‘that was tasty’): a) Das war lecker. (‘That was tasty.’) positive; no emphasis b) Das war voll lecker. (‘That was very tasty.’) highly positive; emphasis c) Das war lecker. Voll. (‘That was tasty. Very.’) highly positive; very strong emphasis As the last structure is a common and sedimented one in spoken German, garden path effects cannot occur in example 8, since the underlying pro11

See also Auer (1992: 50) who analyses an example with a post-positioned attributive adjective as “a loose addition of an appositive structure.”

Online changes in syntactic gestalts in spoken German

147

cesses of a syntax expanding over time are the same. Auer (2007a), too, states that one problem with the analysis of increments or turn-continuations is that many constructions discussed under these headings are in fact standard constructions of spoken German. For the sentence Hier wird ordentlich gegessen HEUTE (‘here we will eat well today’) he states that the in terms of normative grammar misplaced post-positioned adverb heute (‘today’) might not to be analysed as a product of incrementation but as an instantiation of an “ordinary” sentence of spoken German: “It could be argued that they are constructions specific to German, i.e. that the ‘expansion’ looks like an expansion from the normative point of view of written language only” (Auer 2007a: 653). Whether such structures are indeed to be reanalysed as constructions in their own right or whether the strong emphatic effect of a postpositioned adverb is due to the break of a syntactic convention is still open to discussion. Example 9 does not qualify as a real garden path sentence either, though the underlying structure allows for a possible syntactic ambiguity. Verona undresses and comments that die ganzen KAmeras (‘all the cameras’; line 1585) focus on her to sneak a view of her naked. After Sabrina states that this is to be expected (line 1586), Verona expands her previous utterance with angeROLLT (‘rolling in’). This, again, is an insertable which has its canonical position as the right verb bracket in front of the subordinate clause in the post-field of the sentence. Thus angeROLLT (‘rolling in’) would have to be positioned between KAmeras (‘cameras’; the end of the middle field) and weil (‘because’; the beginning of the post field) in line 1585. The prosodic marking (falling intonation, micro pause) as well as the intervening utterance by Sabrina set off the angeROLLT in line 1587. Nevertheless, the restructuring is not very salient: On a syntactic level angeROLLT is displaced, but on a semantic level it does not change the sentence meaning at all: There is no difference whether cameras come or come rolling in. In their discussion of restructuring glue-ons, Couper-Kuhlen and Ono (2007: 524) state that “restructuring glue-ons are quite startling when they occur in English TCU continuation.” As an example they give “they’re doin:g (.) a: comic (.) murder, (…) pla:y,”. What Couper-Kuhlen and Ono call “startling” is the effect of a garden path sentence: The need for restructuring the sentence meaning leads to the feeling of having been led up the garden path. Quite obviously, whether a garden path effect actually occurs or not depends only on the semantics of the added material. If the restructuring process triggers no major change, the garden path effect is nonexistent. The more the unfinished and finished varieties of the utterance vary, the stronger the garden path effect.

148

5

Wolfgang Imo

Increments, garden path structures and online syntax : uncertain constellations

So far, I have shown that increments and garden path structures can both be explained in terms of Auer’s (2000, 2008) online syntax. Yet many open questions remain and several problems have yet to be solved. First, as Auer (2007a: 650) points out, “the role of syntax is unclear. While there is general agreement that increments occur after a point of syntactic closure […], there is no agreement as to whether they integrate into the structure of the host, and if so, how”. This is a problem that came up in the analysis of most of the examples discussed here: How is the second “added” utterance to be interpreted? Is it an instance of a more or less regular pattern (as in the case of post-positioned intensifiers in example 8), or an instance of a gradual syntactic restructuring? We still do not know enough about how spoken language works to answer the question of what counts as “proper” units in spoken language: turns and TCUs, clauses, intonation units, actions or any combination of these (see, for example, Couper-Kuhlen and Thompson 2005 for a discussion of the relevance of the syntactic unit clause for interaction). As long as there is no consensus about what the relevant units are and how they interact with each other, attempts to define garden path sentences are destined to fail. A second – related – problem with increments concerns the role of prosody (see Auer 1996 for a discussion of the interrelatedness of prosody and syntax in the context of turn expansions). This holds especially true for experimental and “arm-chair” work on garden path sentences where the sentences are written down without orthographic markings. If these sentences were spoken, there would hardly be any problem concerning their processability. On the contrary, the analysis of the utterances discussed here has shown that quite often – as is typical for non-add-ons – there is a prosodic break between the utterance and the increment. What is the exact effect of such a break? Can a systematic integration of intonation into the workings of grammar – as Auer (2007a: 650) claims – help clarify the concepts of increments, garden path structures and online syntactic structures in general? A third point Auer (2007a: 651) mentions is the correlation of “syntactic and action dimensions.” This not only concerns questions of the “mapping of syntax on pragmatics or the other way round” but also the role of context: The example of the German garden path sentence involving the horsefly (Glück 2000) discussed above shows that many garden paths become unproblematic main roads if the sentence is not taken out of context. It has yet to be shown whether context in English sentences, too, prevents strong garden paths structures outside comedy or linguistics.

Online changes in syntactic gestalts in spoken German

149

In spite of these problems or, rather, because of them, Auer suggests broadening research to what he calls “unit expansion”: I suggest that the matter of unit expansion in conversational speech is a much larger issue than has been suggested in most work on increments so far, and that it should be approached in a more general way. […] In short, I suggest that the identification of TCU boundaries vs. TCU expansions is a highly interpretive issue which lay and professional analysts alike cannot reduce to a dichotomic distinction. (Auer 2007a: 651).

My analysis is a contribution to the task of classifying the different constructions and patterns which are revealed by an online approach to syntax.

6

Conclusion

The analysis of some candidates for garden path structures in spoken German has shown that the “roots” of this phenomenon are actually very common in everyday interaction. Yet, all of the instances of ambiguity that get resolved over time are strikingly unspectacular and offer no problems for the interactants. Why is it so hard to find strong garden path structures and why do the ones that do come up pose no problems interactionally? The following reasons are proposed for the remarkable lack of strong garden path structures as well as for the unproblematic processing of those that do occur in spoken German: 1. The first reason is given by Auer (2007a): Unlike English, German still has a considerable amount of morphological marking that reduce the range of possible misinterpretations by giving clues as to what follows quite early on in the production of an utterance. Nevertheless, as there have been no studies looking especially at naturally occurring garden path structures in English, it is not clear whether such structures are prevented only by German morphology or if they are also avoided in spoken English. 2. Garden path sentences in which readings differ grossly on the semantic or syntactic level are very rare in everyday conversation: Comedy and linguistic experiments seem to be the “natural” locus of such structures. As the examples show, the differences in meaning or changes in syntax are often so subtle as to be only relevant for linguists and possibly not even noticed by interactants themselves. Context provides a strongly binding factor that prevents vastly diverging meanings. 3. Even for strong garden path structures it can be shown that interactants can live with ambiguity. In those cases (as, for example, in the classic sentence “The horse raced past the barn fell”), the first (“The horse raced past the barn”) and the second (“The horse, which was raced past the barn,

150

Wolfgang Imo

fell”) interpretation differ greatly. Nevertheless, participants in experiments often accept both the former interpretation – which should be overridden – and the latter interpretation at the same time.12 If that is the case with strong garden path structures, it is not surprising that interactants are even less picky when it comes to cases where the diverging meanings or syntactic structures are closely related anyway. One can safely propose the hypothesis that interactants do not realize small inconsistencies at all and have no problems with ambiguity. Norén (2007: 364) also notes that “participants never make apokoinou utterances as such the object of repair” and comes to the conclusion that this “implies strongly that it is an accepted and functional way of doing communicative action and that participants possess communicative and grammatical knowledge of the method.” While it is certainly true that apokoinou constructions can build on sedimented and recognizable structures, I would suggest that the reason for the non-problematic nature of these structures is the same thing that is responsible for the non-problematic nature of all of the examples described in this paper: They pose no problems because fuzziness and constructional ambiguity are tolerated in language anyway. This observation leads away from the small – and esoteric – field of garden path structures to areas of ambiguity, amalgams and hybridizations in general. In recent studies which look at syntax and semantics in spoken language, more and more evidence has been gathered that fuzziness is everywhere: In semantics, as Deppermann (2007) and Norén and Linell (2007) show,13 as well as in syntax, as studies by Imo (2007a,b,c), Günthner (2000, 2006, 2007, 2008), Hopper (2001) or Linell (2005)14 suggest, hybrid structures seem to form a central phenomenon of spoken language. Recent research in psycholinguistics comes to the same conclusion: Taken together, these studies present a challenge for the fundamental assumption in psycholinguistics that comprehension is based on the creation of full, accurate, and detailed representations. It appears, instead, that people work on sentences 12

13

14

Ferreira et al. (2001: 16) state that “the experiments reported here suggest that people are often satisfied with inaccurate representations based on incomplete processing of the sentence” and Gunnar (2005: 39), too, found that “the incorrect representation of the syntactic structure of a garden-path sentence is not fully inhibited and remains active to a certain extent even after analysis of the sentence is finished.” Norén and Linell (2007: 4) define the “meaning potential” of a lexical or grammatical construction as its “set of properties which together with contextual factors […] make possible all the usages and interpretations of the word or construction that language users find reasonably correct, or plainly reasonable in the actual situations of use.” Linell (2005: 217), for example, states that “the generalized linguistics are fuzzy.”

Online changes in syntactic gestalts in spoken German

151

until they reach a point where it subjectively makes sense to them and then processing may cease (Ferreira et al. 2001: 17).

If one looks at data of spoken language which is situated in real-life contexts, traces of this “pragmatic” – in the sense of economic and error-tolerant – processing of utterances can be found, too. Hopper (2001: 125–126) goes as far as to suggest that “these ‘degenerate’ data” – which includes all of the garden path structure examples presented here – are the “true substance of natural spoken language” and that the “fuller ‘canonical’ constructions should rather be seen as highly stylized cultural artifacts, amalgamations of fragments put together.” This leads us back to the quest for a new epistemology for (spoken) language. One central aspect that must be included is that of temporality (Auer 2000, 2006, 2007a,b, 2008; Hopper 1992, 1997, 1998, 2004, 2008; Linell 1998). If language is viewed under the aspects of temporality and use-in-discourse, the next question is what the categories of language are. According to Linell (2005: 43), a view of languages as “inventories of objects” is of doubtful value: Such inventories are nothing more than “cultural artefacts; they exist as the products of those decontextualising activities by linguists and other language cultivators.” This view would reduce any effort of describing the syntactic inventory of any language to a more or less irrelevant – almost artistic – game of inventing structure. Later in his study, Linell (2005: 219) revises his initial radical attitude by conceding that abstract grammar actually has a basis in language-in-use. He poses the question, “What kinds of concepts or entities would be part of a dialogical grammar?” and answers that a multitude of linguistic levels (prosody, lexicology, pragmatics, communicative genres and others) must be taken into account. One of these levels is grammar, of which the basic entities are grammatical constructions, which are abstractions from utterances or utterance types. […] However, they can also be seen as abstract structures underlying concrete, particular utterances and turns at talk, which are built on turn-construction units. […] Grammatical constructions can also be thought of as methods of constructing surface structures and utterance types according to certain patterns or ‘constructional schemas’.

The uncertain status of dialogical grammar and its new epistemology stands out clearly in this quotation, as the relevance of grammatical constructions remains open: On the one hand, they are “abstractions from utterances.” Here, the idea of cultural artefacts – of syntactic patterns as entities that are invented and constructed by linguists – is referred to. In this view, grammar is isolated from actual language-in-use. Two sentences down, Linell revises this view by calling syntactic patterns “abstract structures underlying con-

152

Wolfgang Imo

crete, particular utterances.” Grammar moves much closer to language-inuse and seems to be intertwined with utterances, not an invented set of more or less imagined forms. In a third step, Linell brings yet another aspect into play: Syntactic patterns are actually used as “methods of constructing surface structures.” This is the exact opposite of the first position: Grammar in this view is central – and prior – to language-in-use. The indecision about the actual status of grammar (whether one is talking about categories, syntactic patterns or constructions) reflects the fuzziness and fragmentedness of spoken language on a scientific meta-level. Is it possible that the “new epistemology” is a chimera, that it can only be postulated and applied in some areas of linguistic research but not in others? Quite early, Hopper (1992: 236) saw that the rejection of the old epistemology might lead not to a new one but only to some vague ideas of how to proceed in linguistics: Perhaps the rejection of totality in linguistics means that the only new paradigm will be the absence of paradigms, a kind of disciplinary anarcho-syndicalism of numerous small groups working on limited problems from a variety of perspectives, united only by a collection of practices, a set of linguistic general assumptions. At the heart of most such assumptions would be the presupposition that language owes the way it is to its temporal unfolding through the spoken interaction of historically situated individuals.

Maybe the absence of paradigms is one way of following von Foerster’s (1991: 71) call for developing an epistemology of non-triviality. The downside of this approach, however, is that one must barter unifying descriptions of spoken language for what Hopper calls “anarcho syndicalism.” An alternative which has been proposed in recent years is the theory of construction grammar (or cognitive grammar; cf. Langacker 1987)15 to describe languagein-interaction. If this approach to language-in-use proves to be successful, construction grammar might be just the theory to satisfy corpus-, computer-, interactional-, historical-, pragma-, etc. linguists alike, as Fischer and Stefanowitsch (2006, 2008) illustrate by including approaches from very different subdisciplines of linguistics. Yet, in spite of some promising results, the fundamental problems of how to integrate temporality and how to reconcile the fragmentedness and fuzziness of spoken language with the concept of categories (or constructions) still remains unsolved (cf. Imo 2007c). 15

Although construction grammar and cognitive grammar differ in many aspects, they both share some central assumptions about the structure of language. Therefore, both terms are not always differentiated clearly (e.g. in Goldberg 1998: 205: “Within the theory of Construction Grammar (also cognitive grammar) …”).

Online changes in syntactic gestalts in spoken German

153

Nevertheless, instead of giving up, I suggest we continue the search for a new paradigm. While work on isolated syntactic problems is necessary to provide empirical input, it is the big questions – How does language work? How does grammar work? How do we process language? How do language and interaction/culture interact? – that linguistics should find answers to.

References Auer, P. 1992 The neverending sentence. In: M. Kontra (ed.), Studies in Spoken Languages: English, German, Finno-Ugric, Budapest, 41–59. 1996 On the prosody and syntax of turn-continuations. In: E. Couper-Kuhlen & M. Selting (eds.), Prosody in conversation, Cambridge, 57–100. 2000 On line-Syntax – oder: Was es bedeuten könnte, die Zeitlichkeit der mündlichen Sprache ernst zu nehmen. Sprache und Literatur 85, 43–56. 2006 Construction Grammar meets Conversation: Einige Überlegungen am Beispiel von ‘so’-Konstruktionen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 291–314. 2007a Why are increments such elusive objects? An afterthought. Pragmatics 17, 4, 647–658. 2007b Syntax als Prozess. In: H. Hausendorf (ed.), Gespräch als Prozess, Tübingen, 95–124. 2008 On-line syntax: Thoughts on the temporality of spoken language. Language Sciences 31, 1, 1–13. Bühler, K. 1934/[1982] Sprachtheorie: die Darstellungsfunktion der Sprache, Stuttgart. Couper-Kuhlen, E. & T. Ono 2007 ‘Incrementing’ in conversation. A comparison of practices in English, German and Japanese. Pragmatics 17, 4, 513–552. Couper-Kuhlen, E. & S. A. Thompson 2005 The clause as a locus of grammar and interaction. In: A. Duranti & A. Brown (eds.), Language, Culture and Interaction, Cambridge, 481–505. Deppermann, A. 2007 Grammatik und Semantik aus gesprächsanalytischer Sicht, Berlin/New York. Fanselow, G. 2002 Wie ihr Gebrauch die Sprache prägt. In: S. Krämer & E. König (eds.), Gibt es eine Sprache hinter dem Sprechen? Frankfurt/Main, 229–261. Ferreira, F., K. Christianson & A. Hollingworth 2001 Misinterpretations of garden-path sentences: Implications for models of sentence processing and reanalysis. Journal of Psycholinguistic Research 30, 1, 3– 20. Fischer, K. & A. Stefanowitsch 2006 Konstruktionsgrammatik: Von der Anwendung zur Theorie, Tübingen. Glück, H. 2000 Metzler Lexikon Sprache, Stuttgart.

154

Wolfgang Imo

Goldberg, A. 1998 Patterns of Experience in Patterns of Language. In: M. Tomasello (ed.), The New Psychology of Language, Mahwah, NJ, 203–219. Günthner, S. 2006 Grammatische Analysen der kommunikativen Praxis – ‘Dichte Konstruktionen’ in der Interaktion. In: A Deppermann, R. Fiehler & T. Spranz-Fogasy (eds.), Grammatik und Interaktion, 95–121. http://www.verlag-gespraechsforschung.de/2006/pdf/grammatik.pdf. 2007 Zur Emergenz grammatischer Funktionen im Diskurs – wo-konstruktionen in Alltagsinteraktionen. In: H. Hausendorf (ed.) Gespräch als Prozess, Tübingen, 125–156. 2008 ‘Die Sache ist …’: eine Projektorkonstruktion im gesprochenen Deutsch. Zeitschrift für Sprachwissenschaft 27, 1, 39–72. Hausendorf, H. 2007 Die Prozessualität des Gesprächs als Dreh- und Angelpunkt der linguistischen Gesprächsforschung. In: H. Hausendorf (ed.), Gespräch als Prozess, Tübingen, 11–32. Helbig, G. 1999 Deutsche Grammatik, München. Hopper, P. J. 1992 Times of the sign: Discourse, temporality and recent linguistics. Time & Society 1, 2, 223–238. 1997 When ‘grammar’ and discourse clash. In: J. Bybee, J. Haiman & S. A. Thompson (eds.), Essays on Language Function and Language Type, Amsterdam, 231– 247. 1998 Emergent grammar. In: M. Tomasello (ed.), The NewPsychology of Language, Mahwah, NJ, 155–175. 2001 Grammatical constructions and their discourse origins: Prototype or family resemblance?. In: M. Pütz, S. Niemeier & R. Dirven (eds.), Applied Cognitive Linguistics I: Theory and Language Acquisition, Berlin, 109–129. 2004 The openness of grammatical constructions. Chicago Linguistic Society 40, 239–256. 2008 Die Bedeutung der mündlichen Interaktion für die Linguistik: die Pseudocleft-Konstruktion im Englischen. In: K. Fischer & A. Stefanowitsch (eds.), Konstruktionsgrammatik: Von der Konstruktion zur Grammatik, Tübingen, 179–188. Imo, W. 2006 ‘Da hat des kleine glaub irgendwas angestellt’ – Ein construct ohne construction? In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 263–291. 2007a Zur Anwendung der Construction Grammar auf die gesprochene Sprache – der Fall ‘ich mein(e)’. In: V. Ágel & M. Hennig (eds.), Zugänge zur Grammatik gesprochener Sprache, Tübingen, 3–34. 2007b Construction Grammar und Gesprochene-Sprache-Forschung: Konstruktionen mit zehn matrixsatzfähigen Verben im gesprochenen Deutsch, Tübingen. 2007c Der Zwang zur Kategorienbildung: Probleme der Anwendung der Construction Grammar bei der Analyse gesprochener Sprache. Gesprächsforschung – Online-Zeitschrift zur verbalen Interaktion 8, 22–45.

Online changes in syntactic gestalts in spoken German

155

Jacob, G. 2005 The Syntactic Processing of Garden-Path Structures. Diplomarbeit im Fachbereich Psychologie, Münster. Linell, P. 1998 Approaching Dialogue: Talk, Interaction and Contexts in Dialogical Perspective, Amsterdam. 2005 The Written Language Bias: Its Nature, Origins, and Transformations, London. 2008 Rethinking Language, Mind and World Dialogically: Interactional and contextual theories of human sense-making. Unpublished manuscript. Norén, K. & P. Linell 2007 Meaning potentials and the interaction between lexis and contexts: An empirical substantiation. Pragmatics 17, 3, 387–416. Poncin, K. 2000 Apokoinukonstruktionen: Empirische Untersuchung ihrer Verwendung in aufgabenorientierten Dialogen und Diskussion ihrer grammatischen Modellierbarkeit in einer Unifikationsgrammatik. http://deposit.ddb.de/cgi-bin/ dokserv?idn=960672656. Pritchett, B. L. 1988 Garden path phenomena and the grammatical basis of language processing. Language 64, 539–576. Sandig, B. 1973 Zur historischen Kontinuität normativ diskriminierter syntaktischer Muster in spontaner Sprechsprache. Deutsche Sprache 3, 37–57. Schegloff, E. A. 1996 Turn organization: One intersection of grammar and interaction. In: E. Ochs, E. A. Schegloff & S. A. Thompson (eds.), Interaction and Grammar, Cambridge, 52–133. Scheutz, H. 2005 Pivot constructions in spoken German. In: A. Hakulinen & M. Selting (eds.), Syntax and Lexis in Conversation, Amsterdam, 103–128. Schumacher, H. et al. 2004 VALBU – Valenzwörterbuch deutscher Verben, Tübingen. Selting, M. et al. 1998 Gesprächsanalytisches Transkriptionssystem. Linguistische Berichte 173, 91–122. Selting, M. & E. Couper-Kuhlen 2000 Argumente für die Entwicklung einer ‘interaktionalen Linguistik’. Gesprächsforschung – Online-Zeitschrift zur verbalen Interaktion 1, 76–95. Spiekermann, H. & B. Stoltenburg 2006 ‘Lecker Pilsken trinken’ – Konstruktionen unflektierter Adjektive. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 315–342. Stefanowitsch, A. & K. Fischer 2008 Konstruktionsgrammatik II: Von der Konstruktion zur Grammatik, Tübingen. von Foerster, H. 1991 Through the eyes of the other. In: F. Steier (ed.), Research and Reflexivity, London, 63–75. Weinrich, H. 2003 Textgrammatik der deutschen Sprache. 2nd ed., Hildesheim etc.

156

Susanne Günthner

Susanne Günthner

Between emergence and sedimentation: Projecting constructions in German interactions*

It is easy to understand why many linguists are becoming attracted to the view of language as an emergent behavior. For over forty years, syntacticians have worked to establish a fixed set of rules that would specify all the grammatical sentences of the language and disallow all the ungrammatical sentences. Similarly, phonologists have been trying to formulate a fixed set of constraints that would permit the possible word formations of each human language and none of the impossible forms. However, neither language nor human behavior has cooperated with these attempts. Grammars keep on leaking, language keeps on changing, and humans keep on varying their behavior. (…) Searching for more dynamic approaches, they have begun to think of language as an emergent behavior. (Macwhinney 2001: 447)

1

Introduction

In recent years the concept of Emergent Grammar (Hopper 1987, 1998, 2001, 2004) has become more and more influential in Interactional Linguistics, Cognitive Linguistics, as well as in research on grammaticalization. Instead of viewing grammar as an abstract linguistic system with fixed structures, Emergent Grammar treats grammar as a dynamic phenomenon resulting from interactive usages. Grammar and grammatical constructions arise out of the ways participants routinely choose to perform social actions in everyday conversations: The grammar of a language “comes out of discourse and is shaped by discourse as much as it shapes discourse in an ongoing-process” (Hopper 1987: 2). As usage is inherently unstable, grammatical constructions are not fixed but “open”; i.e. they are “openly extendable and have fuzzy and negotiable boundaries and areas * This paper is based on the research project “Grammatik in der Interaktion” (‘Grammar in Interaction’), funded by the German Science Foundation (DFG). http://noam.uni-muenster.de/gidi/. Thanks to Peter Auer, Lars Wegner and the anonymous reviewer for helpful comments.

Between emergence and sedimentation

157

of overlap with other, structurally or functionally similar, constructions” (Hopper 2004: 1). The concept of grammar as emergent from discourse, the idea of grammatical structure as a continual process of becoming, and the perspective of looking at grammatical constructions as they are produced and interpreted in the process of interaction are very much in line with results from empirical studies of language use in concrete everyday interaction. These studies, however, reveal that in order to account for a dynamic concept of grammar (and for grammatical constructions) as emerging from interaction, new linguistic tools and methods are necessary.1 Analyses of grammar have to reorient away from searching for autonomous structures and from treating grammatical constructions as finished entities; instead – in order to capture the ways in which grammar and grammatical constructions arise out of everyday usage – they have to turn towards structural processes and practices as they unfold in time. On the other hand, studies of grammar in interaction also reveal that participants in formal as well as informal situations orient towards sedimented patterns on various levels. Such prepatterned ‘gestalts’ are important tools with which to accomplish interactional work (Auer 2006, 2007a,b; Günthner 2006b; Feilke 2007; Ehmer/Pfänder 2009; Linell 2009). This, of course, begs the question of how to reconcile the perspective of emergence with the concept of sedimentation, and thus, with the fact that participants orient towards conventionalized and routinized prepatterned formats. In this paper, I will argue that the concept of projection (Auer 2005, 2006) plays a central role in linking aspects of local emergence of grammatical structures with certain features of the sedimentation of linguistic patterns.

2

Between sedimentation and local emergence: The relevance of projections in spoken interactions

In his theory of social action, Luckmann (1992: 156) discusses the advantages of routinized and sedimented solutions to social problems. Sedimentation makes the production as well as the processing of communicative activities easier: It takes the burden away from the participants of having to co-ordinate every communicative action or pattern anew. Speakers do not 1

Cf. Ono/Thompson (1995), Hopper (1998, 2001, 2004, 2005), Thompson (2002a,b), Ford (2004), Auer (2005, 2007), Günthner (2005, 2007, 2008a and b), Günthner/Imo (2006), Deppermann (2007), Imo (2007), Ehmer/Pfänder (2008), Linell (2009), Thompson/Couper-Kuhlen (2005).

158

Susanne Günthner

continuously have to ‘invent’ new ways of speaking, new syntactic patterns, new ways of performing communicative activities, etc. And recipients can anticipate next steps and prepare for dealing with them. The idea that the grammar of a language “consists of the inventory of constructions which speakers use” (cf. Hopper 2004: 1) has been discussed within usage-based approaches of Construction Grammar as well as Interactional Linguistics.2 As Auer (2009) argues, in oral communication, which takes place under time constraints, pre-patterned ‘gestalts’ are important means with which to accomplish interactional work. Within Interactional Linguistics, various terms are used to refer to sedimented patterns, which participants use as resources with which to solve communicative problems: “constructional schemata” (Ono/Thompson 1995), “gestalts” (Auer 2000, 2005, 2007a,b), “grammatical formats” (Thompson/ Couper-Kuhlen 2005), “social action formats” (Thompson 2008), “constructions” (Auer 2006; Günthner 2006a,b, 2007, 2008a,b; Günthner/Imo 2006; Imo 2007; Deppermann 2007; Birkner 2008; Hopper/Thompson i.pr.), etc. Besides work in Construction Grammar and Interactional Linguistics, studies within the Sociology of Knowledge as well as Linguistic Anthropology also argue that in accomplishing social interaction, participants make use of sedimented patterns. These patterns – so called ‘communicative genres’ or ‘discourse genres’3 –, which are part of the social stocks of knowledge, are treated as socially constructed solutions which organize, routinize, and standardize the dealing with particular communicative problems (Luckmann 1986, 1988; Bergmann 1987; Günthner/Knoblauch 1995). They represent central communicative means in the construction of social reality (Luckmann 2002). Thus, routinized patterns – ranging from grammatical constructions to larger genre-like formats – are considered to play a major role in verbal communication (and cannot be reduced to idiomatic expressions): Whenever socially relevant and recurrent actions are to be negotiated, we find forms of routinization.4 2

3

4

Cf. Ono/Thompson (1995), Croft (2001), Hopper (2004), Thompson (2002a and b), Auer (2005, 2006, 2007, 2009), Günthner (2005, 2006a,b, 2007, 2008a,b, 2009), Günthner/Imo (2006), Deppermann (2007), Imo (2007), Thompson/CouperKuhlen (2005), Fried/Östmann (2005), Fischer/Stefanowitsch (2007), Linell (2009), Stefanowitsch/Fischer (2008), Birkner (2008). Cf. Hymes (1974), Bakhtin (1978/1986), Luckmann (1986, 1988), Bergmann (1987), Hanks (1987), Günthner/Knoblauch (1995), Knoblauch (1995), Günthner (2000). Cf. Günthner/Knoblauch (1995).

Between emergence and sedimentation

159

Luckmann as well as Bakhtin (1976/86) claim that the process of sedimentation cannot be abstracted from usage: Sedimentation takes place in the process of a long chain of communicative situations. It is this dialectic process between the local emergence of social actions on the one hand and the sedimentation/routinization of particular action patterns on the other hand, which – according to Berger/Luckmann (1966: 65) – is central to “the social construction of reality”. This ongoing tension between sedimentation and emergence – or according to Bakhtin (1981: 272) between processes of “centralization and decentralization, of unification and disunification” – accounts for recognition and novelty, for reproduction and modification, for standardized/canonical ways of speaking and non-standardized ways. What Bakhtin (1978/1986: 78f.) writes about genres, also holds for grammatical constructions: We know our native language – its lexical composition and grammatical structure – not from dictionaries and grammars but from concrete utterances that we hear and that we ourselves reproduce in live speech communication with people around us. (…) If speech genres did not exist and we had not mastered them, if we had to originate them during speech process and construct each utterance at will for the first time, speech communication would be almost impossible.

In studying how participants orient to sedimented patterns and how grammatical constructions emerge in interactions, the concept of projection plays a major role. In his conception of an ‘on-line’-syntax, Auer (2000, 2005, 2007, 2009) argues that projections of various kinds form a fundamental characteristic of the production as well as the reception of utterances: “Human interaction rests on the possibility of projection; the grammars of human languages provide interlocutors with sedimented and shared ways of organizing projection interaction” (Auer 2005: 8). In the temporal unfolding of syntactic and interactive gestalts, speakers build up expectations about the continuation of these patterns. In German, for example – due to its sentence bracket – the left element (the finite auxiliary verb in ich habe gestern Abend nichts?; ‘I have yesterday evening nothing’?) projects a particular right element (the participle gegessen; ‘eaten’). The production of the feminine article with an inflected adjective (die freundliche?; ‘the friendly?’) foreshadows a noun with feminine gender (Kellnerin; ‘waitress’); the articulation of a lengthened intensifier such as so::::? (‘so::::?’) may anticipate an evaluative adjective; etc.5 Projections, however, are not only based on grammatical but also on sequential knowledge as well as on knowledge about handling complex lin5

Cf. Auer (2000, 2007) and Goodwin (1996).

160

Susanne Günthner

guistic activities: They are a fundamental characteristic of turn taking (Sacks et al. 1974; Houtkoop/Mazeland 1985; Auer 2005), they are important to the production of multi-unit turns, to the construction of ‘larger projects’ and ‘actions’ (Schegloff 1980; Houtkoop/Mazeland 1985; Streeck 1995; Ford 2004) and to the use of communicative genres (Luckmann 1986, 1988; Bergmann 1987; Günthner/Knoblauch 1995; Günthner 2006b). If, for instance, a speaker begins an utterance with the words ‘Have you already heard the news about our neighbour …’, she projects expectations concerning the genre to be constituted (such as gossip). The expectations are related to the form as well as to the content to follow. The neighbour is expected to have done something morally deviate, which the speaker plans to reconstruct in a particular way (as gossip). Thus, projections are important devices in coordinating interaction on various levels. They assist in organizing turns, in announcing activities to come, in modalizing utterances, in expressing a stance, in preparing facethreatening activities, in contextualizing a complex genre or in inviting coparticipants to share a certain reaction and signaling when they are expected to take over the floor (Günthner 2006a,b, 2007, 2008a): While speaking, we constantly foreshadow what is going to come next. We thereby enable our recipients to project these upcoming items, and thereby anticipate next steps, get prepared for dealing with them, and in general, process them more easily. Projections can be weaker or stronger, and the predictability of next items is accordingly high or low. However, projection never equals determination, i.e. even a strongly projected next item may not be delivered, either because the speaker has abandoned the project entirely (in which case a fragment will remain) or because s/he chooses to engage on an unlikely project not easily projectable. (Auer 2007b: 1)

Projections are possible because participants have shared experiences and knowledge about what patterns are routinely used for performing what kind of social actions. This shared knowledge also allows co-participants to complete or co-construct a started pattern (Lerner 1987, 2002; Ono/Thompson 1995, 1996; Auer 2005, 2007, 2009; Couper-Kuhlen/Thompson 2006; Günthner 2006a,b; Günthner/Imo 2006; Deppermann 2007; Imo 2007). However, even though participants orient to sedimented patterns, their actualization takes place in the hic et nunc of the interaction; linguistic patterns are deferred, locally negotiated, and thus, emergent. In the following, I will illustrate this dialectic process of local emergence and sedimentation of linguistic patterns by focusing on constructions used to project upcoming discourse.

Between emergence and sedimentation

3

161

From bi-clausal sentence patterns to complex constructions6

In studying grammatical constructions in spoken interactions, we discover time and again that those seemingly canonical forms presented in reference grammars and in linguistic studies (including some functional variants) are not only based on written language but that they are rarely used in spoken interactions.7 This “written language bias” (Linell 2005) which characterizes mainstream linguistics has lead to the following paradox in modern linguistics: “spoken language is regarded as the primary form of language, yet it is studied by the use of theories and methods that are heavily biased toward written language.” (Linell 2009: 278). The following analysis will look at complex syntactic constructions – German pseudo-clefts, die Sache/das Ding ist … (‘the thing/point/problem is …’)-constructions and extrapositions with es (‘it’) – in the course of their emergence in real time. I will argue that when studying these patterns as they emerge in real-time discourse, they can no longer be treated as ‘bi-clausal sentence patterns’, consisting of two clauses combined into a single construction. Instead, these constructions are especially designed to the manage temporality in interaction: They mainly function as constructions that project upcoming discourse; i.e. as so-called ‘projective constructions’ (Hopper 2005; Günthner 2006a, 2007, 2008a,b, 2009; Günthner/Hopper 2010; Auer 2009). The analysis is based on a corpus of 91 everyday interactions (30 to 180 minutes in length), collected from 1989 to 2006 in different parts of Germany. They include informal face-to-face interactions among friends and family members, office hours at university, genetic counselling sessions, radio phone-in programs, as well as data from the reality TV-series ‘Big Brother’. 3.1 Pseudo-clefts in German interactions Pseudo-clefts in German (e.g. Was ich seltsam finde, ist dass manche ALTE Dateien okay sind und andere nicht; ‘What I find strange, is that some of the OLD data are okay and others aren’t’, E-MAIL communication) are treated by grammarians as complex, bi-clausal sentence patterns, composed of a dependent w-clause, which is positioned in the front field of a matrix clause (a copula sentence with the verb sein) and a complement clause (or a noun phrase) in the middle 6 7

The following analysis is based on Günthner (2008a). Cf. also Thompson (2002b), Hopper (2004), Couper-Kuhlen/Thompson (2006), Günthner (2008a,b, 2009), Günthner/Hopper (2010).

162

Susanne Günthner

field of the matrix clause.8 Lambrecht (2001: 467) argues that the main characteristic of cleft sentences is that “they express a single proposition via bi-clausal syntax”. Traditionally, the initial w-clause (part A) is considered to contain the presupposed or given information, which is “in the hearer’s consciousness” (Prince 1978: 904), whereas the following clause (or the NP) (part B) carries the focus (Lambrecht 2001). The assumption that pseudoclefts are primarily used as focus marking devices is prevalent in linguistic research, since the w-cleft structure explicitly separates the presupposition from the focus:9 A constituent is transferred from its normal position in order to focus it or to assign a contrast focus to it (Collins 1991). In studying the use of pseudo-clefts in spoken German interactions, one discovers that speakers rarely use these so-called canonical forms [w-clause + copula + NP/complement clause]: Neither does the w-clause in spoken German interactions always contain given information, nor can the construction be classified as bi-clausal, consisting of two clauses combined into a single construction. The part following the copula clause cannot be reduced to a NP or a single clause; and even the copula, which is treated as the kernel of pseudo-clefts, is not always realized (Günthner 2006a). In studying pseudo-clefts as structural processes unfolding in time – instead of analyzing them as finished entities from a post factum perspective –, we recognize that the first part (part A) – an incomplete utterance – opens a projection span which draws the recipients’ attention to the missing constituent (an object or subject complement) (part B). This constituent can vary in form from that of a copula clause (i.e. a copula plus a NP) to a complex discourse segment, which stretches over several turn construction units.10 Thus, the realization of the construction initiated by the w-clause is not fixed, but is interactively produced; it is ‘emergent’ and ‘open’ (Hopper 1998, 2001, 2004). First, an example of a canonical pseudo-cleft: STUDENTINNEN: Münster 3 51 Isa: (...) ganz vergEssen. 52 Lore: soll ich SIE=n anrUfen? 53 (-) 54 Isa: hm (.) also was ich WICHtig finde,

8

9

10

Cf. Jespersen (1927, 1937/65, 1949). Cf. also transformation grammatical approaches (Akmajian 1970). Cf. Jespersen (1949: 147f.), Blatz (1886/1970: 896), Paul (1919/68: 64), Collins (1991: 44ff.), Andersson (1993: 41ff.), and Dik (1997: 292). Cf. Günthner (2006a, 2008a) and Birkner (2008). Cf. Hopper (2001) on pseudoclefts in English.

Between emergence and sedimentation 55 56 57 58

163

is (.) da- dass Ihr euch vertrAUt. und nich so ständig schlechtes WITtert, so dieses mIsstrauen. (un so)

STUDENTS: Münster 3 51 Isa: (…) totally forgotten. 52 Lore: shall I call her? 53 (-) 54 Isa: hm (.) actually what I think is important, 55 is (.) 56 th- that you trust each other. 57 and don’t always expect the worst, 58 that kind of distrust. (things like that)

In line 54, Isa constructs her advice in the form of a pseudo-cleft. Part A (also was ich WICHtig finde,; ‘actually what I think is important,’) consists of a dependent clause, positioned in the front field of a matrix clause, and the copula verb sein (Günthner 2006a). The second component of the construction is a complement clause – introduced by the complementizer dass (‘that’) – which fits into the syntactic slot occupied by was (‘what’) in the first component da- dass Ihr euch vertrAUt. (‘th- that you trust each other.’). In studying the construction as it emerges in discourse, we realize that the w-clause opens a syntactic gestalt, which – due to the projecting force of the interrogative pronoun was (‘what’) and the ‘unsaturated’ verb (i.e. particular arguments of the verb are missing) – builds up certain expectations concerning what is to come. The construction is only complete once the expected constituent (i.e. the object) is provided. As in the transcript STUDENTINNEN (‘STUDENTS’), speakers in my data frequently use w-clauses to mark their stances (also was ich WICHtig finde,; ‘actually what I think is important,’) toward the subsequent assertion.11 Introducing w-clauses, thus, come close to what Goodwin (1996: 384) calls “prospective indexical”. Also, in the case of pseudo-clefts, recipients have to gradually detect the specification of the evaluation. The announcement of something ‘important’ not only draws the recipients’ attention to the ‘important’ issue, but it also indicates what recipient reaction the speaker expects: The occurrence of prospective indexicals thus invokes a distributed, multi-party process. The cognitive operations relevant to the ongoing constitution of the event in process are by no means confined to speaker alone. Hearers must engage in an active, somewhat problematic process of interpretation in order to uncover the specification of the indexical that will enable them to build appropriate subse11

For stance-taking in interaction see Kärkkäinen (2006).

164

Susanne Günthner

quent action at a particular place. Moreover this analysis is not static, complete as soon as the prospective indexical is heard, but is instead a dynamic process that extends through time as subsequent talk and the interpretative framework provided by the prospective indexical mutually elaborate each other. (Goodwin 1996: 372)

Even though one can find canonical forms of pseudo-clefts [w-clause + copula + NP/complement clause] in spoken German, our data show that they represent just one type of pseudo-cleft used in spoken interactions. More frequently, speakers start with a w-clause, but what follows (part B) is an independent, syntactically and prosodically non-integrated main clause followed by a longer stretch of discourse: [w-clause + copula + main clause/or longer segment of talk].12 In the next excerpt, taken from the reality TV-series ‘Big Brother’, the speaker produces a pseudo-cleft with a ‘prospective indexical’ (part A) and a complex B-part, stretching over several turn construction units. Christian criticizes his fellow-occupants and emphasizes that from now on, he intends to sein eigenes Ding durchziehen (‘look out for himself ’): BIG BROTHER: KRITIK AM EGOISMUS (bb2–17) 23 Chr: .h dann soll=n se entweder ihren KOFfer packen 24 und hier die FLIEge machen, 25 =oder die sollen mich wieder nomiNIEren, 26 und dann LACH ich da wieder mal drüber; 27 =und dann is GUT; 28 .hhh 29 WAS ich eigentlich damit sagen wollte; 30 (0.5) 31 IST, 32 ich glaube(.) es würde hier VIE:L VIE:L besser ABlaufen; 33 und VIE:len leuten VIEL besser gehen, 34 .hh wenn die EINfach mal MEHR? (0.5) 35

36 (3.0) 37 mehr AN SICH denken; 38 AN SICH denken; 39 NICHT FÜ:R sich denken; 40 weil viele denken FÜR sich, 41 wie kann ich möglichst viel für mich hier ABstauben, BIG BROTHER: CRITICIZING EGOISM (bb2–17) 23 Chr: .h then they should either pack their suitcases 24 and beat it, 25 =or they should nominate me again,

12

For more details see Günthner (2006a). Cf. also Birkner (2008). Cf. Hopper (2001, 2004) and Hopper/Thompson (i. pr.) on pseudo-clefts in English interactions.

Between emergence and sedimentation 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

165

then I will laugh about it again; =and then it’s okay; .hhh what I actually meant to say; (0.5) IS, I believe(.) things would be much much better here; and many people would feel better, .hh if they would just? (0.5)

(3.0) think more about themselves; think about themselves; not in favour themselves; because many people think in favour themselves, thinking how can I get the most out of this for myself,

The pseudo-cleft in line 29ff. is not syntactically bi-clausal. Looking at its ‘on-line’ development, the utterance starts with a w-clause (WAS ich eigentlich damit sagen wollte;; ‘what I actually meant to say;’), which by itself does not represent a complete communicative action: On the syntactic level, the direct object is still missing, and on the semantic level the utterance is incomplete (what is it Christian wants to say?). Thus, the w-clause projects ‘more to come’. However, the project is not continued right away: the initiated gestalt is interrupted by the inclusion of a parenthetical sequence (‘’). As this piece of interaction shows, even though syntactic projection pre-structures the following position(s), it need not be dealt with immediately; a projection can be deferred, and it remains valid across inserted materials.13 Shortly after the parenthetical side sequence, the speaker produces the prosodically marked copula IST (‘is’) and thus, returns to the initiated gestalt. However, instead of a subordinated complement clause, Christian provides a longer stretch of discourse – introduced by the verbum sentiendi ich glaube (‘I believe’): es würde hier VIE:L VIE:L besser ABlaufen; und VIE:len leuten VIEL besser gehen, .hh wenn die EINfach mal MEHR? […] mehr AN SICH denken; AN SICH denken; NICHT FÜ:R sich denken; (‘things would be much much better here; and many people would feel better, .hh if they would just? (0.5) […] think more abour themselves; think about themselves; not in favour of themselves’). Part B shows no signs of hypotactic marking: It is neither introduced by a complementizer, nor does it carry verb-final placement. In interactional use, it is often difficult or even impossible to determine the exact ending of a pseudo-cleft, as the construction is incrementally prolonged, 13

Cf. Günthner (2006a) and Auer (2009).

166

Susanne Günthner

without clear endpoints. Thus, the end of part B is open to negotiation. Instead of treating these uses of pseudo-clefts as deviations from the canonical bi-clausal structure, it might be productive to look at them as patterns which solve ‘specific’ communicative problems in everyday interactions. As Franck (1985: 238) points out, we can avoid deviant categorizations which do not coincide with experiences we have as participants in interaction, if we perceive an utterance from the vantage point of “‘mobile observers’, travelling along with the stream of speech and on the spot producing hypotheses of understanding which change and vary with the point the utterance has reached.” In dealing with pseudo-clefts as patterns unfolding in time, we perceive them as complex constructions, starting with a w-clause which projects upcoming talk by the same speaker. This upcoming talk, however, can be a syntactically and interactionally complex stretch of discourse, often expanding over several TCUs.14 An a priori grammar analysis that postulates a bi-clausal structure misses important characteristics of pseudo-clefts in interactions: Speakers construct grammatical patterns in the process of talk; thus their utterances are open to modifications, insertions, prolongations, hybridizations, amalgamations, etc. The organization of an utterance into a projecting w-part and a succeeding segment provides considerable advantages in spoken German interaction: Until the end of the w-part, speakers have the opportunity to construct this part as syntactically integrated and closely connected to the following segment or as disintegrated and only loosely connected to what follows. In cases in which the following components consist of a longer stretch of discourse, they often follow part A asyndetically and show no signs of hypotactic marking. This asyndetic juxtaposition makes the production of complex segments, expanding over various TCUs, easier to handle. Furthermore, in metapragmatically framing the speaker’s statement, the w-clause not only delays an important argument and draws the recipients’ attention to it, but it also functions to keep the floor pending the upcoming stretch of discourse.15 This division into a framing part and a succeeding part is also convenient for the recipient, because it simplifies her/his task of processing the information, as – due to their 14

15

Cf. Hopper (2001, 2004) and Hopper/Thompson (i.pr.) for similar results in English. Cf. also Houtkoop/Mazeland (1985: 598) about devices speakers use to claim the floor as long as the projected ‘discourse unit’ is not yet completed. Hopper’s (2001: 114) study of pseudo-clefts in English reveals similar results: “The pseudocleft works to delay the delivery of a significant segment of talk. It accomplishes this by adumbrating (foreshadowing) the continuation in general terms without giving away the main point.”

Between emergence and sedimentation

167

projective force – the framing elements limit the possibilities of interpretation and suggest certain expectations. Even though the presence of the copula is treated as a constitutive characteristic of pseudoclefts (Ross 2000: 388; Lambrecht 2001: 467ff.),16 participants in spoken interaction frequently use pseudo-clefts in which the w-clause is not followed up by a copula clause (Günthner 2006a). These forms without a connecting copula appear with a subordinate complement clause following the w-clause: [w-clause + subordinated complement clause] as well as with a main clause following the w-clause: [w-clause + main clause]. In the following segment, which again stems from the reality TVseries Big Brother, the participants are talking about East and West Germans. Alida states that she also had chocolate and new toys (just as West German children had), even though she was brought up in East Germany: BIG BROTHER: OSSIS & WESSIS 66 Alida: h soweit ich mich zuRÜCK erinnern kann; (-) 67 .h ich hatte dann AUCH eben immer schokola:de; 68 =als das alles und vie69 und neues SPIELzeug und die ganzen sachen also; 70 =.h was WAS ich nur SCHAde finde, 71 dass ebend .h 72 auch HEUTzutage; 73 =auch bei den OSTdeutschen; 74 oder überHAUPT, 75 .h bei so VIElen sachen .h (.) 76 der WERT verloren geht; 77 HEUT muss jedes kind n comPUter haben, BIG BROTHER: EAST-& WESTGERMANS 66 Alida: h as far back as I can remember; (-) 67 .h even then I always had chocolate; 68 =when all that and many 69 and new toys and all those things; 70 =.h what what I think is a shame, 71 that well. h 72 even nowadays; 73 =even the east germans; 74 or just in general, 75 .h you don’t appreciate the value .h (.) 76 of many things; 77 today every kid has to have a computer, 16

Cf. Lambrecht (2001: 470) who states: “This entails that the copula, together with its empty subject, serves as a kind of focus marker for the argument of another predicator”.

168

Susanne Günthner

The w-clause in line 70 (=.h was WAS ich nur SCHAde finde,; ‘=.h what what I think is a shame,’) is followed by a complement clause (dass ebend .h auch HEUTzutage; =auch bei den OSTdeutschen; oder überHAUPT, .h bei so VIElen sachen .h (.) der WERT verloren geht;; ‘that well. h even nowadays; =even the east germans; or just in general, .h you don’t appreciate the value .h (.) of many things;’) showing hypotactic markers, i.e. the complementizer dass as well as final positioning of the verb. However, there is no connecting copula between the two parts. Examples like this cause one to question the prevailing thesis that the function of the pseudocleft is connected to the copula. In our example, the pseudo-cleft works well – even without a copula; neither is the utterance marked as ungrammatical nor is it produced in connection with any indications of planning problems.17 In the following segment, taken from a conversation about health food, the w-clause is neither followed by a copula nor by a syntactically dependent complement clause. Instead, part B represents a syntactically independent unit, displaying ‘main clause order’, with the finite verb (KOMMST; ‘come’) in verb-second position (a grammatical feature of independent clauses in German). In this type of ‘de-grammaticalized’ pseudo-cleft, the inherent grammatical cohesion between the two parts is weakened:18 RESTAURANTS IN MÜNSTER II (28–1; 2003) 17 Bert: also ich denke=ja, 18 Udo: hm? 19 Bert: was immer äh- e- entSETZlich is, 20 du KOMMST inne restaurAnt rein, 21 und dat riescht schon so Ü:BEL. 22 Udo: hm. 23 Bert: da biste doch ech=schon beDIENT. ne? RESTAURANTS IN MÜNSTER II (28–1; 2003) 17 Bert: well I think, 18 Udo: hm? 19 Bert: what is always äh- h- horrible, 20 you enter a restaurant, 21 and it smells so terrible. 22 Udo: hm. 23 Bert: then you’ve had enough already. right?

The presented extracts indicate that pseudo-clefts as they are represented in studies based on written or manufactured constructions are by no means the 17

18

In German pseudo-clefts the copula is optional in front of a complement clause and a main clause, but not in front of an NP. Cf. Auer/Günthner (2005) on grammatical cohesion.

Between emergence and sedimentation

169

norm in spoken discourse; instead they are used as projective constructions which are especially designed to manage temporal concerns in interaction.19 3.2 Die Sache/das Ding ist … (‘the thing/point is …’)-constructions Nominal constructions such as die Sache/das Ding/das Problem/der Punkt ist … (‘the thing/point/problem is …’) are another example of seemingly bi-clausal sentence patterns, whose on-line interactive realizations show profound discrepancies from the way they are portrayed in studies based on written language. Die Sache/das Ding ist … (‘the thing/point is …’)-constructions are supposed to consist of a matrix clause followed by a subject complement clause: [matrix clause + complement clause]. The matrix clause (die Sache/das Ding/ das Problem/der Punkt ist …) is not a fully-fledged syntactic ‘gestalt’, as its verb (the copula) requires a further constituent. This constituent is produced in the following complement clause which reveals typical features of subordinate clauses in German (i.e. the subjunction dass (‘that’) as well as by verbfinal constituent order). In Construction Grammar, complex sentences such as ‘the thing/point/ problem is that …’ are referred to as “N-be-that-constructions” (Schmid 2001). They are supposed to consist of an initial noun phrase headed by an abstract noun (‘thing, point, problem …’) functioning as a subject, a form of the copula ‘BE’ and a ‘THAT’-clause syntactically functioning as subject complement: [abstract noun, copula, that-clause].20 In looking at ways in which interactants use this construction in spoken language, it becomes clear once again that the canonical form [matrix clause 19

20

Cf. Hopper (2001, 2005) and Hopper/Thompson (i. pr.: 4) for similar results in English. As Hopper/Thompson (i. pr.: 5ff.) point out: “We have suggested that a major factor in the normativization and persistent perception of bi-clausality in constructions such as the English pseudo-cleft is the strong projectability of the wh-component. That is, in interactional spoken English this construction routinely projects ‘more to come’. As we have seen, the ‘more’ is most frequently not a clause but an indeterminate stretch of discourse. But in normative English, the construction has become standardized so that the ‘more’ is a grammatical clause.” In his corpus-based study of “N-be-that-constructions” in written English, Schmid (2001) argues that the ten nouns that were found to occur most frequently in this construction are “problem, thing, truth, fact, trouble, point, result, view, reason, idea”. Besides the fact that his observations are based on English and, thus, are not automatically transferable to German, all his examples stem from written data. Thus, various forms and functions predominant for spoken language are not taken into account.

170

Susanne Günthner

+ complement clause] represents the exception and not the rule. Participants in everyday interactions make use of die Sache ist-constructions in ways that systematically deviate from standard forms based on written language: The ‘N-be’-part (die Sache/das Ding ist) can no longer be treated as a matrix clause holding the relevant information for the following discourse, nor is the following syntagma (i.e. the complement clause) formally and conceptually subordinate to the preceding clause. Instead, speakers use the first component – the so-called ‘matrix clause’ – to build up a projection span, contextualizing ‘more to come’ (Günthner 2007, 2008a,b). Whereas, the architecture of the first component (part A) is rather stable (it consists of an abstract noun i.e. ‘thing, point, problem …’ functioning as a subject and a form of the copula sein ‘be’), the following component (the seemingly ‘complement clause’; part B) is an emergent product of local management; i.e. it can take on various forms ranging from subordinate clause structure to a longer stretch of discourse, expanding over several TCUs with no discernable righthand boundary. Furthermore, it is the so-called complement part which holds the crucial information. Again, we shall first look at a transcript segment displaying the canonical form. This das Ding ist-construction consists of a first component with an initial NP and the copula ist (in the present tense) and a following component which is introduced by the complementizer dass and shows subordinate clause word order (i.e. final positioning of the finite verb).21 The transcript stems from a talk during a university office hour. Elke, a lecturer, has just proposed to her student Birte that – instead of starting to work on a new topic – she would be better served to write her dissertation about the same topic she explored in her Master’s thesis: PROMOTION (MÜNSTER 88–2; 2005) 1 Elke: und dann auch vie- vielleicht, (.) 2 lieber DAS thema. 3 Birte: das DING ist aber auch4 dass ich in der germanIStik promoVIEren will. 5 (0.5) 6 Elke: [mhm] 7 Birte: [und] deshalb ein germanIStisches THEma brauch.

21

German, which has verb-second as its basic word order in simple and main clauses, requires final position of the finite verb in subordinate clauses. Thus, complement clauses introduced by the subjunction dass (‘that’) – according to German grammar – display verb-final order. German thus provides a clear signal for the grammatical incorporation of one clause into another.

Between emergence and sedimentation

171

DISSERTATION 1 Elke: and then also per- perhaps, (.) 2 THIS topic would be better. 3 Birte: but the THING is also4 that I want to get my PhD in german. 5 (0.5) 6 Elke: [mhm] 7 Birte: [and] that’s why I need a topic within germanistics.

In response to Elke’s proposal (lines 1–2), Birte produces a rejection, introduced by das DING ist aber auch- (‘but the THING is also-’). Already the opposition marker aber auch (‘but also’) foreshadows an upcoming disagreement. The matrix clause das DING ist aber auch- (‘but the THING is also-’) opens a projection space, which delays her main argument, that she wants to write her PhD thesis in German Studies (and not in General Linguistics) and, thus she needs a new topic. However, in starting her utterance with the framing part das DING ist aber auch- (‘but the THING is also-’), Birte does not just foreshadow disagreement but she also sets up a new context for her rejection of Elke’s proposal. Only with the closing of the second component is the syntactic gestalt complete. The two components of the construction are realized in two independent prosodic contours. Even though the construction reveals a complex syntactic gestalt with a main clause and a following subordinate clause, which shows various elements of syntactical integration (i.e. the complementizer dass plus final positioning of the finite verb), it becomes obvious that the focal point is not in the ‘matrix clause’ but in the so-called ‘complement clause’.22 In spoken interaction, we frequently find das Ding/die Sache ist-constructions, which further deviate from the canonical form, the seemingly ‘complement clause’ is no longer introduced by the subjunction dass, but shows the word order of an independent sentence (i.e. verb-second positioning). In the following example, Sven tells his fellow student Tanja about his professor who refuses to give him credit for having attended a seminar: PHILOSOPHIE-SCHEIN (MÜNSTER 90–1; 2005) 21 Tanja: dann wÜrd ich auch nich mehr (.) 22 zu dem PROF gehen, (-) 23 und ihn auch nich als PRÜFer NEHmen.

22

This observation is in line with Thompson’s (2002b: 134) criticism of traditional conceptions of ‘complement clauses’: “In sum, then, the data show that what conversationalists are engaged in doing with their talk crucially involves the complement; in the majority of cases, the complement ‘overrides’ the ‘main clause’, and the ‘main clause’ is there to provide speaker stance towards the assessments, claims, counterclaims, and proposals.”

172 24 25 26 27 28 29 30

Sven: Tanja: Sven:

Tanja:

Susanne Günthner ne. ] [mhm] die sache is; er will mir nich MAL den FSCHEIN anerkennen; (.) weil er sagt, es wäre manipu[lIert.] [mhm]

CREDIT IN PHILOSOPHY 21 Tanja: in that case I also wouldn’t go anymore (.) 22 to that professor, (-) 23 nor would I pick him as your supervisor. 24 Sven: no. ] 25 Tanja: [mhm] 26 Sven: the thing is; 27 he does not even want to give me FCREDIT for the course; (.) 28 cause he says, 29 it would be manipu[lated.] 30 Tanja: [mhm]

Similar to the example in PROMOTION, the die Sache ist …- construction is used to express the speaker’s stance and at the same time to set up a new focus in the interactional proceeding: By referring to the problem that his professor does not even want to give him a credit and has accused him of having manipulated his homework, Sven uses the die Sache ist …-phrase to recontextualize his problem. The beginning of the die Sache ist-construction in this extract corresponds with the canonical form above; however, part A is followed by a syntactically as well as prosodically independent clause, displaying ‘main clause order’, with the finite verb (will) in verb-second position: The syntagma er will mir nich MAL den FSCHEIN anerkennen; (‘he does not even want to give me FCREDIT for the course;’) (l. 27) – incrementally followed by a causal clause (weil er sagt, es wäre manipu[lIert.]; ‘cause he says, it would be manipu[lated.]’) – shows ‘main clause syntax’. Instead of a syntactic and conceptual dependence of the complement clause on the matrix clause, it is now the projecting ‘matrix clause’, which cannot stand on its own. The grammatical cohesion between the “N-be”-clause and the following segment is greatly reduced; each part has its own intonation contour, and part B could stand on its own without being ungrammatical. Even though the die Sache/das Ding ist-part can be left out without the utterance becoming ungrammatical, it still has important interactional functions: On the one hand, it anticipates the following component and guides the recipients’ attention to the focal proposition. On the other hand, it ensures the speaker the opportunity to finish her/his multi-unit turn. This floorholding function is especially important in the case of longer discourse

Between emergence and sedimentation

173

segments. Thus, it is not surprising that die Sache ist-constructions are often used to introduce complex arguments and information which extend over several turn construction units (TCUs). When studying die Sache ist-constructions as emergent products in interaction (instead of treating them as fixed entities), we can detect striking parallels between the die Sache ist-part and pre-sequences as described in Conversation Analysis (Schegloff 1980, 1984; Streeck 1995): With the ‘pre’, speakers foreshadow what might follow; i.e. they “prepare the scene” (Schegoff 1984) and allow co-participants a certain “premonition as to what this actor might be up to next” (Streeck 1995: 87). In marking the subsequent talk as ‘focal’, speakers prepare co-participants to align themselves to the upcoming information. In producing a die Sache ist-construction, speakers exploit the delaying function of the main point; the die Sache ist-construction can provide cognitive and interactional space for thinking through claims: This ‘thinking through’ is, like all discourse, as much interactional as cognitive in nature, since it aims to extend the speaker’s turn and stave off interruption and possible derailment while the argument is being worked out. (Hopper/Thompson i.pr.: 8)

The following segment shows that also in the case of projective die Sache/das Ding ist-constructions, initiated syntactic gestalts need not be dealt with immediately, but can be deferred and still remain active across inserted side-sequences (Auer 2005). This observation supports the thesis of the intertwinedness between participants’ knowledge of sedimented patterns on the one hand, and the local emergence and openness of syntactic patterns on the other. The transcript is taken from an interaction between Olga, a patient who suffers from panic attacks and her friend Eva. Olga describes the difficulties she had driving her car after suffering from a panic attack: PANIKATTACKEN: OLGA-EVA 21 Olga: es hat mich SEHR v- viel überWINdung ge[kOstet;] 22 Eva: [un-] 23 Eva: hm? 24 Olga: d- das ding is hAlt; (-) 25

26 Eva: hm 27 Olga: wenn=du dat EINma hAst, 28 dat LÄSST dich NICH (mehr)los. 29 ECHT. NICH. PANIC ATTACKS: OLGA-EVA 21 Olga: I really had to force [myself;] 22 Eva: [un-] 23 Eva: hm? 24 Olga: th- the thing is; (-)

174 25 26 Eva: 27 Olga: 28 29

Susanne Günthner

hm once you have it, you can’t escape it. you really can’t.

Following Eva’s encouraging minimal response (l. 23), Olga starts with a general statement about panic attacks (l. 24ff.). d- das ding is hAlt; (-) (‘th- the thing is; (-)’) opens a projection span, which is delayed for the sake of a parenthetical insertion (‘’) (l. 25) as well as Eva’s minimal response token (l. 26). We can observe how participants orient to sedimented patterns in producing interactive activities. The actualization of the pattern, however, happens in the hic et nunc of the interaction; it is not fixed but emergent, and thus, responds to local contingencies. In case of the die Sache/das Ding ist-construction in line 24ff., part A is not followed by a subordinate complement clause, but by a syntactically as well as prosodically independent segment, displaying ‘main clause order’, with the finite verb in verb-second position. The die Sache ist-construction allows speakers to maintain the floor even in cases in which they insert sidesequences. The fact that Olga in line 27f. finally finishes her construction by adding a conditional clause (wenn=du dat EIN.MA hAst, dat LÄSST dich NICH (mehr) los.; ‘once you have it, you can’t escape it.’),23 indicates that participants orient to this constructional pattern. 3.3 Extrapositions with es (‘it’) Extrapositions represent a further complex syntactic structure which is traditionally treated as a bi-clausal sentence pattern. The term ‘extraposition’ refers to a syntactic process which moves a syntactic unit (generally a subordinate nominal clause) to the right of the predicate in the superordinate clause and replaces it with a dummy pronoun (such as ‘it’ in English or es in German):24 When for some reason or another it is not convenient to put a content-clause in the ordinary place of the subject, object, etc., the clause is placed at the end in extraposition and is represented in the body of the sentence itself by it. (Jespersen 1937/65: 25)

23

24

In die Sache/das Ding ist-constructions we frequently find modal particles and adverbs (such as halt, nämlich, natürlich, aber etc.), which are used to contextualize speakers’ argumentative direction (such as disagreement) or to back the validity of a following argument. In PANIC ATTACKS: OLGA-EVA the modal particle halt (d- das ding is hAlt;) enforces the validity of the following sententious maxim. Cf. Collins (1994: 8), Bußmann (2002: 210) for German extrapositions with es.

Between emergence and sedimentation

175

As Couper-Kuhlen/Thompson’s (2006) analysis of extrapositions in spoken English shows, ‘movement assumptions’ as postulated by various traditional as well as generative studies have no empirical basis, not even in written English. Rather than assuming that the two parts are clauses in a bi-clausal constructional gestalt, Couper-Kuhlen/Thompson (2006) argue that they are best characterized as cognitively stored and interactionally used as two separate units. An analysis of German extrapositions with es (‘it’) in spoken interaction provides similar results: Interactants use the two segments (A and B) not as a single bi-clausal sentence pattern, but as a projection construction with portion A functioning as a projector phrase anticipating more to come (Günthner 2009). Part A involves an evaluative, epistemic, or evidential statement, framing the component (part B) to follow. Part B however, cannot be reduced to a single clause, instead it is more flexible or “open” (Hopper 2004) than it is portrayed in linguistic studies based on written forms of standard German. As in the case of pseudo-clefts and die Sache/das Ding ist-constructions, even though we find some ‘canonical’ realizations (as in the following excerpt FREUNDINNEN; ‘GIRLFRIENDS’), speakers in interactions do not stick to neatly bounded bi-clausal sentence patterns. Instead we find heterogeneous forms which are much more ‘open’ and dynamic, responding to local, interactive contingencies. Again, I want to argue that these ‘open’ and dynamic forms cannot be treated as deviations or deformations, but are functional – and even sedimented – resources participants use as solutions to specific interactive problems. In the following segment, Betty tells her friend Sarah about her (Betty’s) ex-boyfriend, who broke up with Betty not long ago. She explains to Sarah that she is both ‘angry’ and ‘sad’ about that fact that he left her: FREUNDINNEN 210 Betty: 211 212 Sarah: 213 214 Sarah: 215 Betty: 216 217 218 219 220 221 222 Sarah:

(2003_08_31freunde1_b,MÜNSTER) (.) . h=hm; (2.0) okE:, ja (3.0)

(.) (0.5) .h (.)

hm

176

Susanne Günthner

GIRLFRIENDS 210 Betty: 211 212 Sarah: 213 214 Sarah: 215 Betty: 216 217 218 219 220 221 222

(2003_08_31freunde1_b,MÜNSTER) (.) . h=hm; (2.0) okay, yeah (3.0)

(.)

(0.5) .h (.)

Sarah: hm

In line 217, Betty states that it is TRAUrig (‘sad’), that her friend left her and there is no more contact between them. By presenting a general rule she sets up a new context for her evaluative stance. The evaluative phrase (‘’) with the correlative es (‘it’) forms no complete sentence; its semantic subject is only provided in the following complement clause. Thus, the syntactic gestalt as well as the communicative action projected by part A is only completed with the production of the following segment (part B) (here: (.) (0.5) .h (.) ; ‘ (.) (0.5) .h (.) ’). Traditionally, extrapositions are supposed to consist of two parts:25 the preceding syntagma “es & predicate (…)” (part A) and the following part – a dependent clause, introduced by a complementizer (dass) showing final positioning of the verb; i.e. subordinate clause order. Close grammatical cohesion between the two parts is secured by the matrix clause and subordinate clause-patterning. According to Kay (2007: 4), extrapositions with es (‘it’) belong to the class of constructions, in which “single valence elements are realized as two different constituents of the actual sentence”. Analyzing the ongoing production of extrapositions with es in the process of interaction, however, we frequently encounter B-parts, which do not con25

Contrary to Couper-Kuhlen/Thompson’s (2006) findings on extrapositions in English, German data hardly shows any non-finite syntagmas in part B (Günthner 2009). As Couper-Kuhlen/Thompson (2006) show, 40 % of extrapositions with ‘it’ in English carry non-finite B-parts (“it’s pleasant to run” or “it’s time for me to become a priest”).

Between emergence and sedimentation

177

sist of a simple complement clause, as the canonical examples provided by generative grammarians and reference grammars suggest, but which – similar to uses of pseudo-clefts and die Sache/das Ding ist-constructions – form a complex pattern stretching over several clausal and prosodic units and expanding over various TCUs.26 Again, I want to argue that the construction of part B represents a dynamic process that extends in time. The next excerpt is taken from a telephone interaction between friends. Part A of the extraposition (l. 48) carries an evaluative expression with the modal particles halt schOn. Part B, however, is not realized by way of a subordinate complement clause but has a complex main clause structure and stretches over a longer discourse sequence with no clear boundary: VERLASSEN (EMOTIONEN 2; MÜNSTER 2006) 46 Sina: [ja- glaub ich gern. ] 47 Nine: [kannst auch (echt) ] 48 es is halt schOn (.) sch- be(.)SCHISsen, 49 weißt (.) ER meldet sich NIE; 50 ECHT NIE. 51 Sina: hm. 52 Nine: (ähm) un wenn ich ihn dann mal SEH; 53 ZUfällig, 54 Sina: [hm ] 55 Nine: [is-] isses auch hh’ zIemlisch

LEFT ALONE 46 Sina: 47 Nine: 48 49 50 51 Sina: 52 Nine: 53 54 Sina: 55 Nine:

(EMOTIONEN 2; MÜNSTER 2006) [yea- I can believe that. ] [you can (really) ] actually it’s pretty (.) aw(.)awful, you know (.) he never calls me; really never. hm. (ähm) and when I run into him; by accident, [hm ] [it-] it’s also quite



In line 48, Nine instantiates the first part of an extraposition (es is halt schOn (.) sch- be(.)SCHISsen,; ‘actually it’s pretty (.) aw(.)awful,’) projecting ‘more to come’. However, the expected content clause shows no sign of adhering to rules of syntactic embedding; the speaker continues with a grammatically non-attached main clause, a relatively independent unit weißt (.) ER meldet sich NIE; ECHT NIE. (‘you know (.) he never calls me; really never.’) (l. 49–50), which shows neither a complementizer nor does it have final positioning of 26

Cf. Couper-Kuhlen/Thompson (2006) and Hopper/Thompson (i.pr.) for similar results in English interactions.

178

Susanne Günthner

the finite verb. Thus, typical indications of syntactic subordination are absent. Contrary to the canonical form, part B takes on the shape of a syntactically independent unit. In line 52, Nine incrementally expands part B by adding the conjunction un (‘and’) and a following conditional construction: (ähm) un wenn ich ihn dann mal SEH; ZUfällig, [is-] isses auch hh’ zIemlisch

(‘(ähm) and when I run into him; by accident, [it-] it’s also quite

’) (l. 52–55). In studying extrapositions as they emerge in real-time discourse, we discover that they (much like pseudo-clefts and die Sache ist-constructions) frequently start with a framing element, followed by a rather complex segment whose end is open to negotiation. Also, in the case of extrapositions, the speaker draws the co-participants’ attention to the component to follow. Until part A is completed, the speaker has some leeway and time to decide whether the following part should be syntactically integrated or disintegrated. In the case of syntactical disintegration, the grammatical cohesion of the two parts is weakened. In the following transcript, taken once again from the reality TV-series “Big Brother”, part B covers a complex stretch of discourse, which expands over several TCUs. Harry, Frank, and another participant in the series are talking about good-looking people and the fact that beautiful people have better chances at finding employment: SCHÖNE MENSCHEN (BIG BROTHER bb2–16) 1 Harry: da ist das also (.) schon f- fAst erFORderlich heutzutage, 2 (.) 3 dass DU (.) GUTaussehend bIst. (.) 4 das gilt für MÄNNlein, 5 für WEIBlein, (.) 6 ds=gleichermaß=n 7 Frank: es kommt vO:r dass, 8 SCHÖne menschen [es: ] sEhr einfach haben, 9 ??? [((hustet))] 10 Frank: oder lEIchter haben als andere, 11 denn in viel=n dingen zählt auch der erste EINdruck, 12 das HEIßT, 13 wenn man [äääh ] allEIn durch das AUSsehen, 14 ???: [hehe ((lacht))] 15 Frank: durch ein attraktives AUSsehen, (.) 16 sich das- den Ersten EINdruck (-) 17 d-des gegenüber sIchert, (.) 18 dann hat man schon einen BOnuspunkt, 19 und kann- kann dArauf AUFbauen;

Between emergence and sedimentation

179

BEAUTIFUL PEOPLE (BIG BROTHER bb2–16) 1 Harry: it’s already (.) nearly necessary nowadays, 2 (.) 3 that you (.) are good-looking. (.) 4 whether you are male, 5 or female, (.) 6 doesn’t matter 7 Frank: it happens that, 8 beautiful people [it’s ] very simple for them, 9 ??? [((coughs))] 10 Frank: or easier for them than for other people, 11 because very often the first impression is what counts, 12 this means, 13 if you [äääh ] just because of the way you look, 14 ???: [hehe ((laughs))] 15 Frank: when you are good looking, (.) 16 this- impresses (-) 17 the other person, (.) 18 then you already have bonus points, 19 and can build on them;

In line 7, Frank supports Harry’s claim that nowadays it’s important to be good-looking, because beautiful people have major advantages. Part A of the extraposition (es kommt vO:r dass,; ‘it happens that,’) represents an evidential statement, projecting the following part (SCHÖne menschen [es: ] sEhr einfach haben, oder lEIchter haben als andere, …; ‘beautiful people [it’s ] very simple for them, or easier for them than for other people,’). However, part A does not only foreshadow ‘more to come’, it builds up grammatical expectations about the missing constituent: The subject of the intransitive verb is projected and thus, a specification of what actually ‘happens’. (In this utterance, the complementizer dass is prosodically part of A, i.e. in the same intonation with the evidential formula es kommt vO:r (‘it happens’).) Studying extrapositions with es (‘it’) in the course of their emergence in interaction reveals a much more complex picture than the assumption suggested by a bi-clausal sentence pattern with a grammatically marked relation of dependency between the two parts. Part A, which is lexically rather constrained and formulaic,27 takes over the interactive task of anticipating and 27

Part A provides evaluate, epistemic, or evidential expressions (such as es ist halt traurig, … (‘it is sad …’), es ist gut, … (‘it is good …’), es kommt vor, … (‘it happens …’), es ist herausgekommen, … (‘it turns out …’), es ist wohl möglich, … (‘it is possible …’), etc.). In our data, only a small variety of verbs is used: In the 30 extrapositions analyzed, we find 23 forms of the verb sein (‘be’), twice we find stimmen (‘turns out’), the following verbs appear once: passieren können (‘can happen’), sich zeigen (‘show’), sich treffen (‘turn out/happen’), herauskommen (‘come out’) und vorkommen (‘happen’). Cf. Günthner (i. pr.).

180

Susanne Günthner

framing the salient information, which is then provided in part B. Furthermore, part A helps the speaker to maintain the floor till the end of the construction. This turn-keeping function is particularly relevant in cases in which speakers produce longer stretches of discourse.

4

Conclusion

The constructions discussed reveal various formal as well as functional parallels. In analyzing them as they emerge in the process of interaction, they cannot be reduced to bi-clausal sentence patterns. Such a bi-clausal perspective seems to be based on the prevalent “written language bias” (Linell 2005) and the idea that ‘thoughts are expressed in full clauses and sentences’, which even today characterizes a great deal of linguistic analyses.28 Instead, in everyday interactions, these constructions are used as ‘projecting constructions’ (Hopper 2005; Hopper/Thompson i.pr.; Günthner 2006a, 2007, 2008a,b, 2009, Günthner/Hopper 2010). Projective constructions are interactive resources participants use to solve a multitask problem in the process of talk in time: They help them to organize their talk in terms of indexing communicative activities, integrating aspects of sequential context, temporality, dialogicity and strategic on-line management of interactional contingencies (Auer 2005; Günthner 2008; Günthner/Hoppper 2010; Hopper/Thompson i.pr.). Furthermore, projective constructions reflect the dialectic process of local emergence of grammar on the one hand, and participants’ orientation to sedimented patterns on the other: Even though part A opens a projection span, the projection can be fulfilled in a number of ways; i.e. part B is an emergent product of the locally managed interaction. It can take the form of a canonical pattern, as presented in reference grammars based on written standard German. More frequently, however, part B shows no sign of syntactic integration and stretches over longer segments of talk. In cases in which part A is positioned in front of a fullfledged main clause, the construction loses grammatical coherence; i.e. the two components are juxtaposed without any coding of grammatical dependency. Even though projection does not equal determination and part B is interactively negotiated, it does not mean that what follows is arbitrary either. By applying their knowledge of sedimented patterns of a particular language, re28

The structure of bi-clausal constructions with two full clauses fits the assumption that ‘complete’ thoughts represent propositions, and that propositions are expressed as clauses.

Between emergence and sedimentation

181

cipients can anticipate to a certain degree what might follow. As discourse unfolds through time, they are able to infer the eventual substance of the communication, and can process it more easily.29 Communicative patterns (from grammatical to textual formats) oscillate between processes of local, context bound emergence on the one side and sedimentation, stemming from repeated use in interaction, on the other. It is this dynamic interplay between habitual, routinized ways of communication and the improvised, contingent, and emergent features which characterizes communicative practice in general (Hanks 1996: 233ff.): Viewed from this perspective, the central project of linguistics would be the study not of ‘grammar’, but of ‘grammaticization’ – the ways in which some of the collectively possessed inventory of forms available for the construction of discourse become ‘sedimented’ through repeated use, and eventually are recognized as being to a greater or lesser degree ‘grammatical’. (Hopper 1992: 366f.)

References Akmajian, A. 1970 On deriving cleft sentences from pseudo cleft sentences. Linguistic Inquiry, 1, 2, 149–168. Andersson, S.-G., 1993 Zur Satzspaltung (Cleft) und langer Extraktion in germanischen Sprachen. In: M. Reis (ed.), Wortstellung und Informationsstruktur, Tübingen, 39–62. Auer, P. 2000 On line-Syntax. Sprache und Literatur 85, 31, 43–56. 2005 Projection in interaction and projection in grammar. Text 25, 1, 7–36. 2006 Construction Grammar meets Conversation: Einige Überlegungen am Beispiel von ‘so’-Konstruktionen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 291–314. 2007a Syntax als Prozess. In: H. Hausendorf (ed.), Gespräch als Prozess. Linguistische Aspekte der Zeitlichkeit verbaler Interaktion, Tübingen, 95–124. 2009 Projection and minimalistic syntax in interaction. Disc. Proc. 46, 2, 180–205. Auer, P. & S. Günthner 2003 Die Entstehung von Diskursmarkern im Deutschen – ein Fall von Grammatikalisierung?. In: T. Leuschner & T. Mortelmans (eds.), Grammatikalisierung im Deutschen, Berlin, 335–362. Bakhtin, M. M. 1978/86 The problem of speech genres. In: C. Emerson & M. Holquist (eds.), Speech Genres and Other Late Essays, Austin, 60–102. 1981 The Dialogic Imagination, Austin. Berger, P. L. & T. Luckmann 1966 The Social Construction of Reality, New York.

29

Cf. Auer (2009).

182

Susanne Günthner

Bergmann, J. 1987 Klatsch. Zur Sozialform der diskreten Indiskretion, Berlin. Birkner, K. 2008 Relativ(satz)konstruktionen im gesprochenen Deutsch: Syntaktische, prosodische, semantische und pragmatische Aspekte, Berlin. Blatz, F. 1886/1970 Neuhochdeutsche Grammatik mit Berücksichtigung der historischen Entwicklung der deutschen Sprache. Zweiter Band: Satzlehre, Hildesheim/New York. Bußmann, H. 2002 Lexikon der Sprachwissenschaft, Stuttgart. Collins, P. 1991 Clefts and Pseudo-cleft Constructions in English, London. 1994 Extraposition in English. In: E. Couper-Kuhlen, & S. A. Thompson (eds.), Functions of Language 1, 1, 7–24. 2006 You know, it’s funny: Eine Neubetrachtung der ‘Extraposition’ im Englischen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 23–58. Croft, W. 2001 Radical Construction Grammar: Syntactic Theory in Typological Perspective, Oxford. Deppermann, A. 2007 Grammatik und Semantik aus gesprächsanalytischer Sicht. Methodologischer Rahmen und exemplarische Untersuchungen, Berlin. Dik, S. C. 1997 The Theory of Functional Grammar. Part 2: Complex and Derived Constructions, Berlin/New York. Ehmer, O. & S. Pfänder 2009 Sprache kann in jedem Moment ganz anders sein. Improvisationstechniken im Gespräch. In: M. Gröne et al. (eds.), Improvisation. Kultur- und lebenswissenschaftliche Perspektiven, Freiburg, 175–195. Feilke, H. 2007 Syntaktische Aspekte der Phraseologie III: Construction Grammar und verwandte Ansätze. In: H. Burger et al. (eds.), HSK Phraseologie, Berlin, 63–76. Ford, C. E. 2004 Contingency and units in interaction.Discourse Studies 6, 1, 27–52. Franck, D. 1985 Sentences in conversational turns: a case of syntactic ‘double bind’. In: M. Dascal (ed.), Dialogue. An interdisciplinary approach, Amsterdam, 233–245. Fischer, K. & A. Stefanowitsch (eds.) 2007 Konstruktionsgrammatik: Von der Anwendung zur Theorie, Tübingen. Fried, M. & J.-O. Östman 2005 The cognitive grounding of Construction Grammar. In: M. Fried & J. O. Östman (eds.), Construction Grammars. Cognitive Grounding and theoretical extensions, Amsterdam, Philadelphia, 1–14. Goodwin, C. 1996 Transparent Vision. In: E. Ochs, E.A. Schegloff & S. A. Thompson (eds.), Interaction and Grammar, Cambridge, 370–404. Günthner, S. 2000 Vorwurfsaktivitäten in der Alltagsinteraktion. Grammatische, prosodische, rhetorischstilistische und interaktive Verfahren bei der Konstitution kommunikativer Muster und Gattungen, Tübingen.

Between emergence and sedimentation 2005

183

Grammatical constructions in ‘real life practices’: Wo-constructions in everyday German. In: Hakulinen, A. & M. Selting (eds.), Syntax and Lexic in Conversation. Studies on the use of linguistic resources in talk-in-interaction, Amsterdam, 159–184. 2006a ‘Was ihn trieb, war vor allem Wanderlust’ (Hesse: Narziß und Goldmund). Pseudocleft-Konstruktionen im Deutschen. In: S. Günthner & W. Imo (eds.), Konstruktionen in der Interaktion, Berlin, 59–90. 2006b Von Konstruktionen zu kommunikativen Gattungen: Die Relevanz sedimentierter Muster für die Ausführung kommunikativer Aufgaben. Deutsche Sprache 34, 1–2, 173–190. 2007 N-be-that-constructions in everyday German conversation: A reanalysis of ‘die Sache ist’ (‘the thing is’)-clauses as projector phrases, GIDI-Arbeitspapier (Working Papers in Grammar and Interaction), Münster 2007, http://noam.unimuenster.de/gidi/. 2008a Projektorkonstruktionen im Gespräch: Pseudoclefts, die Sache ist-Konstruktionen und Extrapositionen mit es. Gesprächsforschung – Online-Zeitschrift zur verbalen Interaktion 9, 86–114. http://www.gespraechsforschung-ozs.de/heft 2008/ga-guenthner.pdf. 2008b ‘Die Sache ist …’: eine Projektorkonstruktion im gesprochenen Deutsch. Zeitschrift für Sprachwissenschaft 27, 1, 39–72. 2009 Extrapositionen mit es im gesprochenen Deutsch, in: Zeitschrift für Germanistische Linguistik 37, 15–47. Günthner, S. & H. Knoblauch 1995 Culturally patterned speaking practices. The analysis of communicative genres. Pragmatics 5, 1, 1–32. Günthner, Susanne/Hopper, Paul 2010 Zeitlichkeit & sprachliche Strukturen: Pseudoclefts im Englischen und Deutschen. Gesprächsforschung – Online-Zeitschrift zur verbalen Interaktion 11, 1–28. (http://www.gespraechsforschung-ozs.de/heft2010/ga-guenthner.pdf). Günthner, S. & W. Imo (eds.) 2006 Konstruktionen in der Interaktion, Berlin. Hanks, W. F. 1987 Discourse genres in a theory of practice. American Ethnologist 14, 4, 668–692. Hanks, W. F. 1996 Language and Communicative Practices, Boulder CO. Hopper, P. 1987 Emergent grammar. In: Berkeley Linguistic Society (ed.), General Session and Parasession on Grammar and Cognition, Berkeley, 139–157. 1992 Emergence of grammar. In: W. Bright (ed)., International Encyclopedia of Linguistics. Vol. I, Oxford, 364–367. 1998 Emergent grammar. In: M. Tomasello (ed.), The New Psychology of Language, Mahwah, NJ,155–175. 2001 Grammatical constructions and their discourse origins: prototype or family resemblance?. In: M. Pütz, S. Niemeier & R. Dirven (eds.), Applied Cognitive Linguistics I: Theory and Language Acquisition, Berlin,109–129. 2004 The Openness of Grammatical Constructions, 40th Annual Meeting of the Chicago Linguistic Society, Chicago, 153–175. 2005. Bi-clausal Constructions and Emergent Grammar, Talk presented at the German Department of the Westfälischen Wilhelms-University Münster, April 2005.

184

Susanne Günthner

Hopper, P & Thompson, S. A. (i. pr.) Projectability and Clause Combining in Interaction. To appear in: R. Laury (ed.), Crosslinguistic Studies of Clause Combining: The Multifunctionality of Conjunctions, Amsterdam. Houtkoop, H. & H. Mazeland 1985 Turns and discourse units in everyday conversation. Journal of Pragmatics 9, 595–619. Hymes, D. 1974 Ways of speaking. In: Bauman, R. & J. Sherzer (eds.), Explorations in the Ethnography of Speaking, Cambridge, 433–451. Imo, W. 2007 Construction Grammar und Gesprochene-Sprache-Forschung. Konstruktionen mit zehn matrixfähigen Verben im gesprochenen Deutsch, Tübingen. Jespersen, O., 1927 A Modern English Grammar on Historical Principles, Part III, Heidelberg. 1937/65 Analytic Syntax, London. 1949 A Modern English Grammar on Historical Principles, Part VII, Kopenhagen. Kärkkäinen, E. 2006 Stance taking in conversation: From subjectivity to intersubjectivity. Text and Talk 26, 6, 699–731. Kay, P. 2007 IT-Extraposition, http://www.icsi.berkeley.edu/~kay/bcg/extrap.html, (accessed 5. 6. 2007). Knoblauch, H. 1995 Kommunikationskultur: Die kommunikative Konstruktion kultureller Kontexte, Berlin. Lambrecht, K. 2001 A framework for the analysis of cleft constructions. Linguistics 39, 3, 463–516. Lerner, G. H. 1987 Collaborative Turn Sequences: Sentence Construction and Social Action, Ph.D. thesis, University of California at Irvine. Lerner, G. H., 2002 Turn-sharing: The choral co-production of talk-in-interaction, in: C. E. Ford, B. A. Fox & S. A. Thompson (eds.), The Language of Turn and Sequence, Oxford, 225–256. Linell, Per 2005 The Written Language Bias in Linguistics: Its Nature, Origins and Transformations, New York. 2009 Rethinking Language, Mind, and World Dialogically, Charlotte, NC. Luckmann, T. 1986 Grundformen der gesellschaftlichen Vermittlung des Wissens: Kommunikative Gattungen. Kölner Zeitschrift für Soziologie und Sozialpsychologie, Sonderheft 27, 191–211. 1988 Kommunikative Gattungen im kommunikativen ‘Haushalt’ einer Gesellschaft. In: G. Smolka-Koerdt, G. Spangenberg & D. Tillmann-Bartylla (eds.), Der Ursprung der Literatur, München, 279–288. 1992 Theorie des sozialen Handelns, Berlin/New York. 2002 Wissen und Gesellschaft. Ausgewählte Aufsätze, Konstanz.

Between emergence and sedimentation

185

Macwhinney, B. 2001 Emergentist approaches to language. In: J. Bybee & P. Hopper (eds.), Frequency and the emergence of linguistic structures, Amsterdam, 449–470. Ono, T. & S. A.Thompson 1995 What can conversation tell us about syntax? In: P. W. Davis (ed.), Alternative Linguistics: Descriptive and Theoretical Modes, Amsterdam, 213–271. 1996 Interaction and syntax in the structure of conversational discourse. In: E. Hovy & Scott, D. (eds.), Discourse Processing: An Interdisciplinary Perspective, Heidelberg, 67–96. Paul, H. 1919/1968 Deutsche Grammatik. Band III. Teil IV: Syntax, Tübingen. Prince, E. 1978 A comparison of WH-clefts and IT-clefts in discourse. Language 54, 883–906. Ross, J. R. 2000 The frozenness of pseudo-clefts: towards an equality-based syntax. In: A. Okrent & J. P. Boyle (eds.), Proceedings of the 36th Meeting of the Chicago Linguistic Society, Chicago, 385–426. Sacks, H., E. A. Schegloff & G. Jefferson 1974 A simplest systematics for the organization of turn-taking for conversation. Language 50, 696–735. Schegloff, E. A. 1980 Preliminaries to preliminaries: ‘Can I ask you a question?’ Sociological Inquiry 50, 3–4, 104–152. 1984 On some questions and ambiguities in conversation. In: J. M. Atkinson & J. J. Heritage (eds.), Structures of Social Action. Studies in Conversation Analysis, Camebridge, 28–52. Schmid, H.-J. 2001 ‘Presupposition can be bluff ’: How abstract nouns can be used as presupposition triggers. Journal of Pragmatics 33, 10, 1529–1552. Stefanowitsch, A. & K. Fischer (eds.) 2008 Konstruktionsgrammatik II. Von der Konstruktion zur Grammatik, Tübingen. Streeck, J. 1995 On projection. In: E. N. Goody (ed.), Social Intelligence and Interaction. Expressions and Implications of the Social Bias in Buman Intelligence, Cambridge, 87–110. Thompson, S. A. 2002a Constructions and Conversation, Manuscript. University of California at Santa Barbara. 2002b ‘Object complements’ and conversation toward a realistic account. Studies in Language 26, 1, 126–163. 2008 What are clauses for? Understanding grammar in terms of social action, Manuscript, University of California at Santa Barbara. Thompson, S. A. & E. Couper-Kuhlen 2005 The clause as a locus of grammar and interaction. Discourse Studies 7, 4–5, 481–505.

186

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

Improvisation, temporality and emergent constructions*

1

Introduction

Hopper likens emergent grammar to artistic improvisation: Language is “improvised in much the same way that a jazz performance is improvised – jointly and publicly” (Hopper, this volume, p. 33). However, he does not clearly define the term improvisation itself. Our basic question is thus: What does improvisation mean in the domain of emergence and grammar and how does it further our understanding of grammatical constructions as processed in real time?

2

Improvisation

The concept of improvisation has long been limited to the study of performing arts, such as jazz and street theatre. In recent years, however, the term and concept of improvisation has increasingly been applied to other disciplines (cf. Kurt and Näumann 2008; Gröne, Gehrke, Hausmann, Pfänder and Zimmermann 2009; Zanetti 2009; Bormann, Brandstetter and Matzke, 2010). The term improvisation itself does not have its roots, as is often claimed, in Latin, but in relatinised Italian. More importantly, it did not occur in the nominative, but in the prepositional phrase all’improvviso (‘in the moment of the unforeseen’). It is this special property of merging into the here and now or, as musicians like to put it, the ‘magic of the moment’, which is characteristic of improvisation. * Among the people who gave us inspiration, encouragement and comments we would especially like to thank Peter Auer, Juan Ennis, Hans-Helmuth Gander, Hans-Joachim Gehrke and Martin Hilpert. Our understanding of the cognitive aspects of improvisation was deepened by the collaboration with Gerhard Strube, which was explored in a joint seminar and for which we are very thankful. Finally, we want to thank Carolyn Mackenzie and Elin Arbin for the revision of our English. We would also like to thank a number of institutions for their support. For enabling the interdisciplinary collaboration that formed the basis for this article we thank the Freiburg Institute for Advanced Studies (FRIAS). For long-term financial support, we are grateful to the Fritz Thyssen Foundation (Thyssen Grant “Emergente Konstruktionen”, Az.20.08.0.046).

Improvisation, temporality and emergent constructions

187

The here and now can be reconsidered for music because it is the contact with the emerging sound material that guides the improvisers (Wilson 1999; Essl 2004). The improviser is at a high level of attention and awareness; aware of the potentiality of the structures at every moment, conscious of all the actions that might follow and thus of the very sound material itself, also drawing on cognitive, affective, and social resources (Cunha, Cunha and Kamoche 1999: 3; Ehmer and Pfänder 2009), creating a unique situation of intersubjective attention (cf. Breyer 2011). Improvisation is about the choices1 agents make in the very moment, while the grounds for making these choices are not translated into forms of logical evaluation, but into “immediate action” (Eisenstein 1991: 399).2 The speaker’s thinking “is translated into immediate action, formulated not by formulas, but by form. […] Even in such ‘immediacy’ the necessary groundrules, justification and motivation for precisely that and not another disposition pass through one’s mind (sometimes even burst out in speech!)” (Eisenstein 1991: 399). When modelling improvisation for speech, we might thus be inspired by music, but also by theatre. In theatre, alongside with gaze and gestures (Sawyer 2001), language already has its place. In dance improvisation, the same element can be understood in one way at one moment and in a completely different way at the next moment (cf. Brandstetter 2009: 147; and also Lampert 2007). Improvisation has also been defined as “the degree to which composition and execution converge” in time (Moorman and Miner 1998a: 698; cf. also Moorman and Miner 1998b). Thus, the experience of grammar at times may even have an aesthetic, poetic dimension, which can only be described adequately by combining interactional and cognitive methods. As Brandstetter (2009: 147) goes on to explain, improvisation is a model for effects of emergence in that it constitutes a “Moment […] des größten Risikos, der Unbestimmbarkeit und Unvorhersehbarkeit. Er birgt die mögliche Entfaltung einer anderen, einer fremden Bewegung” [moment of greatest risk, of indeterminacy and unpredictability. It contains the possible unfolding of a different, foreign movement] (2009: 147). Whereas authors may disagree on whether the faculty of improvisation is biological (Biesenbender 1992), cultural (Kurt and Näumann 2008) or so1 2

Cf. Mitchell, Parker and Taborn (2004/05: 69). Eisenstein’s “Vertical Montage” was written in July and August 1940 and first published in Iskusstvo kino, no. 9, September 1940, 16–25; later in no. 1, January 1941, 29–38. It was translated as “Synchronization of Senses”, “Colour and Meaning” and “Form and Content: Practice”, Parts 2, 3, and 4 by Leyda (ed. and trans.), The Film Sense, 69–216.

188

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

cial (Günthner, this volume; Hopper 1998, 2004, this volume; Ehmer and Pfänder 2009; Pfänder and Skrovec, 2010), they all agree on a single point: Improvisation is an in-between phenomenon – or better, some sort of paradox. It is always situated between the known and the unknown, between planned action and unintended behaviour, between the unique and the routine. This, however, does not mean that things are learned, planned or schematised only to a small degree. Rather, it means that both occur at the same time: the expected and the unknown, the newly invented and the entrenched. In the following, we attempt to develop a conception of improvisation that is suited for the linguistic analysis of emergent grammar phenomena. It will be based on three seemingly paradoxical pillars: making and letting go (Section 2.1.), prefabrication and ad hoc montage (Section 2.2.) and a complex time frame, including the before, the coming soon and the very moment of now (Section 2.3.). 2.1 Making and letting go In art, the creative aspects of improvisation are always restricted by narrow boundaries in that the audience must be able to understand the improvised elements in one way or another. Improvisation has an intrinsically social character: Someone improvises for someone else and is thereby bound to the general law of typisation that is required to create something understandable (Raible 2009: 21). For instance, one cannot avoid conforming to the eight beats of the paso doble or the four beats of the jive in dancing. The beat provides something typical for each the dance, and not respecting it means leaving the realm of that specific dance. With reference to the arts, Raible proposes a rather narrow definition of improvisation, focussing on the intentional, individual making: “Improvisation hat … immer einen Urheber und ein Ziel, ein Telos. Ich möchte mit der Stegreifrede jemandem gefallen, mit der Stegreifkomödie mein zahlendes Publikum amüsieren …” [Improvisation always has an originator and a goal, target or telos. In an impromptu speech I want to impress someone, with an impromptu comedy I want to amuse my paying audience] (Raible 2009: 22). This is a point that has also been made by linguistic anthropologists under the term performance (cf. Hymes 1981; Baumann 1977). It is important to note that while emergence happens, improvisation generally has a very strong intentional component. Insofar as emergence represents a surplus that is created when a number of factors come together (in the case of a group brainstorming discussion, it is the individuals getting together), there is no such thing as a framework of expectations from a public audience: The group itself notes the unexpected

Improvisation, temporality and emergent constructions

189

result of the brainstorming which came out of an emergent process. This may or may not happen; it is unforeseeable in the Latin sense of improvisum (Raible 2009: 23–24). However, emergence and improvisation are not separated so clearly by all authors when it comes to the question of intentionality. For Brandstetter, whom we would like to follow in our paper, improvisation comprises both intentional and non-intentional aspects: “Improvisation setzt auf das, was in jedem Moment anders sein kann, oder anders gemacht wird” [Improvisation builds on that which can be different, or is made different, in any moment] (Brandstetter 2009: 147, translation and emphasis TB, OE and SP). From this quote, we can develop a further question, namely regarding the nature of the experience that enters into the act of improvisation – something that is already known, prefabricated and yet different from what was previously given. 2.2 Prefabricated, yet different It is essential for Hopper’s notion of emergence that speakers rely on prefabricated elements in their constructions. Improvisation as we understand it, however, employs elements that are made different compared to previous experience. To explicate this somewhat further, we draw again on Brandstetter, providing a slightly longer version of the citation we already referred to above: Improvisation setzt auf das, was in jedem Moment anders sein kann, oder anders gemacht wird […] Zwischen der Vorschrift […] und dem Chaos-Anteil, der durch Zufall, Unvorhersehbarkeit und Un-Ordnung […] einbricht, entspringt der Überschuss, der Zauber, die Spannung des […] Noch-nicht-erfahrenen. [Improvisation builds on that which can be different, or is made different, in any moment […] Between the prescript […] and the chaos part, which irrupts by means of chance, unpredictability and disorder […] there emerges the surplus, the magic, the tension of the […] not-yet-experienced] (Brandstetter 2009: 147, translation and emphasis TB, OE and SP).

There is a tension between the recourse to the experienced (prefabricated elements) and the not-yet-experienced of improvisation. The quality of the not-yet-experienced does not mean that there is a change in any substantial and enduring manner. It is sufficient that something changes and is different in the moment itself, in the context of its production. This tension can be understood and resolved if one views improvised speaking as a montage of knowledge of experience. The experience can relate to the larger course of the conversation (cf. the upcoming example zurückmalen), or the present moment within a conversation (example schnee).

190

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

Thus, improvisation is best described as a paradoxical configuration: It is at the same time intentional and non-intentional (Section 2.1.) and produces the not-yet-experienced with recourse to the previously experienced (Section 2.2.). This paradox does not disturb the ongoing conversation, but constitutes and fuels it. 2.3 Before, now and soon Let us refer one last time to Brandstetter’s definition by, which was already cited twice, and highlight still another aspect, i.e. the specific temporality of improvisation: “Improvisation setzt auf das, was in jedem Moment anders sein kann, oder anders gemacht wird” [Improvisation builds on that which can be different, or is made different, in any moment] (Brandstetter 2009: 147, translation and emphasis TB, OE and SP). Crucial here is the suddenness of the action – the magic of the moment evoked by many improvisers. To explain this point in the realm of on-line speech production, consider the well-observed practice that speakers go back and forth on the time scale (retraction and projection in the sense of Auer 2000, 2005, 2007; Auer and Pfänder 2007). This practice relies heavily on the speakers’ and, crucially, the hearers’ capacity to precisely remember (here also syntactically) what has just been said and to foresee what is about to be said. Despite this quasi-syntactical control of time by means of anticipation and expectation, the participants in conversations experience surprises in the form of unexpected continuations. Improvisation is always both – surprise and anticipated order: Bestünde sie nur aus Ordnung, also nur aus Vorhersehbarem, so würden wir sie als langweilig empfinden. Bestünde sie hingegen nur aus Überraschung, aus nicht Vorhersehbarem, so bedeutete dies für den Hörer reines Chaos, weil er das Neue nicht einordnen kann [Were it to consist solely of order, i.e. only of what is foreseeable, we would find it boring. Were it to consist solely of surprise, of what cannot be foreseen, this would mean pure chaos for the listener, since he cannot classify the “new”] (Raible 2009: 147, translation TB, OE and SP).

Crucial for improvisation is the effect of surprise created for the audience or – leaving the strict distinction between audience and improvisers behind – for all participants of the situation. Surprise is not only due to the quality of newness, but also to the temporal alignment of the surprising event. A potentially surprising improvisation may be merely odd if not produced at the right time. Improvisation involves a special awareness or increased sensibility of temporal alignment. The continuous building up and possible breaking with expectations can be observed for example in the realm of syntax, here in the linearity of speak-

Improvisation, temporality and emergent constructions

191

ing, manifesting itself in retractions, projections and choral productions, as we mentioned before. Auer’s approach of on-line syntax (2005, 2007) contributes to a better understanding of how the real-time emergence of syntactic projects is fuelled by the continuous projection of incremental syntactic activities which allow listeners to process the contribution of a speaker without delay in time, and proceed in discourse.3 If syntactic projections are possible on the basis of the gestalt principle of “good continuation”, as Auer (2007) emphasises, such gestalts can be interpreted as temporally extended linguistic entities whose comprehension and refiguration into alternative syntactic and/or semantic projects is dependent on anterograde memory. However, the increased sensibility for time may also lead to a momentary violation of the temporal conditions of speaking, i.e. the phenomena described by Auer. In improvisation, participants get into a so-called flow, which may be a joint rather than an individual experience. The term flow expresses the interactors’ experience of the present situation, but also the temporal conditions of being in the world, involving our sedimented experiences on different time spans. Improvisation is furthermore characterised by an enhanced attention of the speaker to language itself and to the other speakers or hearers, the audience: “‘Presence’ means the performer is fully there, ‘present,’ in the present tense, inside the moment. His attention is wholly on the task and yet, most importantly, his awareness extends beyond the immediate space around him to include the audience’s space” (Frost and Yarrow 1990: 100, emphasis TB, OE and SP; cf. also Breyer 2009). Improvisation thus means to be ready for what happens. Improvisation is a permanent acting and reacting (Mäder 1996). Being fully there, ready for what happens and well aware of the others: This here and now of improvisation is reminiscent of Husserl’s Erlebnisjetzt. According to Husserl, the present as it is experienced is not a singular point in time, but has an extension both into the past and into the future (cf. the analyses of the structure of internal time consciousness in Husserl 1991).

3

Projection can be understood as “the fact that an individual action or part of it foreshadows another. In order to understand what is projected, interactants need some kind of knowledge about how actions (or action components) are typically (i.e., qua types) sequenced, i.e., how they follow each other in time” (Auer 2005: 8). In a retraction, on the other hand, “a paradigmatic slot in an emergent syntactic unit is used twice, i.e. in formulating a next verbal component, the speaker re-uses a syntactic position which has already been filled by some other element before, and puts another element in this position” (Auer and Pfänder 2007: 60).

192

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

Jedes Erlebnisjetzt, sei es auch das der Einsatzphase eines neu auftretenden Erlebnisses, hat notwendig einen Horizont des Vorhin. Das kann aber prinzipiell kein leeres Vorhin sein, eine leere Form ohne Inhalt, ein Nonsens. Notwendig hat es die Bedeutung eines vergangenen Jetzt, das in dieser Form ein vergangenes Etwas, ein vergangenes Erlebnis faßt. Notwendig sind jedem neu anfangenden Erlebnis Erlebnisse zeitlich vorhergegangen, die Erlebnisvergangenheit ist kontinuierlich erfüllt. Jedes Ereignisjetzt hat aber auch seinen notwendigen Horizont des Nachher, und auch das ist kein leerer Horizont; notwendig wandelt sich jedes Erlebnisjetzt, sei es auch das der Endphase der Dauer eines aufhörenden Erlebnisses, in ein neues Jetzt, und auch das ist notwendig erfülltes (Husserl 1950: 199–200). [Each experiential Now, be it even the beginning phase of a newly appearing [experiential] process, necessarily has its horizon of Before. But of essential necessity that cannot be an empty Before, an empty form without content, a non-sense. Of necessity it has the signification of a past Now which comprises in this form a past something, a past experience. Every experience that has newly begun is of necessity temporally preceded by [experiential] processes; the past of [experiential] processes is continuously fulfilled. However, every experiential Now also has its necessary horizon of After, and that is also not an empty horizon; of necessity every experiential Now, even if it is the end-phase of duration pertaining to an experience which is ceasing, changes into a new Now, and it is of necessity a fulfilled one] (Husserl 1983: 195, translation F. Kersten, modified4 by TB, OE and SP).

Each experience is embedded in a temporal horizon of previous experiences (retention) and experiences yet to be made (protention). Both retention and protention remind us of Auer’s retraction and projection, respectively. But they go beyond the mere syntactic terms of on-line syntax. Retention and protention belong to the temporal structure of the lived present and allow for a coherent experience of temporal extension – be it a continuously present object of perception, an unfolding melody, a change of colour due to lighting conditions, a syntactic project in the making, etc. The retentional part of experience is the past, insofar as it is remembered in the present and has a thematic connection to what is currently cognised. The protentional aspect refers to the future as it is synthetically connected to the present in order for the stream of consciousness not to be halted abruptly. The protentional extension of the present into the future is necessary for the continuity of experience, i.e. there is always an element of expec4

Our modification of Kersten’s translation makes a crucial terminological distinction. Throughout this text Husserl focuses on experience. Thus the translation of Erlebnis as experiential process seems much more appropriate to us than Kersten’s translation as “mental process”, because this evokes a certain mentalism that is overcome by the phenomenological notion of the intentionality of consciousness, i.e. the directedness of experiences to objects.

Improvisation, temporality and emergent constructions

193

tation in the processing of experiential content: for example, an expectation of how an external object will look from another perspective in the course of perceiving it. In improvisational moments, the threefold structure of time in immediate experience – connecting the present, the past and the future – is highlighted. In other words, the present as it is experienced is already structured and does not exist as an ideal discrete point on an imaginary timescale. Once the experienced present is understood as a temporal duration in which retentional past, impressional now and protentional future are connected, one can consider the special prominence of each of the three constituting aspects in different types of actions and interactions. To sum up this section, we note that emergent grammar as conceived by Hopper emphasises the fact that something happens (and thus is emergent), that something prefabricated is employed and much of language production is expected and formally predesigned. In contrast, the characteristic aspects of improvisation lie in the tension between making and letting go (Section 2.1.), between what is prefabricated and arranged ad hoc (Section 2.2.), and between just before and immediately after the present moment (Section 2.3.) – polar ends that occur together in one jointly experienced moment of now. In other words, what happens in improvisation is not a clash of theoretical opposites, but the constitution of an experiential field of tension that has its own creative productivity. In the following, we will proceed by clarifying in more detail the complex term “emergence” with regard to grammar (Section 3). Thereafter, we shall explore two case studies – one of improvised morphosyntactic montage (Section 4) and a collaborative montage of a clausal syntactic project (Section 5).

3

From emergence to improvisation

The notion of emergence stems from the natural sciences. The following definition is widely used: Emergence is the process that leads to the coming into being of an over-summative quality which cannot be explained by the properties and reactions of the involved elements alone, and which is explainable only through the particular self-organising processual dynamics in each individual case, i.e. by regarding multiple factors of varying importance that occur either sequentially or simultaneously throughout the process. In linguistics, the notion of “emergent grammar” has been popularised by the writings of Hopper (cf. Hopper 1987, 1992, 1998, 2001, 2004, 2005). For emergent grammar, Hopper focuses especially on the second of the two

194

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

properties mentioned above, considering language in the process of its usage at a given moment according to the situational and processual dynamics of a particular conversation. Emergent grammar theory thus differs from a priori theories of language in that it “makes the sign itself subject to the exigencies of communication and assigns ontological priority to the fact of communication itself ” (Hopper 1998: 157). In emergent grammar, the adjective “emergent” designates “a continual movement toward structure, a postponement or deferral of structure, a view of structure as always provisional” (Hopper 1998: 157). In other words, Hopper claims that a structure that is emergent is “never fixed, never determined, but is constantly open and in flux. The term emergent refers to the essential incompleteness of a language, and sees lability between form and meaning as a constant and as a natural situation” (Hopper 1998: 157). For Hopper, then, “systematicity […] is an illusion produced by the partial sedimentation of frequently used forms into temporary subsystems. Grammar is a vast collection of such subsystems” (Hopper 1998: 158). In recent publications (e.g. this volume), Hopper frequently refers to the theoretical work of Giddens and emphasises the similarity of the concepts of “emergence” and “structuration” (Giddens 1984). Structuration in Giddens’ sense is the interplay and interlacing of current social action and social structures. In the course of this interplay over time, meaning and structures are constantly changing, although a primary continuity of the interplay itself is presupposed in order to keep the development going; in contrast, many linguists and theorists of other social disciplines would accept social structures as sometimes moving, and thus emerging; only few would see structure as something inherently emergent. In our view, the research interests of emergent vs. emerging theorists of grammar are complementary and can be schematised as follows: a) Emergent theorists explore how (fragments of) form-function units pass from talk to talk, presuming that grammar exists as (many) grammars, differing according to communicative genres or activity types. Emerging theorists, on the other hand, explore the cognitive foundation of constructions, presuming that frequency leads to entrenchment via processes such as analogy. b) Emergent theorists concentrate on the process whereby new combinations of (previously heard) forms are made in interactive encounters. They want to know how speakers deal with (old) fragments. Emerging theorists, in contrast, are interested in the result, i.e. new form-function units, newly categorised elements and re-organised systems. They aim to describe and explain (new) categories.

Improvisation, temporality and emergent constructions

195

c) Emergent theorists are intrigued by how speakers create grammar as they go along and how they use linguistic resources in concrete situations. They thus focus on the present moment, the événement. Emerging theorists, on the other hand, explore languages as self-organising, complex systems and sometimes postulate an a priori grammar. Hence, their focus is on the conjoncture or the longue durée. One of the strong points of emergent grammar is that it accounts for the sequentiality of speech production in real time. Hopper’s application of the emergence paradigm to linguistics has opened the way to fruitful analyses of syntax in real time, or on-line syntax (Auer 2000, 2009). Substantially more so than other approaches to syntax, emergent grammar focuses on formulas speakers have in their repertoire. Grammar, then, has only an epiphenomenal status; it is no more than the inventory of constructions speakers employ to put ready-mades or prefabricated elements into their ongoing speech activity. What we propose here, however, is to zoom in deeper on the temporal structure and to look closer at the very moment of speech production and perception. We would like to add two aspects to the term “emergence”, namely simultaneity and over-summativity. For emergent constructions in everyday conversations, we find that – when co-constructing the actual conversation in real time, speakers do not only choose ready-made constructions from a ‘basket’ of previously heard utterances. Rather, they go beyond the copy-and-paste activity by combining fragments of constructions or applying frequency-based combinational strategies, such as analogy. The result of this activity is known as montage, i.e. the leap into a new quality. This new quality might be a new form-function unit or construction that has never been heard before and perhaps will never be taken up in the ‘basket’ of ready-mades. We claim that this emergence of new qualities is always over-summative (cf. Section 4). – certain phenomena of on-line syntax can only be explained by going beyond mere sequentiality and also looking at simultaneity, e.g. in moments of choral turn-taking. Due to their ongoing projections, speakers sometimes tend to finish an ongoing syntactic project simultaneously. Hence, we refer not only to the individually rooted cognitive presence of a grammatical structure in conversation, but to the simultaneous presence of a grammatical structure and speakers’ making use of this fact (cf. Section 5). In line with a concept of emergence which encompasses both emergent and emerging characteristics, linguistic structures are to be understood not as unities of form and function, which can be retrieved from a grammar that exists a priori, but as the material that is found and picked up while speaking in

196

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

the communicative sphere between participants in interaction. This does not mean, however, that speakers modify the material found in the communicative sphere at every moment. In everyday speech, actors are rather ‘inert’ – they use the material that was just heard before (Szmrecsanyi 2006). Modifications of the linguistic material itself rarely occur. In certain moments, however, it happens that participants in a conversation play with the potential of emergent constructions and use the material that was just heard in a modified way – they improvise. The following analyses are based on different corpora of spontaneous everyday conversations. We consider mainly moments of collaborative storytelling (cf. Bochner and Ellis 1995; Lerner 1992, 2004) in which the collaboration among the speakers and the joint production of a narrative moment seem particularly important. In our analysis we will not focus on one specific construction or a net of constructions. Rather, we apply the concept of improvisation to conversations to shed light on the way speakers handle emergent constructions in conversations. In examining our corpus5 of collaborative storytelling, we identified two techniques of improvisation in the emergence, i.e. improvisations are emergent in the double sense of developing in real time and having unexpected rather than well-established qualities. We will give an example of the first technique of improvising, namely blending various construction(al fragment)s (Section 4), and an example of the second technique of improvising which consists of isolating one constructional fragment so that it becomes a highly available and variable syntactic pattern (Section 5).

4

Improvising by means of blending various construction(al fragment)s

In this paragraph we will discuss an example from Ehmer and Pfänder (2009) and Ehmer (2011) that shows how an unexpected montage in word formation is improvised in collaborative storytelling. At the beginning of the example, M tells the others that he was portrayed by other participants in a university seminar. We first provide some conversational background and then deliver the improvisational analysis.

5

Ehmer and Pfänder, Schall (www.romanistik.uni-freiburg.de/ehmer/schall/); Pfänder, Corpus del Castellano Cochabambino, CCC (www.espanoldelosandes.org); and Ehmer, Corpus de Conversaciones ESpontáneas PLAtenses, CESPLA (www.cespla.de). For the ease of illustration we will only refer to German examples in this article.

Improvisation, temporality and emergent constructions

197

Example: zurückmalen 01

M:

jetz muss=ich mal kurz was SAGn.= ‘now I have to say something quickly’

02

ich war ich war (.) ich war (.) ich BIN‘I was I was I was I am’

03

das hat jetz nicht mit den mEdien=mehr zu tun SONdern, ‘this now has nothing to do with the media anymore, but’

04

es gIbt, ‘there is’

05

äh bei den (-) (.) ein (dings) des=is (-) um iRAN. I have a (thing/seminar) which is about Iran’

06

und die ham Irgendwie (.) das faible dass die die lEUte alle ABmaln. ‘and they somehow have the penchant for drawing all the people’

07

[wir sitzen da im krEIs und das]=[is] ‘we are sitting there in a circle and that is’

08

U:

[HUAH::::‘huahhh’

]=

09

=[das=is] bei uns AUCH, ‘that is the same with us’

10

[!YE:AH!. ‘yeah’

]

11

M:

[al=das=is dOch-] ‘but that is so’

12

U:

ge[!NAU:!.] ‘exactly’

13

M:

[also ] vor=allen dingen das war jetz !E!cht ‘so this above all this was now really be!SCHEU!ertdaft’

14

15

ich wurde nämLich von drei leuten glEIchzeitich AB[gemalt.] ‘because I was drawn by three people at the same time’ All:

[((lachen))] ((laugh))

M lays out that he was portrayed in the seminar and evaluates this negatively (13). In the further course of the narration, he emphasises that he felt uncomfortable being drawn. Therefore, he himself began to draw the people who were drawing him. In the narration he describes his own drawing as an attempt to defend himself. But his drawing caused a disturbance in the seminar, so he stopped. Therefore, M’s attempt to defend himself eventually failed.

198

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

Before we come to a further analysis of the sequence in terms of improvisation, it is important to note that, like M, the collaborating speakers use the verb malen – which corresponds to the semantic frame of drawing – extensively throughout the whole sequence: – und die ham Irgendwie (.) das faible dass die die lEUte alle ABmaln. (06) – ‘and they somehow have a penchant for drawing all the people’ – ich wurde nämLich von drei leuten glEIchzeitich ABgemalt. (14) – ‘because I was actually drawn by three people at the same time’ – die maln sich Alle gEgenseitich AB (27) – ‘they all draw one another’ – ich hab dann AUch so getan als ob=ich ABmale, (44) – ‘then I acted like I was drawing, too’ – dass wir uns also gegenseitig ABmalen- (55) – ‘that we thus draw each other’ – das=is doch Echt ne k totAl bescheuerte (.) beschäftigung sich da: gegenseitig ABzumalen. (65) – ‘that’s really a totally daft activity there to draw each other’ – da müssn=se dich halt MALN. (104) – ‘there they just have to draw you’ – die maln ja nich nur MICH=ab; (106) – ‘they don’t just draw me’ – die maln sich ja auch (.) gEgenseitig AB. (108) – ‘they also draw each other’

In German, verbs can be combined with particles, which in some tenses stand together with the verb stem and in other tenses are separated from the stem. We refer to the underlying construction of morphological word formation as particle-verb construction. Here, the construction is applied in a canonical form to the verb malen, using the particle ab. The semantics of the resulting verb abmalen differs from the simple verb malen insofar as it denotes the depiction of a real, present object or person. In their contributions to the narration, the other speakers start to make fun of M: Maybe he did not feel so bothered at all, but was flattered by the fact that the female students had chosen to portray him in their drawings? After some turns of teasing, M stresses again that he felt incommoded, summarising the unsuccessfulness of his attempt to defend himself. 133 M:

naja du kAnnst ja (.) man kann ja nich zuRÜCKmaln. ‘well you can’t one can’t draw back (at someone)’

134 All: ((lachen)) ((laugh))

Improvisation, temporality and emergent constructions

199

In his utterance, M creates the neologism zuRÜCKmaln. To do this, he uses the standard morphological particle-verb construction and malen (‘draw’) in the verbal slot, just as before. In the particle slot he uses zurück (‘back’), which is productive in several German particle verbs, but unusual in combination with the verb malen. With the neologism zurückmalen (‘to draw back’) the speaker creates a verb that expresses his action of drawing as well as his intention of defending himself. Emergent grammar sensu strictu does not seem to help here, because zurückmalen is a verb that is unlikely to ever have been heard before. It is constructed by analogy and presupposes knowledge of routines of categorisation. This neologism at first provokes collective laughter (134) and is then directly picked up by speaker E. 135 E:

ab heute wird zuRÜCKgemalt. ‘from today on we draw back [at someone]’

136 All: ((lachen laut)) ((laugh loudly)) 137 K:

138 All: ((lachen laut)) ((laugh loud))

With his utterance ab heute wird zuRÜCKgemalt (135), speaker E not only picks up the neologism created by M, but integrates it into a larger structure: ab heute wird zurück+PARTICIPLE (‘from today on we VERB back’). By using this structure, he establishes an intertextual reference to Hitler’s speech on the occasion of the invasion of Poland on September 1, 1939 (Seit 5.45 Uhr wird jetzt zurückgeschossen! ‘Since 5:45 this morning, we are shooting back!’). This reference is not only made by way of the segmental/syntactical string; the speaker also prosodically mimics Hitler’s style of speech. We can tell by their loud collective laughter that the interlocutors understand this intertexual reference. They are amused by the way in which E blends the context of the narration (drawing in a seminar and M’s intention of defending himself) with the context of war and, more specifically, with Hitler’s aggression.6 Formally, the blend is based on the structure ab heute wird zurück+PARTICIPLE (also including prosody!), which is clearly part of the communicative memory of German speech styles.

6

For a detailed analysis of the cognitive blends created in this sequence, see Ehmer (2011), who applies the cognitive framework of blending theory/conceptual integration networks as developed by Fauconnier and Turner (1998).

200

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

This structure cannot be considered a mere phraseologism, but rather a construction that has become productive. As a Google search for ab heute wird zurück+PARTICIPLE shows, there are about 711 hits of variants of the phrase, such as ab heute wird zurückgefilmt (‘filmed back’), … zurückzensiert (‘censored back’), … zurückgekocht (‘cooked back’), and … zurückgeliebt (‘loved back’), which means that the constructional scheme ab heute wird zurück+PARTICIPLE is also part of the communicative memory. The productively used construction alludes to the context of war and is used to apply the semantic relation of aggression to the semantic frame denoted by a verb, which itself does not per se express aggression, like kochen (‘to cook’) or lieben (‘to love’). Let us now reconsider the sequence in terms of improvisation. 4.1 Making and letting go The speakers in the example are keen on making fun in and of the situation. They try to develop jokes that provoke laughter amongst the others who serve as an audience. They also try to tease and make fun of M. This is visible in the fact that N, for example, picks up a part of M’s speech (for which M has the “authorship” in Goffman’s sense) and blends it with a string whose authorship still somehow lies with Hitler, and thus attributes a highly exaggerated and aggressive behaviour to M in the seminar context. However, the interlocutors also let the situation unfold by itself. Throughout the sequence, M has a distinguished position: He is the main narrator of the story and pushes the conversation forward. The other interlocutors (in the role of making jokes and teasing M) are just waiting for the moment to join in and crack a joke. They do not force these moments, but they are alert and ready for it. This also holds true for the main narrator: He is open to let the audience contribute to the narration and also gets into the flow of his own improvisation. 4.2 Prefabricated, yet different Two speakers use constructional schemes to create emergent constructions. Speaker M uses the standard particle-verb construction to create the neologism zurückmalen. Speaker E repeats the verb that has just been coined and applies a further constructional scheme that has already become productive (ab heute wird zurück+PARTICIPLE). One could also say that he repeats this construction and reactualizes the context associated with it. The use of these constructions is what we would call fabricatedness. The newness in our example consists not only of the production of novel items

Improvisation, temporality and emergent constructions

201

such as a neologism. Rather, the participants experience these elements as something unexpected and express their joyful surprise with laughter. The constructions themselves are well prepared for the need to create new items: Both have an empty slot. But this slot is filled in an unexpected manner, albeit in concordance with the local semantic qualities. The unexpectedness and novelty of these elements is a central part of the conversation and makes the moment enjoyable. The cognitive strategy of creating analogies explains why this surprisingly new element is so quickly understood by all participants and why they are thus able to pick up the strategy and build new analogous structures. In other words, the novel fits into the well-known pattern by way of analogy. This can serve as an argument against the so-called “Lego”-principle, because speakers do not only put together what they have previously heard. Rather, they are pattern seekers, as Bod (2006) puts it. They love to seek, apply and generate patterns. 4.3 Before, now and soon In our example, different discursive patterns and discursive memories (that which has just passed in the current conversation – historical context) as well as mental images (situation of the seminar – Hitler speech) are superimposed and interlaced. Retrospectively, M’s narration receives a completely different meaning. Furthermore, the utterance ab heute wird zuRÜCKgemalt can no longer be attributed to a single speaker, but has become polyphonic in Bakhtin’s (1976/86, 1981) sense of the term. This polyphony and the opening up of the connected semantic spaces continue throughout the course of the conversation. For instance, at a later point of the conversation, when the topic changes to people who colour the grids of their writing pads in seminars, one speaker evokes the image irgendwann schlagen die kÄstchen nämlich Auch mal zuRÜCK (‘at some point the squares will also fight back’). Both the transfer of the antagonism of the seminar situation and the lexeme zurückschlagen (‘fight back’) evoke a military context. The repetition of linguistic elements takes them out of their original context. This variation and interlacing with new contexts and the creation of new hybrid contexts is a technique of improvisation. With this observation in mind, we can now refer back to the two points mentioned above. We suggest that speakers are very alert, due to the paradox of making and letting go, and they are very much aware of construction principles because of the prefabricated-surprise paradox. In this changing twofold motivation, very different layers of the communicative memory are potentially immediately accessible. In addition, we argue that they are very easily ac-

202

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

cessible. In this example, we see how the speakers make use of structures and knowledge that relate to very different time spans: – preceding turn: N repeats M’s neologism; – conversational activity/memory: 1) The neologism serves to elaborate M’s point of the story: that he intended to defend himself. 2) The speakers constantly repeat the verb abmalen and also the constructional scheme of particle-verb, which is then employed again; – historical context (Hitler speech): here it is obvious how the cultural context matters for the analysis of constructions; – the temporal dimension of language: particle-verb construction. These different time spans are brought together in improvisational moments. By using a grammatical construction, speakers reactualize past experience(s) bound to a construction. This can be understood as part of the process of entrenchment (individual, long-term). But the example analysed above illustrates that this reactualization of experience can also involve a re-presentation of a situation or a re-performance of a historical speech, which is then used subsequently as well. The experience bound to a construction is not only activated but is performed and presented to the hearers in such a way that they can savour and re-experience moments of language use. Thus, grammatical constructions should be treated as sediments of experiences, with quite different time depths and structural levels concerning form and content/function. This is very much related to a stance Günthner takes: Statt von starren, fixierten Form–Funktionspaaren als mentalen Konzeptualisierungen, die dann im konkreten Diskurs aktualisiert werden, auszugehen, scheint es nahe liegender, Konstruktionen als Orientierungsmuster zu betrachten, deren Instantiierung im Hier-und-Jetzt der Interaktion erfolgt. Interagierende stellen somit Beziehungen zwischen der sich momentan entfaltenden Interaktion und vergangenen Erfahrungen her, indem sie sedimentierte Muster re-aktualisieren. [Instead of assuming rigid, fixed form–function pairs as mental conceptualisations, which are then reactualized in concrete discourse, it seems to be more appropriate to consider constructions as patterns of orientation, whose instantiations occur in the here-and-now of interaction. Interactors thus create relations between the currently unfolding interaction and previous experiences by re-actualizing sedimented patterns] (Günthner 2007: no page numbers, translation TB, OE and SP).

As Günthner points out, the use of a grammatical construction reactualizes past experiences (of language use). Moreover, grammatical constructions should not be considered fixed form–meaning pairings. Instead, constructions should be seen as patterns to which speakers (only) orient themselves and which are flexible at the speakers’ disposal. We saw above how speakers modify constructions and make flexible use of them. We argue that this is

Improvisation, temporality and emergent constructions

203

generally the case, not only in moments of improvisation. In such moments, however, this fact becomes very obvious. To sum up the analysis of improvisation provided in this section: First, there is the constitutive tension of making and letting go. One participant is allowed to narrate and is accompanied into the imagined scene by the others. Even though the others let things happen in the course of this narration, they seek short moments in which they attempt to transcend the scene with humorous suggestions. This is sometimes done by way of overemphazising a certain aspect (e.g. being drawn, defending oneself), which can then be jointly used for purposes of amusement. Things are tried out which are not predefined in the systematics of the frame. This is not only true, however, for the semantically defined mental spaces, but also for the employed syntactical patterns. Two constructions (zurück + INFINITIVE and PARTICLE + malen) are blended and put together in a montage that draws on two sources. In a second step, this very recent and improvised material is used further and combined with the speakers’ knowledge on different levels of sedimentation. Both the formula ab heute … (‘from today on …’) and its communicative uses in the recent past are reactivated and applied. If we now go one step further, reactivation or reactualization is not only bound to historical moments (hence overt intertextual references) or abstract grammatical structures like the formation of particle verbs in German (e.g. zurückmalen, ‘to draw back’). The reactualization of a grammatical construction can also happen within a single moment in the conversation. We will now analyze an example in which the construction possesses only a very limited time depth: The construction is being created ad hoc in the moment of the conversation. Speakers repeat the speech string and attach a specific meaning to it. The string thus becomes and is made a prefabricated element in conversation and is used and reused in the course of the conversation.

5

Improvising by isolating one constructional fragment which thereby becomes a highly available and variable syntactic pattern

The upcoming example again illustrates how improvisation is done in collaborative storytelling. It is quite similar to the example discussed above, insofar as the participants compete in making humorous contributions to the main story. However, it differs from the previous example because the speakers do not rely so much on a construction that is somehow basic to the German language or anchored in the communicative memory of a social group. In the previous example, the neologism was quite surprising, but its creation was based on a very old morphological pattern/construction. In

204

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

contrast, in the upcoming example the “construction” that is used for improvisation is created ad hoc within the conversation and lacks the moment of surprise. The example therefore much better illustrates that the first step of improvisation might be the making of fragments for further use by isolating, highlighting, i.e. making salient and recognisable a constructional fragment. Example: Schnee In the following sequence, H narrates a story about a budgie his family had when he was a child in the German Democratic Republic and how they got it. Budgies were not very common in that part of Germany, but at some point H, while riding his bicycle in a snow-covered landscape, happened to find a budgie that had flown away. What is crucial for the understanding of this sequence is that H narrates the story twice. During the first narration, only two young women, T and S, are in the room. A fourth person, C, is not in the room and does not hear the story. As C enters the room, the participants are talking about the fact that budgies are very colourful, and T utters that therefore ‘they attract attention when they lie in the snow’: die fallen AUf wenn=sie=im SCHNEE liegen. As C has not heard the complete story of the budgie, he only gets the image of a budgie lying in the snow. He evaluates this image with astonishment: was=is denn DAS | im SCHNEE liegen | OH: (‘what is that | lie in the snow | oh’). It is crucial for the further course of the conversation that C directly picks up the string im schnee liegen (‘lie in the snow’), which was used before by another speaker (T). To make things very clear, C isolates one segment of the sound chain and does not choose die fallen auf (‘they stand out’), what he could have done, but instead chooses im Schnee liegen (‘lie in the snow’) without syntactic integration and with its own prosodic contour: im SCHNEE liegen. The string is located between two elements which mark the astonishment and serve as a bracket: was=is denn DAS on the left and oh on the right. This has an effect of isolating, highlighting, i.e. making salient and recognisable the constructional fragment. The fragment itself, however, does not only consist of a condensed form. In this particular context, it also involves the peculiar semantics of ‘a bird lying in the snow’. After this utterance, S joins in and evaluates the image as mean and nasty: is=das FIE:S (‘that’s mean’). With this evaluation, she is trying to work out the precarious aspect of the image. Her evaluation implies that H somehow caused the fact that the budgie was lying in the snow: He must have behaved in a nasty way. Evidence for this analysis is found in the subsequent course

205

Improvisation, temporality and emergent constructions

of the conversation, where H starts to justify himself and begins to renarrate the story to put things right. It is crucial for the understanding of the following sequence that H wants to tell the correct version of the story, while the others (especially T) try to make fun of this activity and the image of a budgie lying in the snow. Again, we first provide a sequential analysis and then provide a reanalysis in terms of improvisation. 01

H: nEE das [war; ‘no, it was’

]

02

S:

]

03

[(HM)‘hm’

nei [n das [wAr (nur so)- ] ‘no, it was (just like)’

04

H:

05

T:

06

S:

[so hatt ich [den gefUn][den ‘I found him (our bird)like that’ [ne SOCke; ‘a sock’

[(von UNS);

]

] [is von [;] Holger’ 07

C:

08

H: ‘ehm we had’

09

T:

10

H:

[ECHT‘really’

12

]

[ach der] [(.) (is) kaPUTT, ] ‘oh it’s broken’ geKAUFT -= ‘or anything’

f 11

[wir hatt=n den ja]=nich ‘we had not bought him’

=ich bin halt (.) kurz vor wEIhnachten irgendwie mit=m fAhrrad? geFAHRen ‘I was just, shortly before christmas, somehow riding my bicycle and there/then’ T: ((räuspert sich)) ((clears her throat))

f 13

S: [(...) lAg der im SCH[NEE;] ‘he was lying in the snow’

f 14

H: [so (.) ‘just

f 15

T:

f 16

H: = un da:nn WAR das dE[r. ‘and then that was him’

17

]

C:

nebenb[EI: ] (.) was bUntes im schn]EE geSEHN;= by chance saw something colourful in the snow’ [lAg ] ma(l) was im SCHNEE. ] ‘something was casually lying in the snow’ ]

[]ah. ‘whoa’

206 18

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder H: ; ‘yes’

H starts to retell the story, but he is interrupted by the others. T utters ne SOCke (‘a sock’, 05), which is meant as a joke. She animates the voice of H as a child who, riding his bicycle, takes the budgie in the snow for a sock (children’s socks are generally colourful, like budgies). The comical strategy she employs here is to de-animate the animal and refer to it as a thing. S joins in laughing and identifies the child H as the owner of the sock: is von HOLger (‘it’s Holger’s’, 06). T elaborates on this and evokes another image of H as a child examining the ‘sock’: ach der (.) (is) kaPUTT (‘oh it’s broken’, 09). This is also meant as a joke, because for a long time H does not recognise the bird as a bird, but still refers to it as a broken “thing” – one would not normally refer to animals as ‘broken’ (kaputt) but as ‘hurt’, ‘injured’ or ‘dead’. While S and T are busy developing a humorous co-narration, H tries to tell the true story. Since he does not receive the others’ attention, he has to restart his syntactical project several times (01, 03, 04 and 08, 10). However, when he finally manages to regain the floor (10, 11), he does not complete his utterances semantically and syntactically. Instead, he opens a strong projection at the end of utterance 11: un(d) da (‘and there/then’). This projection is completed in the subsequent utterances in a fourfold fashion. The first completion is done by S: lAg der im SCHNEE (‘he was lying in the snow’, 13). This completion is syntactically correct, and S is faster than H in finishing H’s syntactic project. The second completion is done by H, who completes his own utterance: so nebenbEI: (.) was bUntes im schnEE geSEHN (‘just by chance saw something colourful in the snow’, 14). Surprisingly, H does not complete his previous utterance in a syntactically correct way. H narrates the story in the perfect tense and uses the verb sein as an auxiliary (bin gefahren ‘was riding’, verb of motion) in 11. In utterance 14 (the continuation), H uses sehen (‘see’) as a main verb, which requires haben as an auxiliary. Only if the auxiliaries in two subsequent sentences are identical can the auxiliary be omitted in the second sentence. However, H does not change the auxiliary from haben to sein, which he should do according to normative grammar.7 This lack of the auxiliary only becomes obvious in the transcript, not when one just listens to the recording. The ungrammaticality probably also remains unnoticed for the participants in the conversation. 7

The utterance so (.) nebenbEI: (.) was bUntes im schnEE geSEHN can also be seen as a dense construction, in the sense of Günthner (2006), which speakers use in narrations to create a climax. Such constructions are characterized by the fact that speakers employ infinitives (or no verbs at all).

Improvisation, temporality and emergent constructions

207

The third completion is done by T: lAg ma(l) was im SCHNEE (‘something was lying in the snow’, 15). This continuation is again syntactically “correct” and almost identical with the preceding completion of S, who said: lAg der im SCHNEE (‘he was lying in the snow’, 13). With the use of the definite masculine article der, S refers directly to the bird (der Vogel). T, however, employs was (‘something’) as a deictic element to refer to the bird. T sticks here to the punch line of the joke she made before in which she referred to the bird as an inanimate object (a sock): A deictic was can only be used in German to refer to inanimate objects, but not to animate objects such as birds. Finally, the fourth completion is again done by H. This time, he not only completes his utterance, but retraces back to the conjunction und in utterance 11 and repeats the temporal connector da/dann: un da:nn WAR das dEr (‘and then that was him’, 16). This concludes our sequential analysis; in the following, we move on to an improvisational perspective and interpretation. 5.1 Making and letting go As mentioned before, H has to restart his narration several times because the interlocutors, in the mood of teasing and joking, intend to make fun of the situation and try to present their versions of the story. When H, in spite of the others’ interventions, finally gets to narrate, it is clear for S and T that H will tell the story again quickly in three steps, just like he did before: 1. I was riding my bike, 2. saw something lying in the snow and 3. then realised that it is a budgie. Thus, all the participants are very alert, ready to contribute and aware of the upcoming, i.e. emergent, syntax. This on the one hand creates a flow, the speakers let go, and on the other hand, at the same time they are alert to seize upcoming opportunities and to continue at possible points. So far, the setting is one of collaborative making and letting go, just like in the example zurückmalen. In this case, however, things happen much more quickly, and there is a lot of simultaneous and overlapping speech. 5.2 Prefabricated, yet different One aspect of our analysis that was mentioned several times has to be dealt with separately in this final section, namely temporal sequentiality. The contributions of the three speakers in this example overlap; they are uttered in

208

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

extremely rapid sequence and partly simultaneously. In the analysis, we have stressed that the speakers act playfully and competitively and do not allow H to finish narrating his story. In doing so, the speakers have different goals. However, they use the same or very similar linguistic material, namely the previously ad hoc conventionalised prefabricated element im schnee liegen (‘lying in the snow’). In the overlapping utterances it is hard to separate which linguistic material corresponds to which syntactic project or speaker, so we will reproduce them here in a grille-inspired version (cf. Auer and Pfänder 2007): H: ich bin halt (.) kurz vor wEIhnachten irgendwie mit=m fAhrrad geFAHRen un da– der im SCHNEE; S: (...) lAg H: so (.)nebenbEI:(.) was bUntes im schnEE geSEHN; was im SCHNEE; T: lAg ma H: un da:nn WAR das dEr.

Furthermore, each of the utterances is a turn continuation of the utterance H has begun, and T speaks in a stylised voice. Not surprisingly, as we speak about improvisation, things are somewhat paradoxical: On the one hand, each turn-constructional unit (06, 07, 08, 09) can clearly be identified as a turn continuation of 04 and attributed to the different speakers. On the other hand, the outcome can only be understood as a joint product, and this for four reasons: a) the syntactic continuation format b) the overlapping time management c) the constant recycling of the same material d) the animation of “other” voices, which attributes one utterance to another speaker who is in fact speaking at about the same time. This shows the importance of temporality in oral syntactic projects – not only in its sequentiality, but also in its speed, which (as in the example) leads to a form of synchronicity. 5.3 Before, now and soon This example illustrates the notion of an intensively lived-through moment of “now”. Time has to be furthermore understood in its different levels of depth: the moment of speech, but also the recentness effect from the previous conversation and different levels of sedimentation of constructions. Paradoxically, it seems that the ad hoc improvisation is possible precisely because of the deeply sedimented and prefabricated character of the string im schnee liegen.

Improvisation, temporality and emergent constructions

209

With the help of the same or very similar linguistic material, the three speakers do different, but at the same time anticipated and surprising, things. Speakers are syntactically anticipating and thus exploiting the emergent structure just produced by the others, but they provide surprises by changing the course of the narration either semantically or content-wise. T does not give a completely different version of the story, but rather focuses on a particular moment of the story: the moment when H does not yet realize that what he sees is an animal. Thus, the surprise for the speakers lies not so much in the ungrammatical syntactic outcome that goes rather unnoticed because of the overlapping speech production, but in the storyline construction. To sum up, it is important here to stress two facts about the interactional context. First, based on a misunderstanding, the interlocutors gently start teasing and making fun of the situation. Second, the speech situation involves a peculiar image of a budgie lying in the snow, which is attached to the loose construction im schnee liegen. This string becomes a prefabricated element in the current conversation, which is used and reused in the subsequent utterances and is altered quite often so that even a new/ungrammatical version passes unnoticed: ich bin … mit dem Fahrrad gefahren und da Ø nebenbei was im Schnee gesehen (‘I was … riding my bicycle and there/then just by chance Ø seen something in the snow’) instead of … und da habe ich nebenbei was im Schnee gesehen (‘ … and there/then just by chance I have seen something in the snow’). As paradoxical as it may sound, we have here an ad hoc prefabricated element whose evidence lies in the parallel output of very similar constructions in a polyphonic conversational activity. The time depth here is on the level of the conversation itself, and thus conversational memory refers less to the communicative or cultural memory in the sense of Assmann (1992) than in the example zurückmalen (ab heute wird zurückgemalt). Referring back to the notion of reactualizing, a term borrowed from Günthner (re-aktualisieren), the claim might be too strong that speakers reactualize a past experience by reusing the string im schnee liegen. The past is in this case only a few utterances ago. Nevertheless, repeating the string here refers back to speech events that took place some seconds ago and therefore can be seen as retention. By reusing the string im schnee liegen, the current utterance also refers back to past utterances, reactualizes them and therefore somehow resolves and shortens the temporal difference between the currently uttered string and preceding utterances. The speed in which new constructs are formed all’improvviso seems possible only when we assume that the speaker’s language processing is not confined to what is currently heard (the phonetic chain registered by the recorder), but already hurries ahead, i.e. makes projections. The speaker is not

210

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

only able to hear implicitly that which has not yet been uttered but has been projected in the given conversation, turn-constructional unit or syntactic project. Different outcomes are possible for a while. Thus, the concept of a “well-formed sentence” is not ruled out here; it merely needs some time to be realized, and sometimes it is not the well-formed sentence we aimed at, but another one which surfaces. What we would like to question is not so much the existence, but rather the importance of a “set of fixed constructions” (Hopper, this volume). This does not mean that speakers are “creative” at all times in an artistic or aesthetic sense of the term. But the very potentiality of constructing new forms might be more characteristic of language than we have previously thought. Paths that are particularly entrenched can easily be used to create very precise expectations. The production of linguistic structures according to previously established patterns does not imply their identical reproduction, though; rather, it means that such novel structures in the very moment of utterance are easily accessible to the audience even though they have never been heard before.

6

Montage and analogy: two principles of improvisation

What we have tried to show so far is that considering improvisation can enrich the theory of emergence, insofar as it can model those moments in which speakers not only draw from previously stored constructions, but where a potential of novelty, foreignness and incommensurability is opened up (Brandstetter 2009: 147). Thus, a narrow notion of improvisation must be separated from a wider or general one, because the narrow notion is only concerned with the intentional, humorous acting in front of an audience (cf. Raible 2009). A wider notion of improvisation, as we propose it with reference to Brandstetter (2009), on the other hand, opens up a field within the context of emergence that tells us something about the nature of grammar in general. We suggest that there are two important principles of improvisation in conversation, namely montage and analogy. The emergent constructions in our examples can best be described in terms of montage, a term used in art practice to designate “ways in which materials from different sources are put together to produce an object” (Nowell-Smith 1991: xiii). In a montage, elements of different sources are put together in such a way that the result has never been seen or heard before. From this perspective, our data help to show not only how grammatical resources are used by different speakers at one time, but also how new constructions emerge from discourse.

Improvisation, temporality and emergent constructions

211

The principle of montage is reflected and operative in different, specific techniques. Such techniques are, for instance, blending constructional fragments or isolating and thus creating such fragments. However, montage is also the principle for editing emergent syntactic structures8 – in our case the editing of on-line (morpho)syntactic projects – which can be seen as a more or less routine practice. We observed this in the highly collaborative productions in the example Schnee. To explain this somewhat further, the current speaker has to “help the listeners out” and arrange things for them. In terms of syntax, this means that he has to open projections which enable the interlocutors to complete the sentence under construction. On the other hand, however, subsequent completions do not always have to happen in the way they were intended by the speaker. The interlocutors can use a current projection to make “unexpected” continuations. Such unexpected or unintended continuations do not have to be produced because the collaborating speaker has misunderstood or misinterpreted the intention of the first speaker, but precisely because he has understood his intention and does not want to conclude the sentence in the predictable way: Speakers not only fulfil projections, but also play with them in an improvisational manner. In improvisation, things take shape due to the on-line deployment of the knowledge of a given language. This knowledge does not consist of a set of unrelated item-based facts, but is instead structured in an interconnected network. In recent exemplar-based approaches to syntax, which at least in some respects seem similar to emergent grammar theory, basic assumptions have been made about our linguistic knowledge, for instance about the importance of similarity, or as Goldberg, Casenhiser and White put it in their explanation of exemplars: “Exemplars are more likely to be classified together if one exemplar reminds the learner of another” (Goldberg, Casenhiser and White 2007: 73). For Bod, an exemplar is a form of “categorization, classification or analysis of a token […] while a token is an instance of use, an exemplar is a representation of a token […] an exemplar in syntax can be a tree structure of an utterance, a feature structure or whatever syntactic representation one wishes to use to convey the syntactic analysis of a particular utterance” (2006: 293). We do memorize items and strings, but also the routines of their making, as Bod suggests: “We do not only have knowledge of a vast collection of fragments, we also have knowledge of how to combine 8

Nowell-Smith states further: “There is nothing mystical about montage”, montage being the ordinary word for film-editing in French and in Russian (montazh), Italian (montaggio) and Spanish (montaje); “it is a word with strong practical […] overtones” (1991: xiii).

212

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

them, more often than not following the most frequent token or type (analogy)” (2006: 293). Analogy is perhaps the oldest concept if we wish to account for the online making of grammar by means of similarity based on type frequency. It was a strong principle in works as early as Paul’s (1880) Prinzipien at the end of the nineteenth century and especially by Meillet during the early twentieth century (cf. Meillet 1921). In more recent years, the concept has regained popularity (cf., for example, Itkonen 2005; Wanner 2006). Recently, Blevins and Blevins (2009) have defined analogy as “a general cognitive process that transfers specific information or knowledge from one instance or domain (the analogue, base, or source) to another (the target). Sets of percepts, paired, and higher-order generalizations are extracted and carried over to new sets” (Blevins and Blevins 2009: 2). Crucial for our understanding, analogy relies heavily on similarity: One thing reminds us of another, similar one (see Goldberg, Casenhiser and White 2007). However, the similarity does not have to be on one and the same level of cognitive processing. Similarities can also be of an abstract nature and involve functional and causal relations between the analogues. Neither is the recognition of an analogical relationships always predictable, depending heavily on the ongoing speech activities. Rather, it can be described as an “aggressive process driven by a search for predictability” (Blevins and Blevins 2009: 1). In this pattern-seeking mode, structural similarities can be extended beyond the sphere of perceptual objects and employed to infer novel information about the world. Speakers thus actively look for similarities, searching for patterns. In general, the human mind seems to be an “inveterate pattern-seeker” (Blevins and Blevins 2009: 1). Once patterns are extracted from what is perceptually and cognitively given, they are classified and related to other patterns in such a way that still more and different patterns can be predicted. In this way, pattern-seeking is a process which requires some sort of analogical reasoning that goes beyond mere similarity. It requires that structural similarities can be discovered even “between perceptually dissimilar elements” (Blevins and Blevins 2009: 1). For the process of improvisation, this implies that the improviser does not only use elements that are similar to each other in order to combine them in a novel fashion. Often, the novelty of the material that is produced in the act of improvising lies rather in the analogical construction of something which consists of dissimilar parts, thus creating an effect of surprise. The improviser is a pattern-seeker, but the patterns are often found and created on the spot, connecting that which is given in the moment now with previous experiences and setting up a context for further usage. Insofar as

Improvisation, temporality and emergent constructions

213

improvised patterns can be sedimented and habitualized – sometimes even creating standards of improvisation – and then exploited in future improvisations, they themselves form an analogical basis for further processes of pattern-seeking.

7

Conclusion

The present paper – which is explorative in the spirit of FRIAS – started from Hopper’s (1987, 1998, 2001, 2005) model of emergent grammar, focusing on the improvisational aspect of speaking. Our analysis highlighted the improvisation of linguistic structures by looking at those moments in storytelling where the main speaker (the storyteller) starts his narration, but where the storyline takes an unexpected turn, not so much on the plot-level, but on the morphosyntactic (word-formation, Section 4) and syntactic (verb-phrase formation, Section 5) levels, where unexpected (or even ungrammatical) structures emerge. Three aspects seem particularly important here, which follow directly from the concept of improvisation: a) Grammatical acting oscillates between making and letting go within split seconds. b) Participants in (inter)action take recourse to already-uttered material, which was either heard in the immediately preceding moment (example Schnee) or in the wider previous course of the conversation (example zurückmalen); they are not only collectors of prefabricated elements, however, but also pattern-seekers. This holds true for syntactic patterns as well as semantic frames. c) Temporality must be taken seriously in the modelling of spoken grammar – not only in its sequentiality, but also in its simultaneity. Speakers and hearers do synchronize with each other, but they also engage simultaneously in different projects, which can become audible very suddenly. From the presented findings, we draw some consequences for the theory of language: We intended to show in our contribution how speakers deal with constructions by way of improvising, be it intentionally and volitionally or passively and affectively. Constructions are not (only) to be modelled as integrated unities of form and meaning. Rather, the moments of improvisation indicate that the activation of constructions can begin at either the purely formal or semantic end. When we model constructions, we should also include how speakers employ these constructions in improvisation. Emergence does not revert to experiential knowledge in the same manner in all moments and situations; the varying temporal depth of experience as

214

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

well as the grade of sedimentation of linguistic structures play an important role, too. The last consideration may seem surprising at first. The focus on the very moment of improvisation seems to suggest that the perspectives of “emerging” and “emergent” cannot be separated, i.e. the diachronic dimension should not be excluded from a model of emergent grammar. The reason for this is that only by taking into account the different time scales can we obtain a precise analysis of the data: The overlapping speech in the example Schnee demonstrates the crucial importance of each second. The gradual unfolding of different blends in the example zurückmalen shows the central importance of conversational memory. The blend to World War II (‘from today on we draw back’) illustrates the importance of communicative memory, and the question of the construction of analogies in the case of zurückmalen can only be understood in relation to the linguistic constructions within the diachronic development of particle verbs. In sum, our paper is a plea for the further investigation of what can be different, or is made different, in any moment of the emergence of grammar.

References Assmann, J. 1992 Das kulturelle Gedächtnis. Schrift, Erinnerung und politische Identität in frühen Hochkulturen, München. Auer, P. 2000 On-line syntax. Sprache und Literatur 85, 31, 43–56. 2005 Projection in interaction and projection in grammar. Text 25, 1, 7–36. 2007 Syntax als Prozess. In: H. Hausendorf (ed.), Gespräch als Prozess. Linguistische Aspekte der Zeitlichkeit verbaler Interaktion, Tübingen, 95–124. 2009 On-line syntax: thoughts on the temporality of spoken language. Language Sciences 31, 1–13. Auer, P. & S. Pfänder 2007 Multiple retractions in spoken French and spoken German. A contrastive study in oral performance styles. Cahiers de Praxématique 48, 57–84. Bauman, R. 1977 Verbal Art as Performance, Prospect Heights. Bakhtin, M. M. 1976/86 The problem of speech genres. In: C. Emerson & M. Holquist (eds.), Speech Genres and Other Late Essays, Austin, 60–102. 1981 The Dialogic Imagination, Austin. Biesenbender, V. 1992 Von der unerträglichen Leichtigkeit des Instrumentalspiels, Aarau. Blevins, J. P. & J. Blevins 2009 Introduction: analogy in grammar. In: J. P. Blevins & J. Blevins (eds.), Analogy in Grammar. Form and Acquisition, Oxford, 1–12.

Improvisation, temporality and emergent constructions

215

Bochner, A. P. & C. Ellis 1995 Telling and living. Narrative co-construction and the practices of interpersonal relationships. In: W. Leeds-Hurwitz (ed.), Social Approaches to Communication, New York, 201–216. Bod, R. 2006 Exemplar-based syntax: How to get productivity from examples. Linguistic Review 23, 291–320. Bormann, H., G. Brandstetter & A. Matzke (eds.) (to appear) Improvisieren. Paradoxien des Unvorhersehbaren. Kunst – Medien – Praxis, Bielefeld. Brandstetter, G. 2009 Improvisation im Tanz. Lecture-Performance mit Friederike Lamper. In: M. Gröne, H. Gehrke, F. Hausmann, S. Pfänder & B. Zimmermann (eds.), Improvisation. Kultur- und lebenswissenschaftliche Perspektiven, Freiburg, 133– 157. Breyer, T. 2009 Attention and language. Intuitio 2, 1, 245–256. (to appear) Attentionalität und Intentionalität. Überlegungen zu einer phänomenologischkognitionswissenschaftlichen Theorie der Aufmerksamkeit. Ph.D. dissertation, Department of Philosophy, University of Freiburg. Cunha, M. P., J. V. Cunha & K. N. Kamoche 1999 Organizational improvisation: What, when, how, and why. International Journal of Management 1, 3, 299–341. Ehmer, O. 2011 Imagination und Animation. Die Herstellung mentaler Räume durch animierte Rede. Berlin. Ehmer, O. & S. Pfänder 2009 Sprache kann in jedem Moment ganz anders sein. Improvisationstechniken im Gespräch. In: M. Gröne, H. Gehrke, F. Hausmann, S. Pfänder & B. Zimmermann (eds.), Improvisation. Kultur- und lebenswissenschaftliche Perspektiven, Freiburg, 175–195. Eisenstein, S. M. 1989 Montage and architecture. Assemblage 10, 111–131. 1991 Towards a Theory of Montage (Selected Works, Vol. II), London. Essl, K. 2004 Improvisation über “Improvisation”. (http://www.essl.at/bibliogr/impro visation.html#zufall, September 15, 2009). Fauconnier, G. & M. Turner 1998 Conceptual integration networks. Cognitive Science 22, 2, 133–187. Frost, A. & R.Yarrow 1990 Improvisation in Drama, New York. Giddens, A. 1984 The Constitution of Society. Outline of the Theory of Structuration, Cambridge. Goldberg, A. E., D. Casenhiser & T. R. White 2007 Constructions as categories of language. New Ideas in Psychology 25, 70– 86. Gröne, M., H. Gehrke, F. Hausmann, Stefan P. & B. Zimmermann (eds.) 2009 Improvisation. Kultur- und lebenswissenschaftliche Perspektiven, Freiburg.

216

Thiemo Breyer, Oliver Ehmer and Stefan Pfänder

Günthner, S. 2006 Grammatische Analysen in der kommunikativen Praxis – ‘Dichte Konstruktionen’ in der Interaktion. In: A. Deppermann, R. Fiehler & T. Spranz-Fogasy (eds.), Grammatik und Interaktion, Radolfzell, 95–121. 2007 Brauchen wir eine Theorie der gesprochenen Sprache? Und: wie kann sie aussehen? Ein Plädoyer für eine praxisorientierte Grammatiktheorie. gidi Arbeitspapierreihe (http://noam.uni-muenster.de/gidi/ 6, 1–22). Hopper, P. 1987 Emergent grammar. Berkeley Linguistic Society 13, 139–157. 1992 Emergence of grammar. In: W. Bright (ed.), International Encyclopedia of Linguistics, Vol. I, Oxford, 364–367. 1998 Emergent grammar. In: M. Tomasello (ed.), The New Psychology of Language, Mahwah, N.J., 155–175. 2001 Grammatical constructions and their discourse origins: Prototype or family resemblance? In: M. Pütz, S. Niemeier & R. Dirven (eds.), Applied Cognitive Linguistics I: Theory and Language Acquisition, Berlin, 109–129. 2004 The openness of grammatical constructions. 40th Annual Meeting of the Chicago Linguistic Society, Chicago, 153–175. 2005 Bi-clausal constructions and emergent grammar. Talk presented at the German Department of the Westfälischen Wilhelms-University Münster, April 2005. Husserl, E. 1950 Ideen zu einer reinen Phänomenologie und phänomenologischen Philosophie. Erstes Buch: Allgemeine Einführung in die reine Phänomenologie. Husserliana III/1, The Hague. 1983 Ideas Pertaining to a Pure Phenomenology and to a Phenomenological Philosophy. First Book: General Introduction to a Pure Phenomenology. Collected Works II, The Hague/Boston/Lancaster. 1991 On the Phenomenology of the Consciousness of Internal Time, Dordrecht. Hymes, D. 1981 Breakthrough into performance. In: D. Hymes (ed.), In Vain I Tried to Tell You. Essays in Native American Ethnopoetics, Philadelphia, 79–141. Itkonen, E. 2005 Analogy as Structure and Process. Approaches in Linguistics, Cognitive Psychology and Philosophy of Science, Amsterdam/Philadelphia. Kurt, R. & K. Näumann (eds.) 2008 Menschliches Handeln als Improvisation: Sozial- und musikwissenschaftliche Pespektiven, Bielefeld. Lampert, F. 2007 Tanzimprovisation. Geschichte – Theorie – Verfahren – Vermittlung, Bielefeld. Lerner, G. H. 1992 Assisted storytelling: deploying shared knowledge as a practical matter. Qualitative Sociology 15, 3, 213–245. 2004 Collaborative turn sequences. In: G. H. Lerner (ed.), Conversation Analysis: Studies from the First Generation, 225–256, Amsterdam/Philadelphia. Mäder, U. 1996 Im Kreis – drehend, Aarau. Meillet, A. 1921 Linguistique historique et linguistique générale, Paris.

Improvisation, temporality and emergent constructions

217

Mitchell, R., E. Parker & C. Taborn 2004/05 Improvisation ist Komposition. Jazz Thing 56, 68–70. Moorman, C. & A. Miner 1998a Organizational improvisation and organizational memory. Academic Management Review 23, 4, 689–724. 1998b The convergence of planning and execution: Improvisation in new product development. Journal of Marketing 62, 3, 1–20. Nowell-Smith, G. 1991 Eisenstein on Montage. In: S. M. Eisenstein, Towards a Theory of Montage (Selected Works, Vol. II), xiii-xvi. M. Glenny & R. Taylor (eds.) London.. Paul, H. 1880 Prinzipien der Sprachgeschichte, Tübingen. Pfänder, S. & M. Skrovec 2010 Donc, entre grammaire et discours: pour une reprise de la recherche sur les universaux de la langue parlée à partir de nouveaux corpus. In: M. Drescher & I. Neumann-Holzschuh (eds.), Syntaxe de l’oral dans les variétés non-hexagonales du français. Tübingen, 183–196. Raible, W. 2009 Adaptation aus kultur- und lebenswissenschaftlicher Perspektive – ist Improvisation ein in diesem Zusammenhang brauchbarer Begriff ? In: M. Gröne, H. Gehrke, F. Hausmann, S. Pfänder & B. Zimmermann (eds.), Improvisation. Kultur- und lebenswissenschaftliche Pespektven, Freiburg, 19–24. Sawyer, R. K. 2001 Creating Conversations. Improvisation in Everyday Discourse, Cresskill. Szmrecsanyi, B. 2006 Morphosyntactic Persistence in Spoken English. A Corpus Study at the Intersection of Variationist Sociolinguistics, Psycholinguistics, and Discourse Analysis, Berlin/New York. Wanner, D. 2006 The Power of Analogy. An Essay on Historical Linguistics, Berlin. Wilson, P. N. 1999 Hear and Now: Gedanken zur improvisierten Musik, Hofheim. Zanetti, S. 2009 “improvisieren” (http://www.poeticon.net/artikel/improvisieren.html, September 15, 2009).

218

Peter Auer and Jan Lindström

Peter Auer and Jan Lindström

Verb-first conditionals in German and Swedish: convergence in writing, divergence in speaking*

1

Introduction

Like all Germanic languages, German and Swedish have at least two ways of coding conditional relations between two propositions, which are often taken to be functionally and semantically equivalent. One is based on the canonical subordinate clause pattern that makes use of a conjunction (wenn and om, respectively); the other is based on the clause-initial placement of the finite verb. These verb-initial conditional clauses (henceforth: V1-C) show strikingly similar patterns in both languages according to the grammar books. In this paper, we will first show that, other than expected, the usage patterns of Swedish and German V1-C are quite different when the difference between written and spoken language is taken into account: While the construction is frequent in certain written genres in both languages, it is almost absent from modern spoken German, but widely used in spoken Swedish. Comparing the same construction in two closely related languages therefore sheds light on processes of language change, but only when genre differences are taken into consideration. Secondly, we will argue that the online emergence of the V1-C and the conjunctional conditional constructions in spoken interaction is subject to different regularities in the two languages. The advantages and disadvantages of the V1-C construction’s on-line processing in spoken German and Swedish come to the fore as soon as the constructions are analysed from the point of view of interactional linguistics. We conclude by arguing that the disadvantages of the on-line processing of the V1-C in spoken German are a major reason why these seem to be disappearing from the language, with the exception of some fossilised variants which structurally resemble those of English.

* We wish to thank Martin Hilpert and Stefan Pfänder for their comments on a previous version.

Verb-first conditionals in German and Swedish

2

219

What the grammar books say

Verb-first constructions are a versatile resource with a bundle of parallel functions in German and Swedish (Auer 1993; Diessel 1997; Lindström and Karlsson 2005; Önnerfors 1997). The V1 pattern is used at least as the normal form of polar question (1), in the conditional protasis as an alternative to a conjunctional conditional (2), as a possible form for declarative sentences, most typically in a responsive dialogue position (3), and in a number of special pragmatic functions such as for optatives, adhortatives, exclamatives, and desideratives (4).1 (1) Haben Sie Fragen? Har ni frågor? ‘Do you have questions?’ (2) Haben Sie Fragen, können Sie mit mir Kontakt aufnehmen. Har ni frågor, kan ni kontakta mig. ‘If you have questions, you can contact me.’ (3) Haben Sie Fragen? – Ja, haben wir. ‘Do you have questions?’ – ‘Yes, we do.’ Ska vi göra det? – Ja, kan vi göra. ‘Shall we do it?’ – ‘Yes, we can.’ (4) Wäre das nur so einfach. Vore det så enkelt. ‘Were it so simple.’ Standard grammars mention V1-Cs but have very little to say about them. For example, the Swedish Academy Grammar (Teleman, Hellberg and Andersson 1999: 647–648) mainly notes that the V1-C construction is an alternative to the regular om-conditional and that both of these conditional constructions are syntactically and semantically similar. The DUDEN grammar of German (2005: 1093) merely lists V1-C as an alternative to conjunctional conditionals, and so does the monumental Grammatik der deutschen Sprache (Zifonun et al. 1997 vol. 3, 2281). Weinrich (2003: 743) goes a step further towards a genre differentiation and points out that V1-C is typical of scien1

Directives with the verb in the imperative are, of course, another type of a V1 clause (Gib mir das Buch, Ge mig boken). Since these constructions are characterized by a special verb mood and in most cases by the omission of the subject, imperatives are different from the other V1 constructions in (1)–(4). Admittedly, the constructions in (4) also have some distinct properties, including special verb forms like the conjunctive; however, this verb form was used regularly in conditional clauses in older Swedish and is still used in German.

220

Peter Auer and Jan Lindström

tific prose, and that in spoken German, the construction is often linked to the use of the “restrictive conjunctive of the modal verb soll” (translation PA/JL), i.e. sollte. The only structural difference between Swedish and German, from the point of view of the grammars, is that in German, V1-C can be postponed, while this is not possible in Swedish:2 (4) Sie können gern mit mir Kontakt aufnehmen, sollten Sie noch Fragen haben. *Ni kan kontakta mig, skulle ni ännu ha frågor. ‘You can contact me, should you still have questions.’ V1-Cs do not thus look like a terribly exciting phenomenon at first glance. When we widen the scope of comparison to include more Germanic languages, however, an interesting imbalance between conjunctional and V1-Cs emerges (see table 1). Table 1: Distribution of V1-Cs in some Germanic languages (adapted and supplemented from Iatridou/Embick 1994)

Old English Middle English Modern English Middle High German Mod. German (written) Dutch Yiddish Old Swedish Modern Swedish Icelandic

Prepositioned (counterfactual)

Prepositioned (non-counterf.)

Postpositioned (counterfactual)

Postpositioned (non-counterf.)

+ + + +

+ + should +

+ + + +

+ + should +

+

+

+



+ + + + +

+ + + + (+)

+ – + – +

– – – – –

There seems to be a tendency to restrain the contexts in which V1-C can be used, as compared to those in which the conjunctional conditional is possible. Modern English has gone furthest by restricting V1-C to counterfactuals (and in fact, only when applied with a limited set of auxiliaries like had 2

The Swedish Academy Grammar (vol. 4: 467) notes that a concessive conditional clause (e.g. “even if ”) in V1 form which also contains the emphatic adverb så ‘so’ may stand in postposition: Det klarar han inte, håller han så på i tio år ‘He won’t make it, even if he (then) continued for ten years’. This is not, strictly speaking, the V1-C construction we are investigating here, since an additional adverbial element (e.g. så) must be present to explicate the special concessive reading. Hilpert (2010) also cites a few constructed Swedish examples of postpositioned V1-C.

Verb-first conditionals in German and Swedish

221

and were) and should-introduced protases, and Yiddish and modern Swedish only allow prepositioned V1-C. In general, postpositioned V1-Cs are often restricted to counterfactuals. Older stages of English and German show the widest distribution of the construction (cf. Molencki 1999). In sum, V1 clauses have been losing their ability to express conditionality. We argue that spoken Modern German is in fact in the process of restricting V1-C as well, in ways similar to Modern English.3

3

Historical background

The history of V1-C has two phases. In the first phase, the construction grammaticizes; in the second phase, its scope narrows down. The existing literature has exclusively been concerned with the first phase. There is general agreement that the V1 conditional construction has emerged out of other uses of V1 already available in the languages. However, there is considerable disagreement about which of these constructions grammaticized. One immediately appealing theory – Jespersen (1940: 374) probably being its most famous proponent – holds that V1-C emerged from question/answer sequences. With respect to Old High German (OHG) and Middle High German (MHG), already Paul (1920: 270) writes: “Doch besteht seit ahd. Zeit eine Konkurrenz [for conjunctional conditionals, PA/JL] durch Sätze ohne Konjunktion, die aus der direkten Satzfrage hervorgegangen sind (…), z. B. gîst du mir dîn swester, sô will ich ëz tuon, Nib”. This view is supported by the fact that V1-C seems to occur exclusively in languages in which polar questions are marked by verb-initial syntax. Recently, van den Nest (2009) has revisited Paul’s thesis. He argues for a grammaticalization cline from pseudo-dyadic sequences as in the following examples, over V1-Cs with realis meaning to non-realis, non-resumptive V1-Cs. 3

The more restricted nature of V1-C is also due to the fact that it cannot be used in some syntactic contexts. For instance, it is impossible as a an answer to a question: A:

Kommst du auch rechtzeitig? ‘Will you be on time?’ B: (a) *kann ich ein Auto mieten (b) wenn ich ein Auto mieten kann. ‘if I can rent a car’ or when the protasis is in the scope of a focus or negation particle (Iatridou and Embick 1994: 197): (a) *Ich hätte nur dann rechtzeitig kommen können, hätte ich ein Auto gemietet. (b) Ich hätte nur dann rechtzeitig kommen können, wenn ich ein Auto gemietet hätte. ‘I would have only arrived on time if I had rented a car.’

222

Peter Auer and Jan Lindström

(5) (van den Nest 2009) (source: St. Galler Tagblatt 24. 4. 98, “Schalmeien laden ein zum Fest”) Sind Sie neugierig auf die Schalmeien-Musik geworden? have you curious about the shawm music become? Dann lohnt sich ein Besuch am 3. Mai, ab 10.30 Uhr am Krummensea-Mannli-Fest. Then rewards REFL a visit on 3 May, from 10.30, at KummenseaMannli festival. ‘Have you become curious about shawm music?’ ‘Then you will enjoy a visit to the Kummensea-Mannli festival on May 3, from 10:30 on.’ Van den Nest believes that the existence of all three points on the cline in modern German is evidence for their synchronic emergence. His empirical argument is that earlier stages of the language still dominantly adhere to the structure of the V1-question in coding realis protases and using non-integrated syntax.4 Investigating similar parameters, Hilpert (2010) argues on the basis of the structure of V1-C in modern German that V1-Cs are less grammaticized in this language than in Swedish, and that German still shows traces of the Q/A-sequence from which they developed. Other writers have taken different views. For instance, Erdmann (1886: 188) observes: Seit ältester Zeit dient ferner das vorangestellte Verbum zur Bezeichnung eines nur angenommenen Vorganges in conjunctionslosen Bedingungssätzen. Ursprünglich wurden sie wol selbständig dem folgenden Satze vorangestellt: kommt er (=ich will annehmen, dass er kommt), so sehe ich ihn. Dann wurde das vorangestellte Verbum als besonderes Kennzeichen dieses Satzverhältnisses angesehen und machte jede Conjunction entbehrlich.

Here, conditional V1-placement is not linked to Y/N-questions but treated as a general coding device for a “presumed process”: an epistemic status that applies to both conditional protases and questions. Behaghel (1928: 637) dis4

A point that may be raised against this argument is that clause integration in general is much weaker in OHG than in modern German. OHG examples of V1-C (see the list in Blatz 1896: 1171) often seem to be less integrated than modern examples because the protasis does not occupy the position before the finite verb (the “front field” of modern German). This makes it look less integrated, but this lack of integration does not reflect the old question format, and is rather due to the lack of a field structure and the lack of integration of subordinated-into-matrix clauses in general. We give some examples of Old Swedish below, and the same type of structures can be found in OHG.

Verb-first conditionals in German and Swedish

223

tinguishes between realis (indicative present tense) conditionals for which he assumes that “derartige Sätze gehn zweifellos auf alte Fragesätze zurück”; present tense conjunctive conditionals the origin of which he sees in Aufforderungen (e.g. sei getreu bis in den Tod, so wil ich dir die Krone des Lebens geben – Luther Offenb. 2, 10); and past conjunctive conditionals, which he traces back to optatives (Wunschsätze, e.g. o hette ich Flügel wie Tauben, siehe, so wolt ich mich ferne wegmachen Luth., Ps. 55, 7). Here, a variety of old V1 structures is believed to have influenced the V1 conditionals in their various meanings. Erdmann’s claim that questions and protases in conditional clauses share a common feature of non-assertiveness (both questions and protases have no fixed truth value), and that V1conditionals code their non-assertiveness, has been taken up by Harris and Campbell (1995). Even more generally, Hopper (1975: 51) argues that in Old Swedish (as presumably in OHG and Old English), the “clause-initial verb was a possible emphatic alternative to the final and enclitic verb”, i.e. it was one of the normal ways to encode a relationship between two propositions. In this theory, there is no need to derive V1 conditionals from questions or any other specific sentence mode. Wessén (1956: 215ff.) also points out that verb-initial clausal syntax occurred generally in Old Swedish main clauses, for instance in a kind of presentational construction (often called “narrative inversion”), which had to be followed by another sentence. In this stage of development, the first (V1) clause was an independent clause juxtaposed to another main clause that contained a formulation of a consequence of the information expressed in the initial clause: (6) Gangär at stiälä bryti ok präl. bryti skal vppi hängiä ok eigh präl. (VgL I) Go and steal villain and slave. Villain shall be hanged and not slave. ‘A villain and his slave go there and steal. The villain shall be hanged and not his slave.’ (7) Vil konungin af landit fara. Ängin af idar skal honum fölgia. (Birg. aut.) Wants king from land go. No one of you shall him follow. ‘The king wants to leave the land. No one of you shall follow him.’ This paratactic construction consisting of two juxtaposed main clauses in principle has the semantics of a V1-C in modern Swedish, but there is no structural integration yet. The second clause, semantically the consequence, begins with the subject (non-inverted word order, first clause is outside the front field). In the next phase, the second clause tends to be introduced by a resumptive, anaphoric pa ‘then’, pär ‘there’ or pät ‘that’ (Wessén 1956: 218): (8) Är eig sun. Pa är dotter. (VgL I)

224

Peter Auer and Jan Lindström

Is no son. Then is daughter. ‘Is there no son. Then it is the daughter (whose turn it is).’ (9) Dräpär mapär man i kirkiu. Pät är nipingsvärk. (VgL I) Kills another man in church. That is villain’s work. ‘Kills a man another man in a church. That is a villain’s work.’ This can be seen as a step towards a hypotactic construction in which the antecedent is understood to be subordinate to the subsequent clause. The modern, fully integrated V1 construction is reached when the anaphoric marker is dropped, in which case the antecedent is interpreted as a subordinate clause that inhabits the front field of a main clause, followed by the finite verb (in the standard V2 position) of the superordinate clause: (10) Finns det ingen son, står dottern i tur. ‘If there is no son, it is the daughter’s turn.’ Even though we believe that there are good reasons which support the Hopper-Wessén theory on the historical emergence of V1-C (which is fully compatible with the findings which we will present in the following sections), we will not make any strong claims about this historical aspect here. Rather, we will focus on the second phase of the history of the V1-C, and particularly on an explanation of the differences between spoken Swedish (which we take to represent the older state) and spoken German (which we claim to represent a newer state, closer to modern English). These differences have to do with a certain narrowing down of the contexts in which V1-Cs can be used.

3

V1 conditionals in written German and Swedish

In this section we explore the use of V1-C in modern written German and Swedish in quantitative and qualitative terms. In both languages, the construction is quite frequent and occurs in a variety of grammatical shapes, although usage is heavily influenced by text genre. For written German, we performed a corpus search in the annotated database of the IDS Mannheim (TAGGED – Archiv der morphosyntaktisch annotierten Korpora) in two subcorpora: the LIMAS subcorpus of ca. 1970, which is a 1.23-million-word “balanced” corpus modelled on the American English Brown corpus, and the much larger (19.25-million-word) newspaper corpus Mannheimer Morgen (1991–1996). The LIMAS-corpus contains a variety of written genres, from novels to newspaper ads to religious treatises and instruction leaflets. The newspaper corpus also contains a variety of genres,

225

Verb-first conditionals in German and Swedish

but only those that can be found in a modern, regional, daily newspaper. The search algorithm we used looks for initial verb forms in the beginning of a sentence (approximated by verbs after !, ?, .), followed in a distance of no more than 20 words by a comma, which in turn is followed by a verb. This, of course, gives a high number of irrelevant hits; therefore all hits in the LIMAS subcorpus were checked manually. In the case of the MM corpus, 756 examples (all instances extracted from the 1991 subcorpus) were checked manually and the total number was estimated on this basis (cf. Table 2).5 Table 2: Occurrence of prepositioned V1-C and wenn-C in two German written corpora

Total words hits errors V1-Cs per 1000 words wenn-Cs per 1000 words

LIMAS

Mannheimer Morgen

(“Lim-tagged”) 1.23 m.

(“MM-tagged”) 19.25 m. 4577 60 % 2092 (estimated) 0.11 (estimated) 4004 0.21

507 0.41 537 0.44

It is tempting to compare this value with the likelihood of the occurrence of initial wenn-clauses in the same corpus,6 here approximated by the number of occurrences of the conjunction wenn in sentence-initial position. Prepositioned wenn occurs about as frequently as (initial) V1-Cs in the LIMAS-corpus, but about twice as frequently as V1-Cs in the newspaper corpus. Despite the obvious limitations of such a comparison, the numbers make it clear that V1-Cs are a relevant alternative to prepositioned conjunctional conditionals in written German.7 V1-Cs are much less frequent in the newspaper corpus than in the LIMAS-corpus (in a ratio of almost 1:4). Why this difference? First of all, comparison with the wenn-conditionals (ratio 1:2) shows that conditional re5

6

7

We also checked the occurrence of conjunctions before initial verbs (und V1-C, aber V1-C) but they only occur with negligible frequency. Since we only looked at prepositioned German V1-Cs (and since the postpositioning of Swedish V1-Cs is impossible), it seemed useless to compare the V1-C data to the totality of conjunctional conditionals. Since German wenn-clauses are often used with a temporal meaning, the number of conditional wenn-clauses is in fact even lower than the number suggest. Note that in written German (unlike in spoken German), wenn-clauses are often postpositioned (Auer 2000).

226

Peter Auer and Jan Lindström

lationships play a more important role in this corpus (due to certain genres which frequently code conditionality, see below); however, this imbalance only partly explains the difference. In order to answer the question more precisely, it is instructive to have a look at the text genres in the LIMAS corpus and the pragmatic status of the sentences with verb-initial conditional clauses. V1-Cs occur frequently in legal, scientific and regulatory texts. In legal and scientific texts, they are used to express law-like regularities. Regulatory texts give instructions or rules (how to use a typewriter, how to play a card game, etc.). Other genres, particularly literary texts, are notoriously lacking in the list of V1-C examples. Since law-like regularities and instructions are true regardless of the time of their utterance, the V1-Cs often occur in the present tense (realis conditionals), although potentialis and irrealis conditionals can be found in small numbers as well (see Table 5 below). Regulatory, scientific and legal texts are much less common in a daily newspaper, which is the reason for the higher proportion of V1-Cs in the LIMAS corpus as compared to the Mannheimer Morgen. With regard to Swedish, we performed a corresponding search in the Stockholm-Umeå corpus (SUC). This one-million-word corpus from ca. 1990 is also balanced according to the principles of the Brown corpus. The search procedure benefited from the detailed morphosyntactic coding of SUC; the target of the search was verbs in the indicative (or conjunctive) mood, in the present or past tense, and that occurred in the beginning of a graphical sentence which did not end with a question mark.8 This resulted in much less “noise” than in the German data, but approximately 10 % of the hits had to be eliminated by hand (mainly verb-first declaratives and questions). All in all, the search yielded 307 instances of V1-C. This search was then followed by another one in which a coordinating conjunction or the subordinator att was allowed to precede a verb–noun/pronoun combination.9 After sorting by hand, a further 61 V1-C were detected, i.e. a considerably higher count for conjunction + V1-C than in German (see note 5 above).

8

9

We would like to thank Lars Borin at Språkbanken, Gothenburg University, who conducted the basic searches in the version of SUC that is stored in the Språkbanken on-line services (http://spraakbanken.gu.se). We utilized a variety of search strings that were based on the morphological coding of SUC and lexical specifications, for example [word=“och”][msd= “V@IPAS.*”][msd=“P.*”], i.e. the word och followed by a verb in active present tense followed by a pronoun. This yielded instances such as … och åker man långfärdsbuss, tar det lång tid ‘ … and goes one by long voyage bus (i.e. if one goes), it takes a long time’.

227

Verb-first conditionals in German and Swedish Table 3: V1-C and prepositioned conjunctional protases in written Swedish Corpus

SUC 2.0

Total words V1-C instances per 1000 words S-initial om-instances per 1000 words

1.03 m. 368 0.36 667 0.65

In order to calculate Swedish om-conditionals, our search target was om-subordinators that inhabit the beginning of a graphical sentence. We thus eliminated adverbial and prepositional uses of om, but the search nevertheless resulted in some noise – approximately 15 % of the results consisted of sentence-initial embedded polar questions and postpositioned conditional protases that were given a graphical sentence slot of their own. The errors were again eliminated by hand, which eventually left us with 525 instances of om-conditionals. Also, instances in which a conjunction or the subordinator att preceded the sentenceinitial om-protasis were targeted, resulting in a further 142 instances (Table 3). When compared to our German data, the SUC corpus has a higher frequency of V1-C than the Mannheim newspaper corpus but only a slightly lower frequency than the LIMAS corpus. The latter comparison seems particularly relevant since both corpora were composed according to the same principles. Conjunctional om-protases are clearly more favoured in SUC, whereas in LIMAS the distribution between V1-C and sentence-initial wennprotases was practically even. Since SUC can be broken down into text types (other than LIMAS), the Swedish data nicely show the genre dependency of V1-Cs (cf. Table 4). This calculation is based on the results of the first search which gave 307 instances, i.e. instances in which V1-C initiates a graphical sentence without a preceding conjunction. As in German, newspaper texts (reviews, reportage, editorials) do not have considerably high frequencies of V1-C (the combined V1-C frequency for the three newspaper genres is 0.2), and the frequency is equally low in the superordinated category “imaginative prose”, i.e. texts mostly from novels. The construction is particularly common in legal, administrative, regulatory, and instructive texts, which are found in the genre categories “administration” and “skills, trades and hobbies”.10 Surprisingly, however, the con10

Indeed, in a Swedish handbook for authors of legal texts it is noted that conditional clauses with inverted word order (i.e. V1-C) are a “classic”, often “practical” device and should not be condemned as old-fashioned (Bruun et al. 2004: 76).

228

Peter Auer and Jan Lindström

Table 4: Conditionals in written Swedish according to genres

Press: Reviews Imaginative prose Learned and scientific writing Biographies, essays, memoirs Press: Reportage Press: Editorial Popular lore Administration etc. Skills, trades and hobbies

V1-C/1000 words

Om-C/1000 words

0.11 0.18 0.19 0.21 0.22 0.29 0.37 0.56 0.57

0.24 0.37 0.45 0.39 0.18 0.53 0.86 1.06 0.49

struction does not seem to be typical of Swedish academic writing. One reason may be that this corpus category does not contain texts from natural sciences and technology, since these are almost exclusively written in English; instead, natural sciences and technology are represented in the category “popular lore” (popularised scientific texts), which is the third-highest genre regarding the frequency of V1-C. In most text types, the om-conditional is considerably more frequent than the V1-C. In the top frequency category “skills, trades and hobbies”, however, V1-C are more frequent than conjunctional conditionals (this is also the case in the newspaper genre “reportage”). A plausible explanation for this is that the texts in this heterogenic category often have an instructive character; the sources include books and handbooks on interior decorating, pets, sports, food and wine, travel, motor vehicles, outdoor activities, computers, gardening, private finances, religion, as well as publications from non-governmental organisations of various types and trade unions (see Gustafson-Capková and Hartmann 2006). In the following, we take a brief look at the most distinct functional characteristics of verb-first conditionals in the German and Swedish written data. Typical German examples of V1-C expressing law-like regularities are the following ((11) is from a legal text, (12) from a physics text, and (13) from instructions on how to play a card game): (11) Die Vertretungen haben zwei verschiedene Beschlüsse zu fassen: den Beschluß über den Beitritt zum Sparkassenzweckverband und den Beschluß über die Vereinigung der Sparkassen. beide Beschlüsse sind rechtsgeschäftliche Willenserklärungen. Korrespondieren diese Willenserklärungen aller Beteiligten miteinander, so sind zwei öffentlich-rechtliche Verträge zustande gekommen, die die Grundlage für die Bildung des Sparkassenzweckverbandes bzw. die Vereinbarung über die

Verb-first conditionals in German and Swedish

229

Vereinigung der Sparkassen bilden. (LIM/LI1.00136, Rothe, K., Sparkassengesetz für Nordrhein-Westfalen.Kommentar; p. 218–226) ‘The representatives have to come to two different decisions: a decision about joining the association of the savings banks and a decision about the unification of the savings banks. If the declarations of will of all the involved parties correspond with each other, then two public contracts have come into existence, which are the foundation of the formation of the association of the savings banks and the agreement about the unification of the savings banks respectively.’ (12) Die höchste Feldstärke tritt jeweils dort auf, wo der Leitungstyp des Materials wechselt. Erreicht die Feldstärke dort Werte oberhalb von etwa ((Formel)), so setzt für Ladungsträger, die dieses Gebiet hoher Feldstärke durchqueren, Ladungsträgermultiplikation durch Stoßionisation ein. (LIM/LI1.00095, Krumpholz, O.*, Avalanche …; Wissenschaftliche Berichte (AEG):44, 2, p. 73–78) ‘The highest field strength occurs where the conduct type of the material changes. If the field strength reaches values above approximately ((formula)), then multiplication of conduct carriers by shock ionisation begins.’ (13) As ist die niederste Karte, dann folgen Sieben, Acht, Neun, Zehn, Bube, Dame und zuletzt der König als die höchste Karte. Haben zwei Spieler die gleiche niedrigste Karte, so zahlen beide. (LIM/LI1.00090, Grupp, C., 99 Kartenspiele; p. 68–76) ‘Ace is the lowest card, then seven, eight, nine, ten, jack, queen follow, finally the king as the highest card. If two players have the same lowest card, then both pay.’ A special case of this usage is the V1-C with the verb sein (‘to be’) in the present subjunctive which is only used in logical and mathematical texts (in the sense of English ‘let X be Y: then …’): (14) Sei auf Sigma (über ((Term))) eine Metrik Ö erklärt, dann kann man Ö zu einer Metrik omicron auf ((Term)) erweitern. Seien ((Formel)) und ((Formel)), dann ist omicron erklärt durch ((Formel)) und ((Formel)). (LIM/LI1.00103, Ratschek, H.*, Über die …; Archiv für el. Rechnen: Vol.7, 3–4, p. 172–180) ‘Let a metric ö be explained by  (over ((term))). Then ö can be expanded to a metric V on ((term)). Let there be ((formula)) and ((formula)). Then V is explained by ((formula)) and ((formula)).’

230

Peter Auer and Jan Lindström

Like English, Swedish uses the imperative låt ‘let’ and a free-standing consequential clause containing a resuming då ‘then’ in corresponding contexts: (15) Först, låt det vara ett primtal. Vi har då funnit ännu ett primtal utöver de givna A, B och C. Därefter, låt EF vara ett sammansatt tal. Det har då enligt kommentaren ovan en delare som är ett primtal. Kalla detta primtal G. (SUC-jf; Mathematics) ‘First, let it be a primary number. We have then found yet another primary number apart from the given A, B and C. Thereafter, let EF be a compound number. It has then, according to the commentary above, a divisor which is a primary number. Call this primary number G.’ Swedish – like German – favours the V1 construction in instructive and regulatory texts. Typical instances are (16) which is from a food recipe, (17) which concerns private finances, and (18) which is a legal text showing a series of coordinated V1-Cs within one sentence frame. (16) Portionsbröd gräddas i varmare ugn, ca 225°, i 10–15 minuter. Låt helst bröden svalna på ett galler. Får de kallna utan handduk blir brödskorpan knaprig och samtidigt lite seg. Vill man ha mjukare yta på brödet är det bättre att linda in nygräddat bröd i handdukar och låta det svalna så. (SUC-ea; Hobbies, amusements) ‘Portion bread is baked in a warmer oven, ca. 225°, for 10–15 minutes. Let the bread cool off on a wire rack. If they are allowed to cool off without a towel, the crust will become hard and also a bit tough. If you want to have a softer surface on the bread, it is better to wrap a newly baked bread in a towel and let it cool off like that.’ (17) Den effektiva räntan är garanterad om spararen behåller obligationen ända tills den löses in. Men skulle den säljas före den löses in kan räntan bli lägre eller högre. Allt beroende på hur marknadsräntorna utvecklas under tiden. Går räntorna upp efter det att obligationen köpts sjunker värdet på obligationen, och skulle räntorna gå ner ökar värdet på obligationen. (SUC-ea; Hobbies, amusements) ‘The effective interest is guaranteed if the saver keeps the bond until it is cashed. But should it be sold before it is cashed, the interest can be lower or higher. It all depends on how the market interests are developing during the time. If the interests go up after the obligation has been bought, the value of the obligation will decrease, and should the interests go down, the value of the obligation will increase.’

Verb-first conditionals in German and Swedish

231

(18) Avlämnas inte varan eller avlämnas den för sent och beror det inte på köparen eller något förhållande på hans sida, får köparen enligt 23–29§§ kräva fullgörelse eller häva köpet samt dessutom kräva skadestånd. (SUC-ha; Government publications) ‘If the article is not delivered or if it is delivered too late and if it does not depend on the buyer or some condition on his side, the buyer can, according to 23–29§§, claim discharge or cancel the purchase and in addition claim damages.’ Among the remaining cases of non-counterfactual V1-C in German, we often find metalinguistic, text-organising uses such as in the following examples (a total of 6 % of all the examples in the LIMAS corpus): (19) Analysiert man den Begriff des Kerygmas bei Bultmann, so gewinnt er die für Bultmanns Denken typische und notwendige Formalität zurück, die ihn allein rechtfertigt. (LIM/LI1.00056, Sölle, D., Politische Theologie: Hoffnung verändert die Welt) ‘If one analyses the notion of kerygma in Bultmann’s writings, then it regains the kind of formality typical and necessary for Bultmann’s thinking which alone can justify it.’ (20) Als neue Sportart aus Übersee schwappt die Inline-Skater-Welle nach Deutschland. Glaubt man den Sportgeschäften, so sind bereits zwei Millionen Skater auf den Straßen unterwegs. (MM/606.24052: Mannheimer Morgen, 17. 06. 1996, Lokales; Für Oppauer reichte es nur zu Platz zwei) ‘A new kind of sport from overseas: the wave of inline skating reaches Germany. If one believes the (reports of) the sports shops, then two million skaters are skating in the streets already.’ Metalinguistic uses are not uncommon in written Swedish either, especially in scientific texts: (21) Betraktar vi dagens våldsamma nationella eruptioner, möter oss mönster som ter sig egendomligt välbekanta. (SUC-jc; Social sciences) ‘If we look at recent violent national eruptions, we encounter patterns that seem strangely familiar.’ (22) Slår man upp blyg i Svensk Handordbok står det förlägen, försagd, skygg, generad: ord som alla uttrycker en känsla av obehag tillsammans med andra människor. (SUC-fb; Behavioural sciences)

232

Peter Auer and Jan Lindström

‘If you look up shy in the Swedish Handbook Dictionary it says awkward, self-conscious, timid, embarrassed: words all of which express a feeling of discomfort together with other people.’ However, not all V1 subordinate clauses code conditional relationships in written German. In some rare cases, a temporal relation (‘as soon as’, ‘once’) is expressed by the same construction: (23) (instructions on how to build a model aeroplane) Sind die beiden Tragflächenhälften verleimt und trocken, können wir die Flügel zusammenbauen. (LIMTG/LI1.00044 hobby, 1970, Nr. 9; Nr. 16, S. 136–144; S. 115–118; [Zwei Bastelratgeber]) ‘Once the two sides of the wing have been glued and are dry, we can start to attach them (the wings).’ Also, there are some rare instances in which the V1-C expresses a concessive relationship in German: (24) Die ersten fünf Jahre des unabhängigen Kenya sind in der Studie von GERTZEL: “The Politics of Independent Kenya 1963–8” aufgezeichnet. Ist dieses Buch auch ein wenig kompliziert und bisweilen abstrakt geschrieben, so vermögen die 6 Kapitel doch sehr eingehend das politische Geschehen Kenyas in dieser entscheidenden Phase unter neuen Gesichtspunkten (nur das Kapitel 5 ist ein Nachdruck) und zum ersten Mal so umfassend zu analysieren. (LIM/LI1.00194, Goswin, Grundzüge der Geschichte und politischen Entwicklung …; p. 54–63) ‘The first five years of independent Kenya are recorded in Gertzel’s study “The politics of Independent Kenya 1963–8”. Even though this book is written in a somewhat complicated and sometimes abstract way, the six chapters analyse in great detail, and for the first time so comprehensively, the political happenings in Kenya during this decisive phase from new perspectives (only chapter 5 is a reprint).’ Clearly, the clause introduced by V1 contains a proposition, the truth of which is taken for granted by the author, i.e. its semantics is completely different from that of the protasis in a conditional relationship, the truth of which is not presupposed. This also holds for another non-conditional use of V1 dependent clauses in German, which is much more frequent than temporal or concessive ones. We can call it an adversative use: (25) Diese Zeit der Kranken, die bei stets gleichem Tageslauf meist mehrere Jahre im Sanatorium leben, unterscheidet sich weitgehend von der Zeit der Menschen im Flachlande. Gibt es auf der einen Seite Tätigkeit, Veränderung, Ereig-

Verb-first conditionals in German and Swedish

233

nisse, so stehen dem im Sanatorium Untätigkeit, Gleichmaß, Ruhe gegenüber. (LIM/LI1.00014, Karthaus,U.,Der Zauberberg; DVJS 44, 2, p269–275) ‘This time of the ill people, most of whom have been living in the sanatorium for several years with the same daily rhythm, is entirely different from the time of the people on the open plains. While there is activity, change, and events on the one side, there is inactivity, sameness, and quiet on the other.’ There are also a few V1-C instances in the Swedish data which could be classified as concessive-adversative, or which at least only vaguely resemble a conditional meaning, but this use seems to be less widespread than in German: (26) Har spelarna höga löner så innebär det också att deras publik och sponsorer ställer högre krav. (SUC-eb; Society press) ‘While the players have high wages, it also means that their audience and sponsors demand a lot of them.’ (27) Blir vi vad vi tänker, blir vi förvisso också vad vi äter. (SUC/ga; Biographies, memoirs) ‘While we become what we think, we definitely also become what we eat.’ Note that adversativity is a very weak semantic relationship, much less precise than conditionality or concessivity. This renders the V1 construction semantically flexible, and in newspaper German, there is a tendency for it to turn into a mannerism, i.e. a semantically vague way of formally linking two propositions whose actual relationship remains unclear. An example follows in which it is difficult to replace the V1-clause by any more semantically exact conjunction: (28) Hier wie auch bei Volksweisen aus Italien und Rußland gaben Sensibilität und Ausdruckskraft der beiden ideal harmonierenden Künstler den Vorträgen die besondere Note. Stand die Virtuosität stets im Dienst stilistisch ausgewogener Interpretation, so feierte sie Triumphe in der faszinierenden Wiedergabe von Maurice Ravels “Piece en forme de Habanera”. (MM/606.23143: Mannheimer Morgen, 08. 06. 1996, Feuilleton; Exotische Klangfarben) ‘Here like in the folk songs of Italy and Russia, the sensibility and expressivity of the two perfectly harmonizing artists gave a special note to the performances. While ( ? ) virtuosity always served the stylistically balanced interpretation, it achieved triumph in the fascinating rendition of Maurice Ravel’s “Pience en forme de Habanera”.’

234

Peter Auer and Jan Lindström

Table 5 and Figures 1 and 2 summarise some of the semantic features of German V1-dependent clauses for each of the two subcorpora (for the newspaper corpus, the numbers refer to one year only). Conditionality is more typical of the LIMAS corpus, while adversative V1-dependent clauses are more frequent in the newspaper corpus. As Fig. 2 shows, the percentage of potentialis and realis constructions coded as V1-C also differs considerably, and this is mainly due to the high number of sollte-introduced potentialis V1-Cs in MM91. Table 5a/b: The profiles of V1 subordinate clauses in the LIMAS corpus and in the MM corpus (top: total of semantic relationships coded; bottom: only conditionals) LIMAS

MM91

Concessive Adversative Temporal Conditional Total

11 (2 %) 23 (5 %) 1 ( ‘how can I make it stop’

The speaker distinguishes between two alternatives in dealing with a hypothetical situation in which one person in her team “feels unfairly treated”. In the formulation of the first alternative (“acceptable reasons for failure”) she makes use of a wenn-clause and a present tense verb, while in the formulation of the second alternative (“interpersonal or job-related reasons”) she uses a potentialis V1-C marked by sollte. The first alternative is surely the less problematic one in terms of the activity described. There is, therefore, an interactional reason to present the second, more problematic alternative as the more unlikely one. This is what the choice of the sollte-conditional achieves. In sum, the most important qualitative difference between spoken Swedish and spoken German is the absence of realis V1 conditionals in spoken German. While Swedish seems to make use of this construction’s affinity with (written) genres in which laws and regularities are expressed and instructions given in order to achieve expert status in interaction, no such “transfer” can be shown for modern spoken German. There are, however, some reflexes which might point to an older use of this type of realis V1 conditional construction in oral language. In particular, it may be noted that V1-C (with present tense verb) provides a standard format for proverbs, weather rules (Bauernregeln) and sayings in German, formats which surely reflect traditional oral language use. Examples17 are: 17

From www.sprichwoerter.net/content/category.

250

Peter Auer and Jan Lindström

(47) V1-C in German proverbs and sayings: (a) Kannst du nicht Pfaff werden, so bleibe Küster. can you not parson become, so remain sexton. ‘If you can’t become the parson, remain the sexton.’ (b) Kennst du einen, so kennst du alle. know you one, so know you all. ‘If you know one (of them), you know them all.’ The same holds for Swedish as shown in (48). (48) V1-C in Swedish proverbs and sayings: (a) Faller man, så reser man sig igen. falls one so gets one up again ‘If you fall (i.e. fail), you pick yourself up again.’ (b) Känner du en, så känner du alla. (=47b) know you one, so know you all. ‘If you know one (of them), you know them all.’ These uses of the construction again prove that it is associated with law-like regularities or irrefutable truths in spoken Swedish. Indeed, speakers seem to use the pattern productively, to produce proverb-like, new V1-C utterances like the one in (49), taken from a high school student discussion (cf. extract 32 above with the original context): (49) Söker du jobb så får du jobb. Seek you job so get you job ‘If you (really) look for a job, you will get a job.’ Like in many V1-C proverbs, the whole expression is compact and there is structural and lexical parallelism between the protasis and apodosis.

6

Emergent syntax and V1 conditionals in spoken German and Swedish

In his emergent grammar hypothesis, Hopper (1987, 1998, in this volume) argues that grammar is constantly in the making. We build new utterances by making use of ready-made, i.e. formulaic chunks of talk; but these readymade chunks do not determine the shape of our utterances. The outcome of what we “stitch together” from various patterns, the linguistic utterance which finally appears, is not a trivial realisation of a mental model, but rather

Verb-first conditionals in German and Swedish

251

“ad hoc, disparate, and worked out ‘on stage’ in an improvised fashion” (1989/90: 5). The underspecified nature of constructional routines also means that the boundaries between one constructional scheme (routine) and the neighbouring ones are sometimes unclear, and that ambiguities between and blendings of constructions occur frequently. Since utterances emerge in a linear process of (co)construction in which a principal speaker, but also his or her coparticipants are involved, time and speaker/hearer cooperation are of principal importance to such an approach to linguistic structure. In this time course, emerging structures project options for continuation which may be stronger or weaker. As a rule, the range of possible continuations becomes narrower towards the end of an utterance. But even a fully produced utterance may in retrospect be turned into something different by the same speaker or his or her coparticipants who may add further elements which recategorize the already-produced ones.18 How can this idea of the emergent character of linguistic structure be applied to our case? We argue that the idea of emergent syntax and on-line processing can help us to explain the restrictions on the use of V1-C in spoken German. No doubt prepositioned conjunctional conditionals have a strong projecting force which is used for all sorts of conversational tasks (cf. Auer 2000; Lerner 1996). Do verb-initial clauses have an equally strong projecting force which makes a subsequent conditional apodosis likely? The question must be answered separately for sollte-conditionals, irrealis conditionals, and realis conditionals.19 V1 clauses introduced by sollte have indeed a strong projecting force. After an initial V1 clause such as .h äh sollt ich irgendwann mal nächster zeit hier dastehen und heulen ‘uhm should I at some point in the future be standing here crying’

the likelihood of a following clause which can count as a matrix clause and the apodosis of a conditional construction is very high. The auxiliary is already a marker of conditionality, and the placement of the finite verb in the initial position of the syntagma does not carry that functional load alone. The only alternative would be to treat the clause as a self-contained unit, i.e. a question, and such sollte-introduced questions are rare. An initial clause 18 19

See Auer 2009a,b; Auer and Pfänder 2007. Potentialis conditionals introduced by a full verb are exceedingly rare since the synthetic form of the conditional is hardly used in spoken language anyway; they can be neglected in this discussion.

252

Peter Auer and Jan Lindström

which is introduced by the auxilaries hätte/wäre to form an irrealis construction, such as hätte einstein det damals allet so jewusst, ‘had Einstein known about all this at the time’

is already somewhat more ambiguous. It may seem that hätte also enforces a conditional reading. However, consider the following example: (50) Sitting in the garden, the inhabitants of the Big Brother house talk about their most bizarre eating experiences. Jhn:

ah ick gloob det SCHLIMMste was ick jemals jemacht hab war irgendwie (-) n=scheißDING gewesen; ‘uhm I think the most awful thing I’ve ever done was some kind of shitty thing’ ick weiß gar nicht wie ick zu jeKOMMen bin .h ‘I don’t even remember how I came to do that’ auf jeden fall hab ick n=lEbenden fIsch den KOPF abgebissen. ‘in any case, I bit off the head of a live fish’

Jrg:

[((giggling gently

Sbr:

[((giggling, chokes

))] ))]

[hättste was jeSAGT hörmal] ‘had you said something, hey!’ Jrg:

[((laughs))

fSbr:

hätt [ich den immer RAUSgeholt;] ‘I would always have taken it out’

Jrg:

]

[((laughs))

]

‘your friend takes a goldfish out of the aquarium’ Sbr: Jrg:

und ich schmeiß [den kopf immer WEG; ‘and I always throw away the head’

]

[ ‘and he comes back and the fish is swimming around without its head’

SbrfJ:

((laughing inbreath)) boah warum haste DAS denn gemacht; ‘wow why did you do that’

Verb-first conditionals in German and Swedish

253

When analysed ex post, after its full production, Sabrina’s utterance hättste was jesagt hörmal ‘had you said something

/

hätt ich den immer rausgeholt I would always have taken it out again’.

can be understood as a V1-C in which the first clause occupies the syntactic front field of the second one. However, in the course of its emergence, there are other options, and there is evidence that it is not planned as such. The context is rather complicated since narrative, fiction, and references to common experiences interact in an intricate way. Sabrina is responding to John who has just told the story of how he once bit off a live fish’s head. She ironically suggests that since John seems to be fond of fish heads, she should not have thrown away the head of the fish which he has cooked in the house, but rather should have taken it out of the bin. The first clause hättste was jesagt ‘had you said something’

is open to at least two interpretations, none of which is excluded by context. In particular, the meaning oscillates between a syntactically non-projecting interpretation as a slightly reproachful appeal (“why didn’t you say something”), and a conditional meaning in the sense of “if you had said something” in this case projecting an apodosis. The first reading is supported by the final particle hörmal which usually occurs utterance- (and even turn-) finally and also marks the utterance as a reproach. The second reading is supported by the actual continuation of the utterance, with another ambivalent V1 clause: hätt ich den immer rausgeholt ‘I would always have taken it out’

This utterance can be understood as a continuation of the first (providing the matrix clause) or as a second self-contained V1 clause (see below). The point is that the pattern is neither fixed nor predictable, and the conditional reading competes, perhaps even from the speaker’s perspective, but surely from that of her co-participants, with other interpretations. This ambiguity of V1 clauses multiplies in the case of indicative, present tense V1 constructions. It is clearly visible in Swedish where it is even used as a rhetorical device: (51) GIC:16634. Call to a Poison Control Centre, P=pharmacist, C=caller. The caller from a daycare centre suspects that one of the children has eaten seeds from a plant.

254

Peter Auer and Jan Lindström

P:

kan de va mera än fe:m frön då ska man in ti sjukhu:s, can it be more than five seeds ‘if there is a possibility that it was more than five seeds, then one has to go to hospital’

f

kan de vara mer än fem frön? ‘can it be more than five seeds?’

C:

ja de vet >ja inte< ja ska gå: å fråga (dom e) i matsalen ‘well I don’t know about that I’ll go and ask (they’re) in the dining room’

(1.3)

What at first is constructed as a V1-C (kan de vara mer än fem frön) is in the subsequent course of the pharmacist’s turn repeated (the arrowed line), but this time transformed into a regular question. That is, the pharmacist first formulates an instruction on the basis of what she generally knows about possibly hazardous amounts of a plant’s seeds. Having delivered the instruction, the pharmacist seems to realize that she in fact does not know the specific amount of the seeds in the case at hand; thus, she moves on to gather information on the factual amount in the current case. Against this background, the avoidance of realis V1-C in spoken German can be seen as a way of avoiding an unclear projection, since questions play a much larger role in face-to-face interaction than in written texts. But the use of V1 in spoken German goes far beyond the question format and extends to many cases which are impossible in written German. At least two such cases have to be distinguished. In the first case, the V1 clause reports an event. Sometimes, these “narrative” V1 clauses occur in asyndetic pairs in which the parallelism created by the verb-first placement establishes a relatively strong coherence between the two parts. Pragmatically, the construction is highly reminiscent of the Old Swedish (or similar OHG) examples discussed above. Formally, the structure is (on the surface20) identical to that of a realis V1-C. Take the following examples: (52) John has discovered a push-up bra in a box and shows it to Andrea (Big Brother) Jrg:

kummal;=wusstest DU das=s sowas gibt? ‘look; did you know that this exists?’ beeHAAS schon mit siliKON drinne? ‘bras that already contain silicone?’ (0.5)

20

The first clause does not occupy the front field of the second; therefore the resemblance is superficial.

255

Verb-first conditionals in German and Swedish Adr:

(-) ‘no’

Adr:

‘what a scam, huh,’

Jrg:

ja:=alles

Adr:

‘yes, all false promises (lit.: misleading packages)’ [(ae) [ja

((...))

(0.5) [für

Mogelpackung[en;

Adr:

(.) TOLL; ‘great.’

f

packst dich AUS hast nix me d (.) [DRUNter. ‘you take it all off and have nothing left underneath’

Jrg:

[((laughs))

(53) John’s story in (50) about biting off a live fish’s head has now turned into a fictitious story about a fish in an aquarium at John’s friend’s house (Big Brother). Jrg:

and comes he again swims the fish without head around ‘and he comes back and the fish is swimming around without its head’

In the first example the asyndetic double V1 clause construction marked by the arrow can be paraphrased as “you take it all off and then you have nothing left underneath”, in the second case as “he comes back from the bathroom and the fish is swimming around in the aquarium without its head”. The verb-initial prepositioned clause is followed by, and provides some kind of background for, another V1 clause. These narrative V1 clauses are widely used in spoken German, and the paired (asyndetic) construction is only one rather densely-structured manifestation. In the following example, the speaker uses three V1 clauses in a row: (54) Sabrina imagines how she (mis)treats Jürgen like a piece of meat on a hook (Big Brother). Sbr:

ich hab da so=n haken und DU hängst dann da dran so. ‘I have this kind of hook and you will be hanging on it.’ a=hohoho.

256

Peter Auer and Jan Lindström

f

lass ich dich dann so in der sonne trocknen. let I you then like-that in the sun dry. ‘then I’ll let you dry in the sun.’

f

mach ich vorher hier n glatten SCHNITT, make I before here a clear cut, ‘first I’ll make a clean cut here’

f

blutest=de noch n ahaha bleed you still a bit ‘then you’ll bleed a bit longer’

Jrg:

((laughs))

Sbr:

so hier n schnitt, ‘like a cut here’ dann nehm ich dich verKEHRT rum, ‘then I’ll hold you upside down’ dann blutest de richtig schön aus an meinem arm. ‘then you will bleed to death on my arm.’

If V1 clauses were also widely used for coding (realis) conditionality in spoken German, they would in many contexts be ambiguous between the “narrative” (temporal) and the conditional meaning. Modern spoken German has resolved this ambiguity by reducing the likelihood of V1 clauses occuring in conditionals to almost zero. Initial present tense verbs are therefore free to function as in the three examples above. In spoken Swedish, V1 can also code a declarative, but these contexts are rather restricted to a responsive sequential position (cf. ex. 3), i.e. a position in which an interrogative or conditional construction is less likely to occur (Lindström and Karlsson 2005). These V1 responses often have elliptic qualities; for instance, some of them lack a subject, which also distinguishes them from interrogative and conditional V1 constructions (Mörnsjö 2002). There are some further regularities in the discourse emergence of V1-C in German and in Swedish which constrain their occurrence. They point to a stronger integration of protasis and apodosis in V1-C when compared to conjunctional conditionals. One observation is that V1-Cs occur frequently in the front field, but are found to a lesser degree in the pre-front field, particularly when the following clause is an interrogative: (55) GSM:21. High-school students are being interviewed about musical styles. B:

men så om man e rikare, tror du att man lyssnar på? ‘but so if you’re richer, do you think that you’d be listening?’

Compare (55) including a conjunctional protasis with a fabricated V1 version:

Verb-first conditionals in German and Swedish

257

?Är man rikare, tror du att man lyssnar på? The initial V1 clause is not heard as a pre-fronted protasis preparing a question but rather as a Y/N-question. The whole utterance could then be interpreted as a series of polar questions (albeit not a very natural one). The preference for conjunctional conditionals in syntactically loose positions (pre-front field) is e contrario evidence for a tight constructional integration between the protasis and apodosis in the V1 construction. This view is also supported by the fact that if the speaker initiates a turn with a protasis followed by several parentheses and/or reformulations as in (56), the protasis is more likely to be coded with an om-conditional. This is arguably due to the stronger semantic projection of om-conditionals that prepares the listener for the apodosis even in cases where it does not follow immediately. V1-C, on the other hand, requires an immediate realisation of the projected apodosis in order to minimise confusion with competing construction types such as polar questions or V1 declaratives. (56) GSM:24. High-school students are being interviewed about musical styles. A:

för do- dom hänger ju me va, till exempel om man gillar, ‘cos the- they hang around right, for example if one likes,’ om vi gillar nån artist mycke då va, till exempel E-type då ‘if we like some artist a lot, then right, for example E-type then’ som ja tycker ä grym, så ja kommer lyssna på han ‘who I think is awesome, then I am going to listen to him’ när ja blir äldre också ‘when I get older too’

The speaker in (56) presents an example in the guise of a conditional construction. It is initiated by an om-conditional (om man gillar ‘if one likes’); this protasis is then followed by a parenthetical (till exempel E-type då som ja tycker e grym ‘for example E-type who I think is awesome’), which in turn is followed by the projected apodosis (så ja kommer lyssna på han när ja blir äldre också ‘then I am going to listen to him also when I get older’). There seems to be some on-line speech planning involved, as suggested by the rephrasing of the start of the protasis, and the switch from the generic-person reference in the protasis (man ‘(any)one’, vi ‘we’) to the first-person reference in the apodosis. V1-C does not lend itself easily to such exploratory structural revisions and ambiguities, but are rather delivered in “one go”. This in turn relates to the finding presented above that we frequently find V1-Cs in conclusions. They often involve a summary, sometimes even rewording what has been said

258

Peter Auer and Jan Lindström

(cf. ex. 41), i.e. the utterance consists of material which is already available and does not require much planning. In sum, V1 conditionals are fragile objects for on-line emergence. Their projecting force is weak when compared to conjunctional conditionals since they compete with other constructions (more in German than in Swedish). As a result, they are avoided in spoken German unless introduced by an unambiguous modal verb (sollte), and they are produced in Swedish and German in tight packages which minimise ambiguity.

7

Conclusions

In this paper, we have shown that V1 conditionals are used in written Swedish and German with similar frequencies and for the same functions. However, we have also shown that in the spoken language, German has diverged from Swedish both quantitatively and qualitatively and is now similar to (spoken and written) English: English V1-C

spoken German V1-C written German V1-C

Swedish V1-C

On the basis of an emergent, on-line approach to syntax, we argued that the narrowing down of the V1 conditional construction to irrealis and sollte-constructions in spoken German is driven by the need to turn a highly ambiguous first clause, which could trigger multiple projections, into a projecting clause which is less ambiguous. In written German as well as in Swedish, this is partly achieved by making the V1 conditional format genre-dependent (legal, scientific, and regulatory texts) as well as activity-dependent (stating a law-like regularity, making a metapragmatic statement, etc.). Here, the ambiguity of V1 clauses is low since the projections that compete with the conditional one are less frequent (questions) or hardly existent (V1 declaratives). This explains the “survival” of the more traditional pattern in written German (and Swedish). In speaking, German has restricted the V1 conditional format to those verbs which rarely occur in Y/N questions, particularly to sollte, as well as hätte and wäre. Due to a different use of declarative V1 constructions (relative rarity of German “narrative V1”), Swedish does not show the same amount of ambiguity of the V1 construction. When V1 conditionals are used in this language, they give the utterance an institutional, normative overtone. Our study has shown that grammar is not restricted to syntax and semantics. The distinctive features of a construction like V1-C also involve its se-

Verb-first conditionals in German and Swedish

259

quential position, discourse function, and the genre and modality in which the construction occurs. However, these are not static features but constantly in the making as part of the dynamics between the discourse participants and their roles in an interaction. Verb-first conditionals thus serve to enlighten different facets of constructional emergence: as an emerging construction with a path of evaluation and functional specification through language use in time, and as an emergent construction with versatile functional potentials that are specified and negotiated in the local, individual contexts of language use in real time.

Data sources referred to Birg. Aut. = GIC =

GSM =

HUSA =

LIMAS-Corpus (lim) =

MM-Corpus = SUC =

SÅINF =

St Birgitta’s notes in Swedish original. Telephone conversations to the Poison Control Centre in Sweden. Department of Scandinavian Languages, FUMS, Uppsala University. Transcription by Ulrika Sjöberg and Håkan Landqvist, revised transcription by Karin Ridell. Gymnasisters språk- och musikvärldar (The language and music worlds of high school students). Audio recordings collected at the Gothenburg University, Department of Swedish. Several transcribers. Språk och attityder bland helsingforssvenska ungdomar (The language and attitudes among Helsinki Swedish young people). Audio recordings collected at the University of Helsinki, Department of Scandinavian languages and literature. Transcription by Charlotta af Hällström. (compiled by the Forschungsgruppe LIMAS (Bonn, Regensburg). 500 text fragments with 2000 words each; from 1970 and 1971 (http://www.ids-mannheim.de/kl/ projekte/korpora/archiv/lim.html) . Mannheimer Morgen (http://www.ids-mannheim.de/kl/ projekte/korpora/archiv/mm.html): newspaper corpus. Stockholm Umeå Corpus version 2.0, SUC 2.0. Distributed by Språkbanken, University of Gothenburg, http:// spraakbanken.gu.se/parole/. Video recorded conversations collected in the project Samtal, åldrande och identitet (Conversation, ageing and identity). Department of Scandinavian Languages, FUMS, Uppsala University. Transcription by Karin Ridell.

260 TemaK =

VgL I =

Peter Auer and Jan Lindström

Audio recorded conversations from a courtroom process (A51), a consultation between a midwife and a pregnant woman (KBU), a focus group conversation between five farmers on genetically manipulated food (GML) and a police interrogation (P5). Tema Kommunikation, Linköping University. Transcription by Niklas Norén, Karin Ridell, Henric Bagerius. Äldre Västgötalagen (The older West Göta law).

References Auer, P. 1993 Zur Verbspitzenstellung im gesprochenen Deutsch. Deutsche Sprache 21, 193–222. 2000 Pre- and postpositioning of wenn-clauses in spoken and written German. In: Couper-Kuhlen, E. & B. Kortmann (eds.): Cause – Condition – Concession – Contrast. Cognitive and Discourse Perspectives, Berlin, 173–204. 2009a On-line syntax: thoughts on the temporality of spoken language. Language Sciences 31, 1–13. 2009b Projection and minimalistic syntax in interaction. Disc. Proc. 46, 2, 180–205. Auer, P. & S. Pfänder 2007 Multiple retractions in spoken French and spoken German. A contrastive study in oral performance styles. Cahier de Praxématique 48, 57–84. Behaghel, O. 1928 Deutsche Syntax – eine geschichtliche Darstellung, Heidelberg. Bergeå, H., C. Martin & F. Sahlström 2008 I don’t know what you’re looking for: Professional vision in Swedish agricultural extension on nature conservation management. Journal of Agricultural Education and Extension 14, 4, 329–345. Blatz, F. 1896 Neuhochdeutsche Grammatik. Vol. 2, Karlsruhe. Bruun, H., A. Malmsten & S. Palmgren (eds.) 2004 Svenskt lagspråk i Finland, Helsingfors. Declerck, R. & S. Reed 2001 Conditionals: A Comprehensive Empirical Analysis, Berlin. Diessel, H. 1997 Verb-first constructions in German. In: M. Verspoor, K. D. Lee & E. Sweetser (eds.), Lexical and Syntactical Constructions and the Construction of Meaning, Amsterdam, 51–68. DUDEN 2005 Duden – die Grammatik, Vol. 7, Mannheim. Erdmann, O. 1886 Grundzüge der deutschen Syntax nach ihrer geschichtlichen Entwicklung, Stuttgart. Gustafson-Capková S. & B. Hartmann 2006 Manual of the Stockhom Umeå Corpus version 2.0. http://spraakbanken.gu.se/ parole/ (retrieved May 18, 2009).

Verb-first conditionals in German and Swedish

261

Harris, A. & L. Campbell 1995 Historical Syntax in Cross-Liguistic Perspective, Cambridge. Hilpert, M. 2010 What can synchronic gradience tell us about reanalysis? Verb-first conditionals in written German and Swedish.. In: E.C. Traugott & G. Trousdale (eds.), Gradience, Gradualness and Grammaticalization, 181–201, Amsterdam. Hopper, P. 1975 The Syntax of the Simple Sentence in Proto-Germanic, The Hague. 1987 Emergent grammar. Berkeley Linguistics Society 13, 139–157. 1998 Emergent grammar. In: Tomasello, M. (ed.) The New Psychology of Language, Mahwah, NJ,155–175. Iatridou, S. & D. Embick 1994 Conditional inversion. Proceedings of the North East Linguistic Society 24, 189–203. Jespersen, O. 1940 A Modern English Grammar on Historical Principles, Part 5, Vol. 4, Syntax, Copenhagen. Lassus, J. 2010 Betydelser i barnfamiljsbroschyrer. Systemisk-funktionell analys av den tänkta läsaren och institutionen, Diss. E-thesis – Electronic Publications at the University of Helsinki, , accessed 17 June 2010. Lerner, G. H. 1996 On the “semi-permeable” character of grammatical units in conversation: Conditional entry into the turn space of another speaker. In: E. Ochs, E. A. Schegloff & S. Thompson (eds.), Interaction and Grammar, 238–276, Camebridge. Lindström, A. 2003 Skärningspunkter mellan sociala och språkliga strukturer i studier av tal-i-interaktion. Folkmålsstudier 42, 91–112. Lindström, J. & S. Karlsson 2005 Verb-first constructions as a syntactic and functional resource in (spoken) Swedish. Nordic Journal of Linguistics 28, 97–131. Mörnsjö, M. 2002 V1 declaratives in spoken Swedish: syntax, information structure, and prosodic pattern, Lund: Department of Scandinavian languages, University of Lund. Molencki, R. 1999 A History of English Counterfactuals, Katowice. Önnerfors, O. 1997 Verb-erst-Deklarativsätze. Grammatik und Pragmatik, Stockholm. Paul, H. 1920 Deutsche Grammatik, Halle. Schmitt, R. 1992 Die Schwellensteher. Sprachliche Präsenz und sozialer Austausch in einem Kiosk, Tübingen. Teleman, U., S. Hellberg & E. Andersson 1999 Svenska Akademiens grammatik, Stockholm. Van den Nest, D. 2009 Should conditionals be emergent …: Asyndetic subordination in German and English as a challenge to grammaticalization research. In: A. van Linden, J.-C.

262

Peter Auer and Jan Lindström

Verstraete & K. Davidse (eds.) Formal Evidence in Grammaticalization Research, Amsterdam, 93–136. Weinrich, H. 2003 Textgrammatik der deutschen Sprache, Vol. 2, Mannheim. Wessén, E. 1956 Svensk språkhistoria III. Grundlinjer till en historisk syntax, Stockholm. Wirdenäs, K. 2002 Ungdomars argumentation. Om argumentationstekniker i gruppsamtal. Nordistica Gothoburgensia 26, Göteborg. Zifonun, G., L. Hoffmann, B. Strecker et al. 1997 Grammatik der deutschen Sprache. Vol.3, Berlin.

Action, prosody and emergent constructions: The case of and

263

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

Action, prosody and emergent constructions: The case of and

1

Introduction

As Bybee has pointed out, language may have evolved from a set of relatively short utterances consisting of first one and then two units which bind together “via the concatenation of preformed chunks” to produce much longer utterances (2002: 131). Bybee proposes a Linear Fusion Hypothesis to capture this process, claiming that “elements that are frequently used together bind together into constituents” (2002: 109). It is repetition, she claims, which serves as the “glue” that binds the items into an emergent constituent (2002: 111). And because the amount of repetition encountered with units which are used together can vary, i.e. be gradient, so too the constituency which emerges from this repeated togetherness is gradient. These observations are highly pertinent for the topic of our paper, the emergence of constructions with and. The point we wish to make, both with respect to and-patterns and in general with respect to emergent constructions, is that as a contributor to structural emergence togetherness implicates not only syntactic/semantic cohesion but also two further dimensions: (a) togetherness of action, and (b) togetherness in prosodic/phonetic form. We will present evidence to support the argument that these other forms of togetherness are also necessary for the emergence of structure in use, using data from a collection of ands culled from a set of American English telephone conversations.1 For present purposes, we restrict ourselves to cases of and conjoining verb phrases only.2 That is, we include predicate conjunction as e.g. in “the little edge had curled up and was showing red” (SBL 2:1:8:6) or “Missiz Kelly looks a little younger when you get in and kind of size her up” (SBL 2:1:8:4) but exclude clausal conjunction as found in “I used a suppository yesterday morning and nothing happened” (SBL 2:1:8:2). 1

2

Our primary source has been 29 transcripts from the CallHome corpus, a set of telephone calls recorded in the US in the middle of the 1990s, supplemented by the SBL and NB corpora, two collections of telephone calls recorded in California in the early 1960s. Many of our findings hold as well for conjoined noun phrases, but for reasons of space we are unable to treat them here (see Barth-Weingarten, 2010, under revision).

264

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

The task we have set ourselves with respect to the set of widely divergent realizations of VP and VP found in our data is to study the degree of cohesiveness between the conjoined parts and consequently to identify potentially different layers of use (Hopper 1998) of and as a VP-conjunctional. As we will see, not all VP conjoins linked by and are used in ways which, given frequent recurrence, would facilitate fusion between the two parts. Our aim is thus to work out under what circumstances the two parts of VP and VP can be said to be “used together” such that through repetition, they could conceivably “fuse together” and become construction-like, possibly even yielding hendiadics such as go ahead and X, sit down and X as described by Hopper (2001a, 2001b). In the following, we will first discuss togetherness of action (section 2) and then togetherness in prosodic/phonetic form (section 3). Thereafter, we will point to some wider implications of our findings (section 4). It is to be noted throughout this paper that the separation of togetherness in prosody and togetherness of action is not just induced by the linearity of writing, but also by the fact that, while they often co-occur, there are also instances in which the two phenomena can appear separately (cf. section 4).

2

Togetherness of action

With the notion of action we refer to what speakers are doing when they use a “burst of language” to take a turn-at-talk (Schegloff 1996: 53–54). In other words, we assume that the units which speakers use to construct turns-attalk house actions: as a rule, one turn-constructional unit implements one action. Turn-constructional units can be formed by sentences, clauses, phrases or words – whatever, in the given context, it takes to carry out the requisite action. Togetherness of action is thus present when the string of words used is part of a simple turn-constructional unit and implements a single action. With respect to VP and VP, the object of study here, it may be tempting to think that this string inherently enjoys togetherness of action simply because and VP is not syntactically viable on its own. However, when we examine conversation, we find that speakers do not always use the linear sequence of elements VP and VP to carry out a single action. In the following case, for instance, one speaker produces a first VP as part of a clause substantiating the assessment she is making, but it is her interlocutor who adds a second VP to it, thereby displaying affiliation:

Action, prosody and emergent constructions: The case of and

265

(1) and VP2 as a separate action by a next speaker Spaghetti (SBL027: 223+224)3 (Claire and Marylou are weighing up the pros and cons of preparing a bigsize enchilada dinner for an upcoming gathering at Marylou’s home.) 1 Cla: you’re gonna have to have a lot of people Fworking it’s not 2 like having a spaghetti dinner where you have a great 3 f big bunch of (.) sau:ce all ready and all you have to do is 4 f hhhhh pull the spaghetti out of the hot wa:te[r. 5 Mar: [ri:ght’n tee6 ye:s. 7 f and just keep making another batch as you need it. 8 Cla: that’s Fri:ght.

With pull the spaghetti out of the hot water delivered on a final falling pitch contour (line 4), Claire has completed her point that preparing an enchilada dinner requires more work than merely cooking spaghetti. In next turn Marylou acknowledges Claire’s point and begins to expand on it right’n tee-4 but breaks off to confirm her agreement with a simple yes. She then recycles and repairs the aborted unit, which in the event turns out to be more detailing in support of the argument that making a spaghetti dinner is easier than an enchilada one. Coming as it does after an initial token of agreement, Marylou’s addendum provides strong affiliation with Claire’s assessment. The way she accomplishes this is by designing a unit built on the syntax of the turn which she is affiliating with: Claire: all you have to do is pull the spaghetti out of the hot water. Marylou: and just keep making another batch as you need it. So in (1) it is the recipient of a possibly complete VP1 who makes use of and VP2 as a device to extend the interlocutor’s prior turn and thereby display affiliation with it. This recipient’s and VP2 implements a separate action above and beyond the action which the prior speaker’s clause with VP1 implemented, in the sense that it is not prolonging the prior action but rather taking up an affiliative stance towards it. Although there is of course a sense in which and VP2 depends syntactically on VP1 and the clause it is a part of (they share the same understood subject and infinitival verbal form5), nevertheless the linear sequence VP1 and VP2 is produced by two different 3

4 5

This and all further excerpts from the SBL and NB corpora are transcribed according to Jefferson’s system (cf., e.g., Schenkein 1978) but with normalized orthography. The tee- in line 5 is arguably a slip of the tongue for keep (see line 7). Cf. Givón (1993), Stefanowitsch (1999, 2000), Hopper (2008).

266

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

speakers and its parts are carrying out separate actions: assessing and subsequently agreeing with that assessment.6 The pieces VP1 and VP2 are thus not being “used together” here in a way which would facilitate fusion in Bybee’s sense. Yet even when VP1 and VP2 is produced by a single speaker, it is not necessarily carrying out a single action: (2) and VP2 as an increment to same-speaker talk Bridge game (SBL 022 2:1:8) (Earlier in this conversation Bea has casually introduced the possibility of her friend Nora coming over to play bridge that afternoon. Several minutes later Nora returns to the topic.) 1 2 3 4 5 f 6 7 8 f 9 10 11 12 13 14

Nor: Bea: Bea: Nor: Bea: Nor: Bea: Nor: Bea:

Well who would you drum u:p. (.) I don’t kno:w. (0.2) I would have to think that over. (.) [eh hu]h heh] [ehhh ]h e h]heh and FFI(h)ND FFOU::T. WELL i:f you WANNA play why fi:ne uh: becuz uh (1.6) [I-] [I:]-: should work but I: can (.) put it o:ff. (.) I’ll see what I can do:.

Nora implies with her question in line 1 that she might be interested in a bridge game that afternoon depending on who else would be there. Bea’s response, however, is somewhat non-committal. In line 3 she professes not to know who else she might enlist to play, and in line 5 she indicates that she would have to think that over (VP1). This may be a tacit allusion to the fact that only some of the women they know would be desirable as bridge partners, an implication which Nora’s subsequent laughter implies she understands. At this point Bea’s responsive action is complete, yet following the shared laughter with Nora, Bea extends her turn in line 8 by adding and find out (VP2). This VP2 refers back to the VP1 in line 5 for its syntactic/semantic interpretation, yet at the same time it constitutes a separate action from the one which was brought to conclusion there. Bea is now suggesting that 6

As we shall see below, production by a different speaker is, however, not a sufficient criterion for being a separate action, as witnessed by the phenomenon of joint turn construction (Lerner 1996).

Action, prosody and emergent constructions: The case of and

267

not all the bridge players she might contact would be able to come, another possible obstacle to the spontaneous bridge game being considered. When does a same-speaker turn continuation constitute a separate action and when not? This is a delicate issue fraught with unresolved problems (Couper-Kuhlen and Ono 2007). But this much can be said about the case at hand: As a stretch of speech displaying syntactic and semantic symbiosis with a prior potentially complete turn-constructional unit, and find out in (2) above would qualify as an increment (Schegloff 1996). Increments, it has been argued, do not always merely extend the prior action; on occasion they step aside from it and evaluate or take a stance with respect to it. This then constitutes a new and separate action (Ford, Fox, and Thompson 2002). Increments which extend the prior action have been shown to be prosodically and phonetically fitted to their hosts, in the sense that they are not significantly higher or louder than the prior unit nor do they display marked changes in articulation or rate of articulation. Instead, they are produced in such a way as to continue the prior prosodic delivery (Walker 2004). In (2) there are both semantic and prosodic reasons for treating and find out as a separate action: for one, the scope of the event find out is dependent on the result of the event think that over (line 5). It is only for those potential players whose names her reflection process comes up with that Bea will need to determine availability. This temporally separate event thus builds on the result of the first. Prosodically, the phrase and find out is significantly louder and higher than the phrase think that over. In this respect it is not delivered like a prolongation of what precedes but like something new. If we accept on these grounds that the syntactically incremental and find out is a separate action from that of its host, then here too the use of the pieces VP1 and VP2 is not one which would further fusion, since they lack togetherness of action. To summarize the discussion so far, we have identified two inter-turn patterns for the use of VP-conjunction with and : (I) Cross-speaker production of VP1 and VP2 A: ( clauseVP1 ) . (. = final intonation) (extension of interlocutor’s prior turn) B: and VP2 (II) A: B: A: 7

Same-speaker incremental production of VP1 and VP2 ( clauseVP1 ) . (. = final intonation) (.)7 and VP2 (extension of speaker’s own prior turn)

In our example (2) there is a micro-pause following the provisional end of the speaker’s turn constructional unit. However, an increment can of course also be produced without this micro-pause (Couper-Kuhlen and Ono 2007).

268

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

What these patterns make clear is that the linear sequence VP1 and VP2 is not always produced as a unified whole but, as here, can be composed of two pieces which differ significantly in terms of dependency. In patterns (I) and (II) the first piece, VP1, is typically an integral part of a clause with independent status. In fragments (1) and (2), which instantiate these patterns, the clause containing this piece is treated as transition-ready. Recipients orient to this transition-readiness by coming in next. There is no projection of any second piece to come. When forthcoming, the second piece, however, is marked as dependent on the first piece in that (a) it is introduced by and, and (b) its VP requires localization of a prior VP and inspection of the clause containing it for syntactic parsing and semantic interpretation.8 In sum, the patterns (I) and (II) show that the two pieces VP1 and VP2 can be produced for the implementation of separate actions and that when this happens, these pieces differ significantly in terms of dependency. In the case of (I), the two parts are even produced by different speakers. This draws our attention to the fact that the linear sequence VP1 and VP2 is not inherently a single piece, or unified action, but that, being produced in real time, it can come about through the collaborative work of two speakers or through the online work of a single speaker and constitute two separate actions. Togetherness of action with VP1 and VP2 is not a given in conversational interaction, but an achievement.9 In this sense, when describing the role of linear fusion in the grammaticization process, it is crucial to specify that linear fusion can only come about if the elements being used are together not only syntactically/semantically but also in the sense of action. Interestingly, the VP conjunction with and behaves in this respect similar to other “conjunctionals” (Jefferson 1980), including because (CouperKuhlen 1996), so (Raymond 2004) and but (Mulder and Thompson 2008). These linking elements all have in common that they introduce bits of talk which are dependent on prior talk for their full interpretation. All can be produced both by the same speaker of the prior talk or by another speaker.10 8

9

10

It is true that all coordinative structures have a dependent second part, but the point is that not all have a first part which is initially presented as an independent stand-alone. Cf. Hopper and Thompson (2008) for a similar conclusion with regard to other “biclausal” constructions, e.g. pseudo-clefts and extraposition. So can be used unproblematically in a turn-final position or as a stand-alone (Raymond 2004), while but is still regionally marked when it ‘dangles’ (Mulder and Thompson 2008). (There is some indication that dangling but may be coming into use also in varieties other than Australian English: Ursula Lenk personal comment, Barth-Weingarten 2007.). Because and and do not appear to have moved as

Action, prosody and emergent constructions: The case of and

269

And in their conjunctional use, all have in common that their speaker may be carrying out a separate action by deploying them to introduce further talk. All thus have the potential to be parts of patterns which are multi-actional.11 So the question which arises now is: When can it be said that two pieces such as VP1 and VP2 are implementing a single action? There appear to be at least two different ways in which VP1 and VP2 can be deployed to create a single turn-constructional unit and implement a single action in talk. In one case a speaker projects prosodically at the end of the first part that more is to come12 and this “more” turns out retrospectively to be another VP linked to the first by and. The result is a kind of compound turn-constructional unit, produced on the fly: (3) VP1 and VP2 as compound turn-constructional unit Gladys (SBL 2:2:3:7) (Claire and her friend Chloe are talking about a recent bridge game and how poorly Gladys, one of the other women there, played.) 1 2 3 4 5

Cla: Chl: Cla: Chl: Cla:

6 7 8 9 10

Cla: Chl: Cla: Chl:

11 Cla: 12 f Chl: 13 Cla:

11

12

FI can just get e-uh ((smiling)) hhehhh hh[ehh [she say[G:s uhG ] [d i d y][ou notice]( ) [is that al]right with you dear? hhh she ta:lks [so En]glish you know]hhh [yah. ] she’s Fsu]ch a do:ll a[nd you know= [uh-h =she almost makes me pull my hair ou[t when]I w a]tch her pla:y [ehhhhh]Fheh hehF] hhhh I shoul[dFget u:p, [ hhhh

far down this putative cline of independence – at least not in English. In Japanese, by contrast, kara (=‘because’) can dangle unproblematically in final position (Mori 1999) as can the Finnish turn-final topic-closing particle että (Koivisto 2006) and Icelandic eja (‘or’) (Blöndal 2006). To the extent that they are produced by two speakers, or at least allow for this possibility, they are also eminently inter-actional. There is presumably an implicational relationship here between actionality and prosody such that multi-actionality requires two parts separated by a strong prosodic break (i.e. final intonation). There are to our knowledge no cases of multi-actionality without this kind of strong prosodic break. However, there are cases of single-actionality which have strong prosodic breaks, as we show later in this paper (cf. ex. (8) in section 4).

270

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

14 f Chl: and go (to) the bathroom or put some ma(h)ake[(h)up (h)on= 15 Cla: [u 16 Cla: =h[n hn 17 Chl: [bec(h)uz she Fjust about Fhhehhh= 18 Cla: =hhhh 19 Chl: dri(h)ves me[inFsane she won’t get those Ftrumps out hhe:hh 20 Cla: [°he° 21 Cla: I know

Chloe’s complaint about her bridge friend Gladys begins in line 10 with she almost makes me pull my hair out when I watch her play, to which Claire provides appreciative laughter (line 11). Underlining how excruciating she finds Gladys’s game, Chloe now proposes that rather than watch Gladys play, she should get up (line 12) and go to the bathroom or put some make-up on (line 14). Retrospectively, these two lines form a VP1 and VP2 pattern: that is, at the end of line 12 the speaker projects, via continuing intonation, that more will follow, although at this point it is not yet clear what exactly it will be. In line 14 it turns out to be the addition of a further event, or alternative events, (VP2), which together with the first event of getting up (VP1) are being proposed as a substitute to watching Gladys play bridge. This VP1 and VP2 pattern, although it is produced on the fly, is implemented by the speaker retrospectively as a single action, namely proposing a set of alternative activities to sitting there and watching Gladys play. In this sense, it instantiates a togetherness of action; the two pieces are produced in a way that makes them susceptible to further merging. Yet there is another possibility, namely that the two parts, VP1 and VP2, are welded together and produced in one go within a simple turn-constructional unit: (4) VP1 and VP2 as part of a simple turn-constructional unit Money borrower (SBL 011: 38) (Bea and her friend Dinah are talking about a mutual friend who is a compulsive money borrower.) 1 2 3 4 5 6 7 8 9 10 11 12 f

Bea: Din:

Bea: Din:

=Well that’s what I mea:n. It isn’t the: money as [much as ] [No money] doe:s: didn’t mean anything or I’d do it for FM-ARty only ee-it’s: it’s just this: uh:: (0.7) hh Fyou kno:w now for instance wushe: used to borrow from me she borrowed twice:: from me once. Uh huh hhh A:n:: (0.5) I wu-I was sitting in her hou:se and: Re:j Oakley came’n deFlivered something.

Action, prosody and emergent constructions: The case of and 13 14 15 16 17 18 19 20

Bea: Din:

271

and she: w- h said she didn’t have the cha:nge would I loan her the money to pay him and she’d pay me later:. and I: said well you already borrowed from me twice: and never offered to pa:y Oh huh[: [hhh And Fshe produ:ced money enou:gh to pa:y Rej Oakley and me bo:th.

To substantiate her friend’s odd behavior with respect to borrowing money, Dinah announces a story in lines 6–8 with the preface now for instance she used to borrow from me she borrowed twice:: from me once. After a go-ahead from Bea (line 9), Dinah depicts the setting with I was sitting in her hou:se (line 11) and the precipitating event with Re:j Oakley came’n delivered something (line 12). Significantly, the VP1 and VP2 pattern here, came and delivered something, is part of a simple turn-constructional unit carrying out a single action, viz. delivering the precipitating event of the story. In fact, these two VPs are not only linked together for the implementation of a single action, they have actually fused together semantically to the point of referencing a single event: Rej is understood not to have first come and then delivered something, but to have come and delivered something at one single point in time.13 This is an indication of its hendiadic status. Our point is that the kind of semantic fusion which has taken place here with come and deliver can only happen if the two pieces in question VP1 and VP2 are used together frequently as part of a simple turnconstructional unit implementing a single action in talk. To summarize then, we have identified two further, intra-turn patterns of use for VP1 and VP2, which could be schematized as follows: (III) Compound turn-constructional unit with VP1 and VP2 A: ( clauseVP1) , and VP2 (, = prosodic boundary marker of incompleteness) (IV) Simple turn-constructional unit with VP1 and VP2 A: ( clause VP1 + and VP2 ) (+ = no prosodic boundary marker) These two patterns show an increasing amount of integration between the component pieces. In (III) the first piece on its completion projects more to come prosodically, yielding post-hoc a compound turn-constructional unit produced on the fly and implementing a single action. In (IV), the two pieces are produced in one go as part of a simple turn-constructional unit implementing a single action. Our argument is that the fusion of the two pieces, 13

For further criteria for monopredicativity cf. Hopper (2008), Stefanowitsch (2000) and section 4.

272

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

which might then lead to the emergence of a hendiadic construction, can only take place if the pieces of the linear sequence VP1 and VP2 are deployed as part of a single turn-constructional unit implementing a single action. It is this kind of togetherness of action which, we would argue, is a first sine qua non for the emergence of an and-construction. Taken together, the patterns shown in (I)-(II) and (III)-(IV) illustrate four attested degrees of integration between two pieces, VP1 and VP2, ranging from multi-actional to uni-actional use. With only two of these, namely (III) and (IV), are the pieces produced in such a way that they could be said to deliver a single action. Yet, delivery as a single action is not the only prerequisite for potential fusion. Just as important, if not more, is the degree of prosodic/phonetic integration. We turn now to the notion of prosodic/phonetic togetherness as it applies to the linear sequence VP1 and VP2. As the previous examples have already shown, there are different degrees of prosodic/phonetic integration. Our argument will be that patterns (III) and (IV) represent idealized positions along a gradient of prosodic/phonetic integration, and that a high degree of prosodic/phonetic togetherness between VP1 and VP2 is in fact a second sine qua non for the emergence of and-constructions.14

3

Togetherness in prosodic/phonetic form

Prosodic/phonetic togetherness is understood here as the integrated production of potentially separate linguistic units by means of connective features such as smooth continuation of pitch, tempo and loudness as well as coarticulation and liaison effects. To our knowledge, in most of the prosodic literature on English, the integration vs. separation of intonation units is thought of as a binary distinction: linguistic material is said to be presented in either one, or two, prosodic units as the case may be.15 We provided clear 14

15

It goes without saying that some degree of semantic togetherness is also a precondition for the emergence of and-constructions. We do not discuss this aspect here because our approach is an external one: we take the perspective of a language recipient, who, upon the occurrence of appropriate cues, interprets and stores something as one or two units. An alternative perspective would be that of the speaker, who produces two events/actions in one or two units because for him/ her they are semantically connected. In this perspective, it is a closer semantic relationship which is reflected in closer prosodic integration (cf. Givón 1997, 1993). Cf., for instance, Crystal (1969), Cruttenden (1986), du Bois et al. (1992), but for a more differentiated view Jun and Fougeron (2000), Degand and Simon (2009). Schönherr (1997) and Birkner (2008) have also identified a continuum between prosodic integration and disintegration for German (see also Barth-Weingarten, submitted).

Action, prosody and emergent constructions: The case of and

273

cases of these two options in excerpts (3) and (4) above.16 In excerpt (3) VP1 and VP2 are realized as parts of two clearly separate prosodic units: at the end of get u:p, there is final lengthening, a level pitch movement which remains high in the speaker’s range and a clear release of the final plosive in up. This is followed by a micro-pause and a glottal onset at the beginning of and go (to) the bathroom, which is produced on a lower pitch level. Hence, except for the projecting intonational contour, none of the integrative prosodic/phonetic features mentioned above are present at the boundary between VP1 and and VP2 in (3). Instead we find prosodic breaks in pitch and tempo as well as phonetic disjunctions through audible plosive release and glottal closure. Excerpt (4), by contrast, illustrates maximal prosodic and phonetic togetherness of the linguistic material in VP1 and VP2. The VPs came and delivered something are realized in one go. There are no pitch or tempo changes between them which would indicate a prosodic boundary. The final pitch contour begins on delivered, which is the main accent. VP1 is not accented. VP1, and and the beginning of VP2 belong to the same rhythmic foot. In addition, and is cliticized onto the preceding came with a reduced vowel immediately following the nasal [m]. However, the binary distinction illustrated by (3) and (4) does not sufficiently reflect the varying degrees of prosodic/phonetic integration we find in our material. Rather, these clear cases appear to be only the endpoints of a cline with differing degrees of integration, or togetherness, in-between, similar to what has been described for German by Schönherr (1997) and Birkner (2008). What is of particular interest here are the micro-levels of not only prosodic, but also phonetic detail at the potential boundary between VP1 and and VP2. Consider excerpt (5) as a first illustration: (5) Intermediate degree of prosodic/phonetic integration between VP1 and VP2 Dust (CallHome, 4666: 425–433 sec.)17 (Marianne has been complaining to her mother about the noise and inconvenience created by the paving of their road, which was previously apparently rather dusty. Now she finds something good to say about it.) 16

17

Excerpts (1) and (2), of course, also provide instances of VP1 and VP2 separated by clear prosodic boundaries. But, in this part of our paper, we will concentrate on those cases that already meet the criterion of togetherness of action. This and all further CallHome excerpts are transcribed according to GAT 2 (Selting et al. 2009, to appear). A list of symbols used can be found in the appendix. A close transcription of the relevant phonetic details are provided in curly brackets and italics in a separate column next to the transcript.

274

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

1 MAR: °hh and it Also: u:hm2 °h wEll; 3 it it’s so deLIGHTful.= 4 f =to come down the /rOA:d | and not be / /FfOllowed by a clOud o’/ f /FDUST. f 5 [you kno]w, 6 MO: [yeah; ]

{rOAden}

After some hesitation (lines 1–3), Marianne now mentions a positive effect of the paving of their road which she can take delight in, namely her being able to come down the rOA:d | and not be F fOllowed by a clOud o’ FDUST (lines 3–4). Note that VP1 and VP2 here meet the criterion of togetherness of action and they depict a single event: Marianne is not so much delighted by her coming down the road, but by the fact that she is not followed by a cloud of dust when doing so. This one-eventness is reflected not only in the shared temporal reference point of the two VPs but also in the cataphoric use of it to refer to their conjunction (cf. section 4). In terms of delivery, VP1 and VP2 are produced with features which cue both prosodic integration and separation: On the one hand, there is final lengthening at the end of come down the rOA:d, and one might be tempted to put a unit boundary there. Yet, while the prominent syllable rOA:d has a falling pitch movement, all the following material up to FDUST simply continues this movement without any further prosodic break in terms of pitch or speech rate. This then speaks for prosodic integration. Furthermore, the material surrounding and is noticeably rhythmic, falling into three rhythmic feet. Yet, the foot boundaries are nowhere near the syntactic boundary which orthodox grammar would put between VP1 and VP2.18 So the rhythm established seems to further integrate the VPs connected by and. This prosodic integration is paralleled by integrative phonetic features: and does not have a glottal beginning. Instead, it is glued onto the preceding rOA:d in a way which resembles liaison: the final alveolar plosive [d] of rOA:d is released into the vowel of and, so that in effect and is cliticized to the previous word. In addition, and has a reduced vocalic quality and its final /d/ is deleted altogether.19 Thus, in cases such as this one, we find a degree of prosodic/phonetic togetherness which lies somewhere between that of patterns (III) and (IV). Even if we decided in favour of a prosodic boundary, it would be difficult to know where to place it. 18 19

But for a less fixed view on syntax and syntactic constituents cf. Steedman (2000). Noticing such phenomenon of course also poses questions with regard to the way this, and further, instances of its kind should be transcribed (cf. Barth-Weingarten, submitted).

Action, prosody and emergent constructions: The case of and

275

Other instances involve prosodic boundaries which are similarly weak and have an indeterminate location, if for slightly different reasons: While fragment (5) exhibits clear final lengthening, there are other cases where this disintegrative cue is attenuated and the juncture is accompanied by more integrative features. Example (6) is a case in point: (6) Further intermediate degree of prosodic/phonetic integration between VP1 and VP2 Clinton (CallHome, 4247: 182–207 sec.) (Alec is updating a friend abroad on recent political events in the US, including a statement by then-president Bill Clinton concerning affirmative action as a means of countering racial discrimination.) 1 2 3

ALEC:

4

f f

5 6 7 8 9 10

but at the same tIme_2uh the anAlysis in the nEws°h shall we say the conFSERvative analysis in the news is that; {Ouɾen} °hh clInton just came /Ou:t | Gand | / / /FsAi:d that /he: _FdOesn’t believe°h in quOta systems°h and in revErse discriminAtion-= =but that he dOEs believe=that affirmative Action is nEcessary°h to move-uh you know black amEricans | °h fOrward-= =and to give them the opportUnities | that they’ve been denIED.

The relevant instance of VP conjunction occurs in the quotation of the conservative news analysis of Clinton’s statement: clInton just came Ou:t | G and| FsAi:d that … (line 4). While there are a sufficient number of prominent syllables to divide the VPs conjoined here into two prosodic units, doing so is problematic. The reason for this is that around the potential unit boundary we find various kinds of prosodic and phonetic cues integrating the units. The and itself is somewhat more integrated with the preceding item than in excerpt (5) – it is rhythmically cliticized to the previous word and also has a reduced vowel. In addition, the final consonant of Ou:t is reduced to an alveolar flap, producing a liaison effect between Ou:t and and. At the same time, and exhibits a minor dip in pitch: retrospectively, it could be heard either as prolonging the pitch movement on Ou:t or as initiating the pitch movement on said. What is also noticeable here is that the material surrounding and is strongly rhythmic. Similar to excerpt (5), this integrates the two verbs surrounding and. Yet, in (6) the rhythmic feet are rather short, with rhythmic

276

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

beats on either side of and. This results in a potentially disintegrative foot boundary following and. On the other hand, the VPs surrounding and and providing the ictus syllables have equally prominent accents, as reflected in their parallel, slightly lengthened realization. This cancels out the potentially disintegrative effect of the attenuated final lengthening in Ou:t. The result is a kind of balanced realization that could be said to distinguish instances of this kind from those with clear final lengthening as in fragment (5). Yet, perhaps more important than such deliberations is that it is again difficult to know whether and where to place a prosodic boundary in cases such as (6). We are witnessing another “in-between” case of prosodic/phonetic togetherness in VP conjunction. Schematically:

ex. 5 pattern III

pattern IV ex. 6

Schema 1: Prosodic-phonetic togetherness with patterns (III–IV) and examples (5–6).

The pattern of variation represented in schema 1 is not unusual in conjunction with processes of language change.20 A similar kind of phonetic variation has also been observed by Bybee and Scheibman (1999) in word-level patterns undergoing change.21 They found four categories into which they could group tokens of I don’t according to the degrees of phonetic reduction of the consonant and the vowel. With high-frequency verbs, greater reduction correlated, to a certain extent, with a tighter internal structure and a different pragmatic function. While we are not (yet) in a position to make specific claims about the direction of change in the case of VP conjunction,22 the variation observed here is at least suggestive of emergent constructionhood.23 20 21 22 23

Cf. Hopper (1987, 1991), also Barth-Weingarten and Couper-Kuhlen (2002). Also cf. Bybee (2001). But cf. Barth-Weingarten (in prep.). Hopper (2008) describes a similar fusional process culminating in the take NP and construction.

Action, prosody and emergent constructions: The case of and

277

The prosodic/phonetic gradience shown with the examples above would seem to accomplish an important task in the emergence of and-constructions: In order for the two conjoins, here VP1 and VP2, to be able to fuse, they must not only be deployed in a single turn-constructional unit but, in addition, must be part of a single prosodic contour. This means that any prosodic boundary between VP1 and VP2 must weaken, if not disappear altogether. This in turn will allow further phonetic links to establish themselves between the constituents. Excerpts (3)–(6) show just such prosodic and phonetic operations at work. For a more detailed consideration of this process, it will be helpful to examine the prosodic and phonetic features described above separately. First, prosodic togetherness: While Bybee and other scholars of grammaticization have concentrated so far on phonological processes at the word level,24 we would like to stress that in looking at changes involving units beyond the (phonological) word, prosodic processes shift to the centre of attention. In the case of VP conjunction, the adjacent VP conjoins can only merge into a simple turn-constructional unit (pattern IV) if the prosodic boundary cues that are present in their realization as a compound turnconstructional unit (pattern III) are reduced or indeed eliminated. In essence, this can be assumed to happen via processes similar to those which Bybee has argued apply in the automatization of word-size units, i.e. via “the creation of a neuromotor routine that is processed as a single unit and can undergo phonological reduction, defined as a decrease in magnitude of gestures and an increase in their overlap” (Bybee 2006: 720). If the magnitude of prosodic gestures, such as pitch movement and speech rate change, which are usually present at prosodic unit boundaries is reduced25, this will result in a dissolution of those boundaries. And because prosodic boundary cues function in clusters, the dissolution is likely to happen, not all at once, but gradually, with differing cues as well as differing extensions of these cues being involved at different stages. This might then explain the varying degrees of prosodic togetherness encountered in excerpts (3)-(6) above: the variation in the realization of final lengthening, pitch movement and rhythmic integration would not be happenstance, but a clustering of continuously variable features at points along a continuum. 24

25

Cf., e.g., Bybee (2001), Bybee and Scheibman (1999), Bybee and Thompson (2000) and Scheibman (2000). For an earlier extension of Bybee’s notion of gesture reduction in language change to prosodic features on the level of the (phonological) word cf. Wichmann (2005) and Wichmann et al. (2009).

278

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

Second, phonetic togetherness: Bybee and others have mentioned a range of phonological processes that arise as part of the automatization process, among them alternations (a/an), liaison, coarticulation, lenition such as flapping, vowel reduction (centralization, shortening), and loss of vowels and consonants.26 Similar phonetic phenomena are observable in VP conjunction: encliticization of and connected with shortening, vowel centralization, at times even vowel assimilation, flapping of pre-and /t/, loss of the and-final /d/ and sometimes even of the vocalic quality of and altogether. What is to be noted is that these phonological processes co-occur with prosodic boundaries that are weak and/or indeterminate. In fact, Bybee and Scheibman (1999) claim that phonetic reduction only occurs when something is already stored as a processing unit27; according to them, such phenomena are “phonological indicators of constituency” (1999: 578)28, i.e. the effect of fusion. If this is the case, how then can phonetic togetherness at the same time be a pre-requisite for fusion, as we have claimed? We would like to argue that many of the phonetic phenomena mentioned above can, as such, occur at different stages in the process of language change. In point of fact, most of these phonetic changes are due to fast speech rules.29 In contrast to phonological rules of connected speech (external sandhi), fast speech rules operate independently of syntactic and morpho-lexical structure. As such they are omnipresent in language production and begin functioning as soon as adjacent items, of whatever kind, are produced fluently, i.e. without pause or separation by prosodic boundaries. In principle then, they can occur already on the very first exemplar of a construction (to emerge). If such an item is stored upon its first occurrence and in the form it occurred in (Bybee 2006), then the reduction and coarticulation occurring with fast speech can shape that form’s representation (cf. Bybee and Scheibman 1999: 578), and hence can be a major factor in its emergence. After storage, the new units can become entrenched through frequent recurrence of similarly reduced and coarticulated forms, or more new forms can develop through further reduction and coarticulation (cf. Bybee 2006). In this sense then, phonological processes may not only occur after fusion, but can set language change in motion, and instances of phonetic togetherness such as described above can 26

27 28 29

Cf., e.g., Bybee (2001, 2002), also Bybee and Scheibman (1999), Bybee and Thompson (2000), Scheibman (2000). Called ‘chunking’ by Haiman (1994: 8). Also cf. Bybee (2001: 161). Cf. Kaisse (1985), also Shockey (2003).

Action, prosody and emergent constructions: The case of and

279

lead to fusion. Support for this hypothesis can be found in two observations: one by Bybee, who claims that phonological reduction is somewhat ahead of other change, such as the extension of pragmatic use (2001: 160). The second one is by Kaisse, who points out that fast speech rules can become conventionalized and turn into phonological rules of connected speech (1985: 35). Both observations suggest that phonetic processes originating in fast speech can happen first, i.e. before entrenchment. To sum up, we have argued in this section that not only togetherness of action but also prosodic and phonetic togetherness are pre-requisites for the syntactic/semantic fusion of VP1 and VP2. Moreover, we have suggested that prosodic togetherness enables phonetic togetherness. At the same time, phonetic reduction can occur with the first instance of a construction.

4

Wider implications

In the case of VP conjunction with and, togetherness in action and togetherness in prosodic/phonetic form have facilitated syntactic/semantic fusion and led to the emergence of a special hendiadic construction with the rough schematic form of VP and VP.30 Core examples are: go and X, come along and X, come up and X, stand there and X, sit around and X, try and X. According to Hopper (2008: 260), the main criterion distinguishing hendiadic constructions from synthetons is their monopredicativity: although two verbs are used, only one event is predicated. Hendiadic status is also reflected in a variety of more specific semantic, morpho- and lexico-syntactic features, such as semantic focus on VP2, (incipient) bleaching of VP1, semantic and syntactic interlacing, i.e. the sharing of predicates, tense, aspect and mode, actants and location, a preference for uninflected forms with VP1, a tighter morphosyntactic integration31 as well as a lack of syntagmatic and paradigmatic versatility.32 Although Sánchez (1999) observes greater prosodic and phonological integration between the parts of a hendiadic, Hopper claims that “there is no evidence” that the constituents of hendiadic constructions “are linked by 30

31

32

Cf. Hopper (2001b: 151), also (2001a, 2008), also Stefanowitsch (1999, 2000). In grammars this phenomenon is referred to as pseudo-coordination (Quirk et al 1985: 987–988, cf. also Sánchez 1999) or coordinated binomial phrases (Biber et al. 1999: 1030–1036). This refers both to the possibility of inserting adverbs or discourse markers between VP1 and VP2 as well as to their negation, gapping etc. as a complex unit. The reference here is to specific verbs occurring in a specific order. Cf. Hopper (2000, 2001a, 2001b, 2008), also Quirk et al. (1985: 979), Lehmann (1988: 204), Pullum (1990), Stefanowitsch (1999), Sànchez (1999).

280

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

any special suprasegmental feature such as intonation” (2008: 261). Yet, a pilot study of 96 instances of VP conjunction from the CallHome corpus showed significant correlations between monopredicativity and features of prosodic togetherness, among them lack of pitch reset with and and lack of prosodic breaks such as filled/unfilled pausing but presence of latching at the potential unit boundary. Those VP conjunctions which were interpretable in context as two events, in contrast, exhibited significantly more disintegrative prosodic/phonetic features (see Tables 1 and 2; cf. Barth-Weingarten in prep.).33,34 Table 1: Correlation of event structure with pitch reset on and in 96 candidate instances of the hendiadic construction35 in the CallHome corpus (*, ** and *** indicate 10 %, 5 % and 2 % chance of error respectively) 1 event (n = 12)

1–2 events (n = 34)

2 events (n = 50)

2 (17 %)**

15 (44 %)

25 (50 %)

pitch continued on and

10 (83 %)**

18 (53 %)

25 (50 %)

pitch not analyzable



pitch reset on and

1 ( 3 %)



Table 2: Correlation of event structure with prosodic breaks around and in 96 candidate instances of the hendiadic construction in the CallHome corpus prosodic breaks around and

1 event (n = 12)

no break

1–2 events (n = 34)

2 events (n = 50)

8 (67 %)***

9 (26 %)

14 (28 %)

break

4 (33 %)***

25 (74 %)

36 (72 %)

selected kinds of break: – unfilled pause – breathing – latching

– – 3 (75 %)**

2 ( 8 %) 1 ( 4 %) 11 (44 %)

4 (11 %) 9 (25 %)** 16 (44 %)

33

34

35

Cf. Givón (1991: 86, 1993) for pauses with serial verb constructions and the proximity principle; for the phenomenon of grammaticalization and fusion of intonation units in general cf. also Croft (1995). The results also show that there are a significant number of cases where event structure is ambiguous between one- and two-event interpretation. The statistics included only instances of VP conjunction with the subject elided in the second conjunct.

281

Action, prosody and emergent constructions: The case of and

These findings suggest that the degree of prosodic/phonetic integration between VP1 and VP2 might be a further cue to the presence of a hendiadic construction. Adopting a prosodic/phonetic perspective could even provide us with a tool for spotting instances of incipient hendiadys, namely conjoined VPs with prosodic/phonetic integration which do not (yet) belong to the hendiadic core but may nonetheless exhibit some degree of syntactic/ semantic fusion. (7) Conjoined, potentially incipient hendiadic VPs with prosodic/phonetic integration Older home (CallHome, 4065: 1002–1020 sec.) (Ava is telling her friend Ben about meeting a former fellow student, Julie, who has become a professional architect and now remodels old homes.) Ben: Ava: Ava: Ben: Ava:

Ben: Ava:

Ben: Ava:

[good. and [then we saw leo and JULie:; at: FCHR[ISTmas ti:me. [uh-huh, .hhh and FTHEY’RE doing grEa:t.= =uh:m– they had jUst [moved to[he’s in new; YORK now rIght? a- REALly nice house in wEstchester. YEA:H, an o[l[gOod. an older hom:e that | you know;= = /jUlie is of/ /cOurse | Fand making / {Up n} /bEAuti / .hh u[hm:: [uh_huh– now she had a job (...) e

1 2 3 4 5 6 7 8 9 10 11 12 13 14 f f f 15 16 17

Ben: Ava:

The first piece of news that Ava delivers in (7) concerns her meeting up with their former friend and this friend’s partner, who she introduces with the recognitionals Leo and Julie (lines 2 and 3). A second piece of news follows, namely that the two are doing well (line 5). With some delay, Ben now proffers a piece of candidate knowledge about Leo for confirmation (line 8). But Ava pursues her telling by announcing that they have moved to a new house in Westchester (lines 7 and 9), before confirming Ben’s proffer (line 10). Next, she extends her telling with the detail that their new house is an older one, which Julie is carving up and making beautiful (lines 11, 13–14).

282

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

Observe now that this linear sequence of VP1 and VP2 is produced by one speaker in the construction of a simple turn-constructional unit; that is, it has togetherness of action. Second, the sequence is delivered with a good measure of prosodic/phonetic togetherness. Although we find some final lengthening on U:p and a small pitch step-up on and, the encliticization of and and its phonetic reduction speak more for prosodic integration. There is also a certain amount of syntactic fusion between the two parts: Ava’s VP2 (making beautiful) not only shares a grammatical subject with VP1 (carving up) but also a grammatical object36, which, in addition, appears in front of the two VPs.37 In fact, it could even be argued that the VPs display some semantic fusion in this context: for Ava, Julie’s remodeling efforts appear to have a single purpose and result – beautification. Conceptually, we are thus not dealing with the temporal sequence or simultaneity of two separate events but rather with one event. The VPs in (7) are thus positioned and realized in a manner which, with further semantic bleaching of carve up, could conceivably yield a productive hendiadic construction. Admittedly, this has not yet happened: we do not have a construction carve up and X similar to other hendiadics such as, e.g., go ahead and X, sit down and X, etc. However, a realization such as that in (7), were it to recur frequently, would facilitate such a development. In this sense, the criterion of prosodic/phonetic integration can open our eyes to potentially “new” verbs which are susceptible to being absorbed by the hendiadic pattern. More generally, prosodic/phonetic integration may help us identify candidates for emergent constructionhood. This said, two caveats are in order: First, we clearly cannot expect every single instance of VP conjunction with prosodic/phonetic togetherness to instantiate an entrenched hendiadic pattern. As has been argued above, the varying degree of prosodic/phonetic togetherness in the instances of VP conjunction discussed can be (a) the effect of a first, but fluent, occurrence of two conjoined VPs whose specific lexical filling may, or may not, be repeated and become entrenched more deeply; or (b) the effect of (frequent) repetition of tokens of a conjoined VP pattern with specific lexical filling which has led to a neuromotor routine and to the pattern being stored as a unit. Second, we cannot expect that clearly hendiadic constructions with and will always be produced with prosodic and phonetic togetherness. Any

36 37

A form of referential cohesion as described by Givón (1993, vol. II: 16). Cf. violation of the ‘island constraint’, Stefanowitsch (1999: 124).

Action, prosody and emergent constructions: The case of and

283

number of contingencies can arise in conversational exchanges which may result in: – a speaker splitting up a hendiadic construction into two separate prosodic contours (an action pattern similar to (III)), as illustrated in example (8). (8) Conjoined hendiadic VPs lacking prosodic integration Boyfriends (NB 008: 922) (Following an extended troubles telling about Nancy’s problems with her exhusband, Emma is now inquiring about her friend’s social life.) 8 Emm: You got any(b) frie:nd boyfrie:nds? er any°thing 9 [goin:g [steady’r:°] 10 Nan: [Oh::: [°Gh*ell n ]*o.°? 11 Emm: °Nothin° 12 (0.3) 13 Nan: .t Oh I’ve gotta lot of (0.2) frie:nds,= 14 Emm: =But n[othin’ you’re] dating. 15 Nan: [But n o: ] 16 Nan: .hhhh Oh hu-E:mma GI: don’t wanna get? = 17 Emm: =GY[eh ((compressed)) 18 Nan: I [just am: not emotionally: 19 (0.2) 20 Emm: °Mm-mm[:.° 21 f Nan: [I: don’t wanna get invo:lved I: don’t wanna go 22 f ou:t and .hhhhhhhhh Fput myself in the position where: some 23 ma:n’s gonna think oh boy she’s really hard up you know, hh 24 Emm: Mm [ m m : ] 25 Nan: [I mean y’nu I’ve h]ea:rd this so: mu:ch.h[.hhhhh 26 Emm: [°Mm: [hm :° 27 Nan: [That 28 this see:ms to be the na:tural (.) you know 29 Emm: (.) 30 Nan: re[action,]

In fragment (8), Nancy makes several attempts to account for why she is not dating anyone at the moment (lines 16, 18, 21–23). The last one of these involves a VP1 and VP2 pattern: I don’t wanna go out and put myself in the position … (line 21f). Semantically, much speaks for interpreting this pattern in a oneevent fashion: for one, the verb go out as used here has little semantic content. What Nancy is saying is that after a period of romantic abstinence she does not want to put herself in an uncomfortable position by beginning to see men again.38 Yet in terms of prosodic/phonetic togetherness, we can observe a range of disintegrative features: final lengthening on go out, a lengthy 38

Cf. the description of ‘Go-out-and-V’ in Stefanowitsch (2000). Were go out being used in the sense of ‘date’, we would expect Nancy to say go out with someone.

284

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

inbreath after and, a marked pitch step-up on put and creaky voice on only part of the construction, namely on and. In fact, it may be precisely the disintegrated delivery of this hendiadic construction which indexes the emotional turmoil Nancy is describing. – a speaker using the second part of a hendiadic as a means of jointly completing another speaker’s provisionally incomplete turn, as illustrated in (9). (9) Conjoined hendiadic VPs in co-construction Bring students (CallHome, 5648: 406) (Ida has asked her friend Maria, a nun currently working in Chile, how her family feels about her being away from home.) 1 MAR: uhhh:: it’s so SO. 2 it’s INteresting; 3 cuz most of the family was really exCITed? 4 and in FACT; 5 gladys and mich would like to come down at some point, 6 IDA: o[h good. 7 MAR: [nita teaches spanish up in alaska? 8 IDA: oh ter[rifi[c. 9 f MAR: [°hh [shE would like to cOm:e – 10 f (0.18) 2e[n wOrk-h (in a sh-/you ] know she’[s) ] 11 f IDA: [en brIng her (STUdents);] [rIght?] 12 [((lau[ghs)) 13 MAR: [°hh [((laughs)) 14 to do (2 sylls)15 =and then (Also) uhm (0.1) u:h:m: – 16 cIndy; 17 uh: nEd’s wife; 18 °hh uhm is a m-spAnish profEssor; 19 =and so shE would like to come dOwn –

In response to Ida’s inquiry, Maria explains that most of her family are excited about her being in Chile (line 3). By way of exemplification, she now mentions that Gladys and Mich, presumably relatives, intend to come down (line 5). Next she mentions Nita, described as a Spanish teacher in Alaska (line 7). This detail, proffered with rising intonation in the form of a trymarker (Sacks and Schegloff 1979), is an attempt to achieve recognition, but Ida treats it as a piece of news to be assessed (line 8). Maria now adds that she too would like to cOm:e – projecting with non-final intonation that her turn will continue (line 9). But when she does not carry on immediately, Ida uses the opportunity to co-complete the turn with and bring her students (line 11). This co-completion proves in retrospect to be the VP2 of a hendiadic pattern: come and bring her students. Interestingly, Maria herself was also working on a

Action, prosody and emergent constructions: The case of and

285

similar construction: [come] and work-h, which, however, she does not bring to completion (line 10). So in (9) it is the recipient who co-completes a hendiadic construction which was incipient in the interlocutor’s turn.39 To sum up, we have argued here that prosodic/phonetic togetherness may frequently be found with potential one-event patterns and can even cue emergent hendiadic constructions. Yet, it cannot be the only diagnostic. Prosodic delivery – and therefore also phonetic realization – both of hendiadic constructions and of constructions in general, is always a contingent accomplishment in real-time interaction.

5

Conclusion

This paper has presented the results of a study of cohesiveness between the conjoined parts of VP conjunction with and. We have shown that there are different layers of use of and as a VP-conjunctional, ranging from systematic cross-speaker uses to single-speaker hendiadic constructions. Inspired by Bybee’s Linear Fusion Hypothesis (2002), we have inquired into the specific pre-requisites for fusion apart from formal adjacency and have pointed to the fact that only those VP1 and VP2 patterns are susceptible of fusion which exhibit togetherness of action and togetherness in prosodic/phonetic form. In our argument, we have extended Bybee’s fusion hypothesis from the level of the phonological word to items whose constituents can be realized as separate predicates in separate prosodic units and/or in separate turns and turn-constructional units. We have also amended Bybee’s more syntactically/semantically oriented notion of being “used together” to include the notions of togetherness of action and togetherness in prosodic/phonetic form. Togetherness of action refers to a situation in which the items in question are used as part of a simple turn-constructional unit and contribute to implementing a single action. Whereas there are regular patterns of VP conjunction in which the potential pieces occur on their own, VP1 and VP2 need to co-occur in a simple turn-constructional units and be used to accomplish a single action in order to fuse together. Furthermore, they need to meet the criterion of togetherness in prosodic/phonetic form. This refers to the fact that they must be produced in one intonation unit and undergo phonetic integration. Both togetherness in action and togetherness in prosodic/phonetic form are pre-conditions for the emergence of hendiadys. In addition, we have identified prosodic togetherness as the more basic pre-condition, enabling phonetic togetherness to come into play, and have 39

Cf. Lerner (1996) for other cases of anticipatory TCU completion.

286

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

highlighted phonetic processes, in particular fast-speech rules, not only as resulting from frequent use and storage as a unit, but also as contributing to the sedimentation of a unit in a particular form. Needless to say, however, all three types of togetherness – action, syntax/semantics and prosody/phonetics – must be considered aspects of one and the same process of emergence. Methodologically, we have stressed that insight into the functioning of linguistic items can be gained from looking beyond a single speaker’s contribution. Absent this perspective, the regular occurrence of, e.g., crossspeaker uses of VP conjunction with and, as well as comparable uses of other conjunctionals, would remain undetected. We have furthermore stressed that insight into the emergence of constructions can be gained from the study of fine prosodic and phonetic detail. Here it may be necessary to examine single prosodic and phonetic features separately in order to do justice to their characteristic, often gradient nature. Absent this perspective, degrees of lengthening, extent of pitch movement, shades of vocalic quality, etc. and their relevance for emergent constructionhood escape notice. Finally, we have argued that prosodic delivery and phonetic realization are accomplishments which are sensitive to the contingencies of talk-in-interaction. Even if in the aggregate they play an important role in the recognition of emergent constructional patterns, they cannot be a blanket diagnostic for individual cases. Apart from this, our observations on the fine detail of prosodic/phonetic realization at potential unit boundaries and their relevance to developing a model for the emergence of constructions have repercussions for a different, but related area of language study, namely current practices of transcription. To our knowledge, none of the more widely used transcription systems to date (GAT, Jefferson, du Bois) provide for the systematic notation of weak prosodic boundaries: they either more or less neglect the question (Jefferson)40 or force the transcriber to make a binary yes/no decision (GAT, du Bois)41. When confronted with weak boundaries, transcribers therefore usually fall back on (orthodox) syntactic criteria. The problems created by this practice are obvious: it does not adequately reflect the prosodic-phonetic features of the material at hand; and more importantly in the context of emergent constructions, it veils a phenomenon which may provide us with crucial evidence for identifying constructions in the making.

40 41

Cf. Schenkein (1978), Atkinson and Heritage (1984). Cf. Selting et al. (1998), du Bois et al. (1993).

Action, prosody and emergent constructions: The case of and

287

Appendix Selected GAT 2 conventions (cf. Selting et al., 2009, also Selting et al., 1998, to appear) [ [

] ]

= (.) (-), (--), (---)

(2.0) (2.85) .h, .hh, .hhh h, hh, hhh anduh, and_uh :, ::, ::: uh, ah * \ so(h)o hm, yes, yeah, no hm_hm, yea_ah ACcent Accent ? , – ; .

F or FF G or GG

((coughs))

(

)

overlap (beginning and end) latching of units micropause short, middle, long pauses of approximately 0.25–0.75 seconds, up to approximately one second estimated pause of more than one second measured pause in-breath, according to its duration out-breath, according to its duration cliticization lengthening, according to its duration hesitation signals, filled pauses glottal closure break-off without glottal closure laughter particles integrated into speech one-syllable continuers two-syllable continuers primary, or main accent secondary accent unit-final intonation, rising to high unit-final intonation, rising to mid unit-final intonation, level unit-final intonation, falling to mid unit-final intonation, falling to low pitch up-step (smaller or larger) pitch down-step (smaller or larger) allegro, fast lento, slow creaky voice paralinguistic and non-linguistic actions and events paralinguistic and non-linguistic actions and events accompanying speech unintelligible passage

288

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

unintelligible passage with indication of the number of syllables it contains (‘xxx’ represents one syllable) presumed wording presumed sound or syllable alternative presumed wordings unidentifiable speaker line in transcript referred to in the text parallel interaction = ˜ one intonation phrase

(xxx xxx …) (such) al(s)o (such/which) X1:

f @ one line

Additional phonetic details: | weak intonation unit boundary subscript/superscript symbols (e.g.F or : or { }

)

e

feature present to a smaller degree articulatory detail in IPA

References Atkinson, J. M. & Heritage, J. (eds.) 1984 Structures of Social Action, Cambridge. Barth-Weingarten, D. 2007 Prosody, construction grammar and language change. In: S. Volk-Birke & J. Lippert (eds.), Anglistentag 2006 Halle, Trier, 421–433. 2010 On the role of phonetics in signaling discourse functions in spoken English: The many shades of and. In: Proceedings Anglistentag 2009 Klagenfurt, Trier, 21–34. (in prep.) Intonation units, turn-taking an language change. A parametricized approach to cesuring in talk-in-interaction. Unpubl. Postdoc Thesis. (under revision) Of ens ’n’ ands – observations on the phonetic make-up of a coordinator and its uses in talk-in-interaction. To appear in Language & Speech (special issue ed. by Richard Odgen). (submitted) The fuzziness of intonation units: Some theoretical considerations and a practical solution. To appear in InLiSt (http://www.inlist.uni-bayreuth. de). Barth-Weingarten, D. & E. Couper-Kuhlen 2002 On the development of final though: A case of grammaticalization? In: G. Diewald & I. Wischer (eds.), New Perspectives on Grammaticalization, Amsterdam, 345–361. Biber, D., S. Johansson, G. Leech, S. Conrad & E. Finegan 1999 Longman Grammar of Spoken and Written English, Essex. Birkner, K. 2008 Relativ(satz)konstruktionen im gesprochenen Deutsch. Syntaktische, prosodische, semantische und pragmatische Aspekte, Berlin/New York.

Action, prosody and emergent constructions: The case of and

289

Blöndal, T. 2006 Dinner or coffee or …? The interactive role of turn-final ‘eja’ (e. ‘or’) in Icelandic. Presentation at the International Conference on Conversation Analysis (ICCA) 2006 Comparative perspectives in Conversation Analysis, 10–14 May 2006, Helsinki. Bybee, J. 2001 Phonology and Language Use, Cambridge. 2002 Sequentiality as the basis of constituent structure. In: T. Givón & B. F. Malle (eds.), The Evolution of Language out of Pre-language, Amsterdam, 109–134. 2006 From usage to grammar: The mind’s response to repetition. Language 82, 711–733. Bybee, J. & J. Scheibman 1999 The effect of usage on degrees of constituency: the reduction of don’t in English. Linguistics 37, 575–596. Bybee, J. & S.A. Thompson 2000 Three frequency effects in syntax. Berkeley Linguistic Society 23, 378–388. Couper-Kuhlen, E. 1996 Intonation and clause combining in discourse: The case of ‘because’. Pragmatics 6, 389–426. Couper-Kuhlen, E. & T. Ono 2007 ‘Incrementing’ in conversation. A comparison of practices in English, German and Japanese. Pragmatics 17, 505–552. Croft, W. 1995 Intonation units and grammatical structure. Linguistics 33, 839–882. Cruttenden, A. 1986 Intonation, Cambridge. Crystal, D. 1969 Prosodic Systems and Intonation in English, Cambridge. Degand, L. & A.C. Simon 2009 Mapping prosody and syntax as discourse strategies: How basic discourse units vary across genres. In: D. Barth-Weingarten, N. Dehé & A. Wichmann (eds.), Where Prosody meets Pragmatics, Bingley, 79–106. Du Bois, J. W., S. Schuetze-Coburn, D. Paolino & S. Cumming 1992 Discourse transcription. In: Santa Barbara Papers in Linguistics, 4, Department of Linguistics, University of California, Santa Barbara. 1993 Outline of discourse transcription. In: J. A. Edwards & M. D. Lampert (eds.), Talking Data: Transcription and Coding in Discourse Research, Hillsdale, New York, 45–87. Ford, C. E., B.A. Fox & S.A. Thompson 2002 Constituency and the grammar of turn increments. In: C. E. Ford, B. A. Fox, & S.A. Thompson (eds.), The Language of Turn and Sequence, New York, 14–38. Givón, T. 1991 Serial verbs and the mental reality of ‘Event’: Grammatical vs. cognitive packaging. In: B. Heine & E. Traugott (eds.), Approaches to Grammaticalization, Amsterdam, 81–127. 1993 English Grammar. A Function-Based Introduction, Vol. 1 & 2. Amsterdam/Philadelphia.

290

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

1997 Grammatical relations: an introduction. In: T. Givón (ed.), Grammatical Relations. A Functional Perspective, Amsterdam, 1–84. Haiman, J. 1994 Ritualization and the development of language. In: W. Pagliuca (ed.), Perspectives on Grammaticalization, Amsterdam/Philadelphia, 3–28. Hopper, P. 1987 Emergent grammar. Berkeley Linguistic Society 13, 139–157. 1991 On some principles of grammaticalization, Vol. 1. In: E.C. Traugott & B. Heine (eds.), Approaches to Grammaticalization, Amsterdam, 17–35. 1998 Emergent grammar. In: M. Tomasello (ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure, Mahwah, 155– 175. 2001a Grammatical constructions and their discourse origins: prototype or family resemblance?. In: M. Pütz, S. Niemeier & R.Dirven (eds.), Applied Cognitive Linguistics I: Theory and Language Acquisition, Berlin, 109–129. 2001b Hendiadys and auxiliation in English. In: J. Bybee & M. Noonan (eds.), Complex Sentences in Grammar and Discourse. Essays in Honour of Sandra A. Thompson, Amsterdam, 145–173. 2008 Emergent serialization in English: Pragmatics and typology. In: J. Good (ed.), Linguistic Universals and Language Change, Oxford, 253–284. Hopper, P. & S.A.Thompson 2008 Projectability and clause combining in interaction. In: R. Laury (ed.), Crosslinguistic Studies of Clause Combining, Amsterdam, 99–123. Jefferson, G. 1980 On a failed hypothesis: “Conjunctionals” as overlap vulnerable. Two explorations of the organization of overlapping talk in conversation. Tilburg Papers in Language and Literature 28. Jun, S. & C. Fougeron 2000 A phonological model of French intonation. In: A. Botinis (ed.), Intonation: Analysis, modelling and technology, Amsterdam, 209–242. Kaisse, E. M. 1985 Connected Speech. The Interaction of Syntax and Phonology, Orlando. Koivisto, A. 2006 Finnish että as a turn-final particle: negotiating topic-transitions. Presentation at the International Conference on Conversation Analysis (ICCA) 2006 Comparative Perspectives in Conversation Analysis, 10–14 May 2006, Helsinki. Lehmann, C. 1988 Towards a typology of clause linkage. In: Haiman, J. & S.A. Thompson (eds.), Clause Combining in Grammar and Discourse, Amsterdam/Philadelphia, 181– 225. Lerner, G. H. 1996 On the “semi-permeable” character of grammatical units in conversation: Conditional entry into the turn space of another speaker. In: E. Ochs, E.A. Schegloff & S.A. Thompson (eds.), Interaction and Grammar, Cambridge, 238–276. Mori, J. 1999 Negotiating Agreement and Disagreement in Japanese: Connective Expressions and Turn Construction, Amsterdam.

Action, prosody and emergent constructions: The case of and

291

Mulder, J. & S.A.Thompson 2008 The grammaticalization of but as a final particle in English conversation. In: R. Laury (ed.), Crosslinguistic Studies of Clause Combining: The Multifunctionality of Conjunctions, Amsterdam, 179–204. Pullum, G. K. 1990 Constraints on intransitive quasi-serial verb constructions in modern colloquiual English. Working Papers in Linguistics 39, 218–239. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik 1985 A Comprehensive Grammar of the English Language, Cambridge. Raymond, G. 2004 Prompting action: The stand-alone “so” in ordinary conversation”. Research on Language and Social Interaction 37, 185–218. Sacks, H. & E. A. Schegloff 1979 Two preferences in the organization of reference to persons in conversation and their interaction. In: G. Psathas (ed.), Everyday Language. Studies in Ethnomethodology, New York, 15–21. Sánchez, I. 1999 English pseudo-coordination. Paper presented at the 2nd Annual High Desert Student Conference in Linguistics. Albuquerque, NM. Schegloff, E. A. 1996 Turn organization: one intersection of grammar and interaction. In: E. Ochs, E. A. Schegloff & S.A. Thompson (eds.), Interaction and Grammar, Cambridge, 52–133. Scheibman, J. 2000 I dunno: A usage-based account of the phonological reduction of don’t in American English conversation. Journal of Pragmatics 32, 105–124. Schenkein, J. 1978 Studies in the Organization of Conversational Interaction, New York. Schönherr, B. 1997 Syntax – Prosodie – nonverbale Kommunikation. Empirische Untersuchungen zur Interaktion sprachlicher und parasprachlicher Ausdrucksmittel im Gespräch, Tübingen. Selting, M., P. Auer, B. Barden, J. Bergmann, E. Couper-Kuhlen, S. Günthner, C. Meier, U. Quasthoff, P. Schlobinski & S. Uhmann 1998 Gesprächsanalytisches Transkriptionssystem (GAT). Linguistische Berichte 173, 91–122. Selting, M., P. Auer, D. Barth-Weingarten, J. Bergmann, P. Bergmann, K. Birkner, E. Couper-Kuhlen, A. Deppermann, P. Gilles, S. Günthner, M. Hartung, F. Kern, C. Mertzlufft, C. Meyer, M. Morek, F. Oberzaucher, J. Peters, U. Quasthoff, R. Schmitt, W. Schütte, A. Stukenbrock & S. Uhmann 2009 Gesprächsanalytisches Transkriptionssystem 2 (GAT 2). Gesprächsforschung Online 10 (http://www.gespraechsforschung-ozs.de/heft2009/heft2009. htm). (to appear) A system for transcribing talk in interaction: GAT 2 (translated and adapted for English by Elizabeth Couper-Kuhlen and Dagmar Barth-Weingarten). Gesprächsforschung – Online (http://www.gespraechsforschungozs.de). Shockey, L. 2003 Sound Patterns of Spoken English, Oxford.

292

Dagmar Barth-Weingarten and Elizabeth Couper-Kuhlen

Steedman, M. 2000 The Syntactic Process, Cambridge/Mass. Stefanowitsch, A. 1999 The GO-AND-VERB-construction in a cross-linguistic perspective: Imageschema blending and the construal of events. In: D. Nordquist & C. Berkenfield (eds.), Proceedings of the Second Annual High Desert Linguistics Society Conference (http://www.stefanowitsch.de/anatol/research. html, 24. 02. 2006). 2000 The English GO-(PRT)-AND-VERB construction. Berkeley Linguistic Society 26, 259–270. Walker, G. 2004 On some interactional and phonetic properties of increments to turns in talk-in interaction. In: E. Couper-Kuhlen & C. E. Ford (eds.), Sound Patterns in Interaction, Amsterdam, 147–169. Wichmann, A. 2005 Prosody and discourse: a diachronic approach. In: Proceedings IDP05 “Interface Discours-Prosodie (Discourse-Prosody interface)”, 8–9 September 2005, Aix-enProvence, France 2005 (http://aune.lpl.univ-aix.fr/~prodige/idp05/actes/ wichmann. pdf; date of access: 21. 12. 2007). Wichmann A., A.-M Simon-Vandenbergen & K. Aijmer 2009 How prosody reflects semantic change: a synchronic case study of of course. In: H. Cuyckens, K. Davidse & L. Vandelanotte (eds.), Subjectification, Intersubjectification and Grammaticalization, Berlin, 103–154.

293

On the emergence of Hebrew adverbial connectives

Yael Maschler *

On the emergence of adverbial connectives from Hebrew relative clause constructions with participation of Susan Shaer

1

Introduction

In this study, we wish to explore a particular emergent construction (Hopper 1987, 1988, 1998, 2001, 2004, this volume) in Hebrew discourse, one that started out as a relative clause construction and has come to function as an adverbial connective. A relative clause consists of a clause modifying a nominal kernel – i.e., noun, noun phrase, or pronoun – known as ‘the antecedent’ or ‘head’. Examine, for instance, Excerpt 1, a conversation between three young Israelis, a couple (Ron and Goni), and Goni’s cousin, Idit. The three are reminiscing about their childhood pets: Excerpt 1 (‘Pets’)1: 34 Ron:

35 Goni:

36 Ron:

.. ve--nidme li shegam hayu lahem 'e--h, and seems to me that also there were to them uh, ‘a--nd it seems to me they also had u--h,’ ... hayu lahem there were to them ‘they also had goldfish.’

gam also

degey zahav. fish gold

tuki. ‘[a] parrot.’

* The paper was written by Yael Maschler, who further developed some of the findings of an M. A. thesis written under her direction by Susan Shaer. The work was first presented at the FRIAS workshop on Emergent Constructions, May 9, 2008, and then at the annual meeting of the Israeli Linguistic Society, at the Haifa Language Forum, and at the colloquium of the Linguistics department at the University of Caliafornia at Santa Barbara. I thank the participants at those forums for their questions and comments, particularly: Mira Ariel, Peter Auer, John Du Bois, Carol Genetti, Paul Hopper, Irit Meir, Adina Moshavi, Bracha Nir, Ora Schwarzwald, Sandra Thompson, Paul Woodward, and Tamar Zewi. 1 For transcription conventions and list of abbreviations, see appendix. The square brackets in the transliteration lines mark the head of the relative clause.

294

Yael Maschler

37

... hayu lahem there were to them ‘they had fish.’

38 Idit:

... vehaya lahem 'ogrim, and there was to them hamsters ‘and they had hamsters,’

39

.. midey pa'am, ‘occasionally,’

40

lo? ‘no?’

41 Ron:

dagim. fish

naxon. ‘right.’

42

.. [kavyot]. ‘guinea pigs.’

43

.. (hand clap)

44

... she'atem natatem lahem. that you gave to them {you = Goni’s family} ‘that you had given them.’

The clause she'atem natatem lahem (‘that you had given them’, line 44) modifies the noun kavyot (‘guinea pigs’, line 43). Another example is found a few seconds later on in the same conversation, still discussing the guinea pigs: 67 Goni:

.. vehexziru ‘and they returned

lanu, to us,’

68

.. 'et ma shenish'ar. what that was left ‘those that were left.’

69

...

70

.. lakaxnu we took, ‘we took,’

71

le[vet hasefer to house [of] the book ‘to the open school,’

72

... sheyaron lamad bo? that Yaron studied at it (3P MASC SG PREP) ‘that Yaron studied at (where Yaron studied)?’

ve'et ma shenish'ar, and what that was left ‘and those that were left,’

hapatuax], the open

On the emergence of Hebrew adverbial connectives

295

In lines 69–71, we find the main clause ve'et ma shenish'ar, lakaxnu, levet hasefer hapatuax (‘and those that were left, we took, to the open school’). The noun phrase consisting of definite nominal compound2 + adjective bet hasefer hapatuax (‘the open school’), which functions as adverbial complement to the verb lakaxnu (‘we took’), is modified by the relative clause sheyaron lamad bo (‘that Yaron studied at’ (i.e., ‘where Yaron studied’)). ‘The open school’, then, is the head, or antecedent of this relative clause. The relative clause is a construction much studied by scholars of Hebrew in its various periods (e.g., Peretz 1967, Landau 1975, Blau 1977, Ornan 1979, Schwarzchild 1990, Waltke and O’Connor 1990, Goldenberg 1995, Azar 1995, Fassberg 1996, Tsarfati 1996, Shlezinger 1999, Huehnergard 2006). With the exception of Ariel 1999 and Shaer 2007, it has not heretofore been studied in spoken Hebrew discourse. In Modern Hebrew, according to prescriptive Hebrew grammar, a relative clause must be connected to its antecedent via two means: 1) a relativizer, which in the present corpus of spoken Hebrew is always the proclitic complementizer she- (‘that’) attached to the first element of the relative clause3. This functional morpheme is known as the general Hebrew ‘subordinator’, that which connects also complement clauses to complement-taking-predicate clauses (e.g., nidme li she- (‘it seems to me that’, e.g., Excerpt 1, line 34), 'ani xoshevet she- (‘I think that’)). The form she- also connects the various types of adverbial clause to their main clause, constituting part of the connective, such as lifney she-, 'axarey she- (the temporal connectives ‘before’ and ‘after’), mipney she-, (the causal connective ‘because’), or lamrot she- (the concessive connective ‘although’). 2) The second means connecting a Hebrew relative clause to its antecedent is an element in the relative clause which is coreferential with the antecedent and agrees with it in person, gender and number. Hebrew prepositions can be declined for person, gender and number. In Excerpt 1, line 72, e.g., the coreferent is the declined preposition bo, consisting of the preposition b(e)- (‘at’) declined in third person masculine singular (to agree with the antecedent ‘the open school’), and therefore taking the suffix -o, creating the resumptive pronoun bo (‘at it’). This coreferent functions as adverbial complement in the relative clause. When the coreferential element functions 2

3

The Hebrew word for ‘school’ consists of the nominal compound (smixut construction) bet (‘house’) sefer (‘book’). The noun bet undergoes phonological change to vet when it immediately follows the preposition le- (‘to’). The relativizer can be 'asher or ha- in written Modern Hebrew discourse, particularly of higher registers. In this type of discourse we also find asyndetic relative clauses, i.e., with no relativizer at all.

296

Yael Maschler

as subject of the relative clause, it is the subject pronoun of the relative clause which constitutes the coreferent. Since Hebrew is a so-called ‘Pro-drop’ language, the subject coreferent is often reduced to an obligatory bound subject morpheme on the verb. According to prescriptive Hebrew grammar (e.g., Tsadka 1989: 106), a clause is considered a relative clause only if the coreferential element either appears or could be added. For instance, in lines 42–44, kavyot. she'atem natatem lahem. (‘guinea pigs. that you had given them.’), we find the relativizer she- but no element which is coreferential with the antecedent kavyot (‘guinea pigs’). Such an element would be the declined direct object marker ’otan, consisting of the direct object marker 'et declined in third person, feminine, plural, to agree with kavyot (‘guinea pigs’). Because this resumptive pronoun could be added here, this construction is considered a relative clause according to prescriptive grammar. Prescriptive Hebrew grammar allows this deletion of the resumptive pronoun only when it denotes a direct object in the relative clause. A common phenomenon lamented by prescriptive Hebrew grammarians is the lack of a resumptive pronoun in the case of verbal complements other than the direct object. This lack of resumptive pronoun is by no means restricted to spoken discourse. It can often be encountered reading the newspaper, as in the following example form Ha'aretz, known as ‘the intellectuals’ newpaper’: gam be[mikrim] shehaxug ’eyno nisgar also in cases that the department does not close down ‘Even in the event that the department does not close down’ holxim vene'elamim txumey mexkar mesuyamim go and disappear fields research certain ‘certain fields of research are disappearing’ ketotsa'a mehakitsuts bemispar xavrey as a result of the cut in number members ‘as a result of cuts in the number of faculty members. (Ha’aretz, Jan 4, 2008).’

hasegel the faculty

The relative clause haxug 'eyno nisgar (‘the department does not close down’) modifying the head mikrim (‘cases, events’) appears with no resumptive pronoun, which is obligatory here according to prescriptive grammar, since the correferent functions as an adverbial complement in the relative clause (‘the department does not close down in these cases’). The ‘correct’ form should have been: nisgar bahem close down in them (3P MASC PL PREP) ‘Even in cases in which the department does not close down’ gam also

bemikrim in cases

shehaxug 'eyno that the department does not

On the emergence of Hebrew adverbial connectives

297

or in different word order: 4 gam bemikrim shebahem haxug also in cases that in them (3P MASC PL PREP) the department 'eyno nisgar does not close down ‘Even in cases in which the department does not close down’ ‘Even in the event that4 the department does not close down’

The lack of a resumptive pronoun in Hebrew has been studied within the generative framework by Doron (1982), Borer (1984), and Erteschik-Shir (1992). Givón has studied it based on intuitions regarding Hebrew examples (1973, 1975), and Ariel (1999) has investigated the phenomenon in written and spoken discourse from the standpoint of accessibility theory, arguing that when the head is highly accessible when the relativized position is processed, we find a Ø resumptive pronoun. Here we take an interactional linguistics approach, which will show some problems with previous approaches. After a brief description of the data, in Section 3 we describe the extent of the phenomenon in a corpus of spoken Hebrew discourse. In Section 4, we suggest an explanation for it in terms of the functions of relative clauses in discourse as they relate to the typology of Hebrew. In connection with the topic of this volume, in Section 5 we then show that in the most common environment in which the phenomenon occurs, the lack of resumptive pronoun results in grammaticization involving the emergence of adverbial connectives from relative clause constructions in Hebrew. We conclude in Section 6 with a discussion of this case of grammaticization.

2

Data

The study is based on a corpus of 55 naturally-occurring casual Hebrew interactions among students, their friends and relatives, collected during the years 1994–2002. The conversations range from approximately one to nine minutes in length. Altogether 158 minutes of talk among 141 different participants were investigated, two to five participants per interaction. All cases following the pattern

4

Note that this English ‘equivalent’ also omits the element denoting the adverbial complement of the relative clause (the preposition in + case marking on the relativizer). In English the relativizer itself is often omitted (see below), an option not available in spoken Hebrew discourse.

298

Yael Maschler

[antecedent + she- + relative clause] (where ‘clause’ includes at least subject and predicate, but the antecedent need not be part of a ‘main’ clause) were collected. Altogether, 115 relative clauses satisfying this pattern were found in the corpus. Note that in lines 67–68 of Excerpt 1 67 Goni:

.. vehexziru ‘and they returned

lanu, to us,’

68

.. 'et ma shenish'ar. DIR OBJ MRK what that was left ‘those that were left.’

the clause ma shenish'ar (‘what that was left’), composed of the question word ma (‘what’), the relativizer she- (‘that’) and the clause nish'ar (‘was left’) (‘Prodropped’ subject + verb), does not modify any head in the main clause. The entire clause functions as direct object of the main verb hexziru (‘they returned’), to which it is connected via the direct object marker 'et. Such clauses are considered independent relative clauses since they do not modify any head in the main clause (Waltke and O’Connor 1990, Zewi 2008). There were 63 such independent relative clauses throughout the database. The present database includes only dependent relative clauses. Thus, cases opening with question words (e.g., mi she- ‘whoever’ lit. ‘who that’, ma she- ‘whatever’, lit. ‘what that’) were not considered. Note that the relative clause at line 44 of Excerpt 1 appears as an increment (Schegloff 1996), following a sentence-final intonation contour kavyot. (‘guinea pigs.’) and is further separated from the antecedent by a clap. Of the 115 relative clauses found throughout the database, only five (four percent) had their antecedent separated from the relative clause by sentence-final intonation.

3

Lack of resumptive pronoun in spoken Hebrew discourse

In this section, we explore the discourse processes leading to emergence of Hebrew adverbial connectives from relative clause constructions. First, we describe the extent and environments in which the resumptive pronoun is lacking.

299

On the emergence of Hebrew adverbial connectives

3.1 The syntactic role of the coreferent in the relative clause All 115 relative clauses were classified according to the syntactic role which the coreferent occupies in the relative clause: Table 1: Distribution of relative clauses according to syntactic role of the coreferent Syntactic role of coreferent in relative clause

N (%)

Subject

68 (59 %)

Direct Object

19 (17 %)

Indirect Object

12 (10 %)

Time Adverbial

12 (10 %)

Place Adverbial

3 (3 %)

Manner Adverbial

1 (1 %)

Total

115 (100 %)

The most common syntactic role of the coreferent is subject (59 %). Examine, for instance, Excerpt 2: Excerpt 2 (‘Violence’): 3 Tom:

.. bekitsur, ‘anyway,’

4

.. yesh ‘there’s

5

.. she'over meharakevet, ‘that passes (3P MASC SG) from the train,’

6

.. lekivun ‘to the direction of

[gesher], [a] bridge,’

ha'itstadyon. the stadium.’

The coreferent of the antecedent gesher (‘bridge’) consists of the third person masculine singular bound subject morpheme of the predicate 'over (‘passes’), which in Hebrew consists of the ø morpheme. This morpheme functions as subject of the relative clause. From Table 1, we see that there are more direct object coreferents (17 %) in comparison to indirect object ones (10 %), and that object coreferents (both direct and indirect, 27 %) are almost twice as common as adverbial complement coreferents (12 % + 3 % + 1 % = 14 %). However, the number of indirect object coreferents is identical to that of time adverbials (10 % each).

300

Yael Maschler

Keenan and Comrie (1977) have argued that languages impose restrictions on the grammatical role of the NP or Ø coreferential with the head of a relative clause. They propose an accessibility hierarchy to account for relativizing possibilities in the world’s languages: Subj > Direct Obj > Indirect Obj > Oblique > Genitive > Obj of Comparison The higher on this scale a grammatical category is, according to Keenan and Comrie, the more often it can be relativized. Note that if we take Keenan and Cormier’s ‘Oblique’ as refering to adverbial complements, Table 1 indeed confirms this hierarchy for the Hebrew data, as the frequency of relativized grammatical roles is represented by the following scale: Subj > Direct Obj > Indirect Obj > Adv Compl, with the exception that indirect objects are relativized as frequently as time adverbials. 3.2 The part of speech constituting the coreferent Next, all relative clauses were classified according to the part of speech constituting the coreferent: Table 2: Distribution of relative clauses according to part of speech of the coreferent Part of speech constituting the coreferent

N (%)

Bound morpheme on predicate

52 (45 %)

Ø resumptive pronoun

36 (31 %)

Subject pronoun

16 (14 %)

Resumptive pronoun

11 (10 %)

Total

115 (100 %)

We see that almost half the coreferents (45 %) are realized as a bound morpheme on the predicate, as in Excerpt 2 above. The second largest category is that of the Ø resumptive pronoun (31 %) as in kavyot. she'atem natatem lahem. (‘guinea pigs. that you had given them.’) of Excerpt 1, lines 42–44. A subject pronoun coreferent was found in 14 % of the cases. The smallest category consisted of the resumptive pronoun coreferents, as in bet hasefer hapatuax sheyaron lamad bo (‘the open school that Yaron studied at it’) of Excerpt 1, line 72. They constituted only 10 % of the cases. Thus, the lack of a resumptive pronoun is about three times more common than a resumptive pronoun in this database.

301

On the emergence of Hebrew adverbial connectives

3.3 The part of speech realizing the coreferent as a function of its syntactic role Let us now investigate the part of speech realizing the coreferent as a function of its syntactic role in the relative clause: Table 3: Part of speech of coreferent as a function of its syntactic role in relative clause Subject Bound morpheme on predicate Ø resumptive pronoun Subject pronoun Resumptive pronoun Total

Direct Object

Indirect Object

Adverbial Complement

52 (76 %) 17 (89 %)

4 (33 %)

15 (94 %)

2 (11 %)

8 (67 %)

1 (6 %)

19 (100 %)

12 (100 %)

16 (100 %)

16 (24 %)

68 (100 %)

3.3.1 Subject coreferents We see that for subject coreferents, 76 % consist of a bound morpheme on the predicate, whereas 24 % appear as subject pronouns, as, e.g., in the following excerpt, in which Hanna tells a story of how she was wounded during her army service at the Lebanese border. In the orientation (Labov 1972) to the narrative, she describes what she and her women soldier companions were carrying – heavy equipment which caused her to stumble and “be wounded”: Excerpt 3 (‘Wounded in Lebanon’): 91 Hanna: .. hem ‘they

'e--h u--h

sàmu 'aleynu put on us

. . . 103

.... xagór, ‘an ammunition belt,’

shaxpáts. [a] shaxpats {a type of bullet-proof vest}.’

302

Yael Maschler

104

... [shtéy meymiyòt], ‘two canteens (FEM PL),’

105

.. shehem tsrixót COMPL+ 3P MASC need PL SUBJ PRO (FEM PL) that they need ‘that need to be full’

106

.. kòl ’axat, ‘each one,’

107 Galia: 108 Hanna:

lihiyot to be (INF) to be

mele'ót, full (FEM PL) full,

mhm. .. ve--kaplá--d! ‘and [a] kaplad {fused form for ‘steel helmet’}!’

A list of seven items which the women soldiers were supposed to carry is co-constructed in lines 91–108. The item before last, shtéy meymiyòt (‘two canteens’, line 104), is modified by the relative clause shehem tsrixót lihiyot mele'ót, kòl 'axat, (‘that need to be full, each one’). The coreferent of ‘the two canteens’ functions as subject of this relative clause. It is realized by the subject pronoun hem (‘they’, MASC5) (in addition to the obligatory subject morpheme on the predicate tsrixót (‘need’, FEM PL). The fact that approximately a fourth of all subject coreferents appear as a subject pronoun contradicts Ariel, who argues that such subject pronouns are quite rare even in conversational Hebrew data (1999: 220, 257). Ariel claims this in support of Keenan and Comrie’s accessibility hierarchy (1977) (see 3.1 above), according to which if a language has resumptive pronouns6, they will be found for NPs low on the accessibility hierarchy (to aid in their processing). Furthermore, “once a language imposes [resumptive pronouns] on some position, it invariably imposes [them] on positions lower on the hierarchy” (Ariel 1999: 220). As an example, Ariel gives Hebrew, in which “subject resumptive pronouns (at the top of the accessibility hierarchy) are banned” (ibid.). Keenan and Comrie claim that an overwhelming majority of languages they surveyed (92.3 %) ban resumptive pronouns in subject position of the relative clause. However, Table 3 shows that our data does not support their theory, since a considerable number of subject coreferents (24 %) 5

6

As is very often the case in spoken Hebrew, gender agreement is not maintained in this environment and we find the subject pronoun hem (‘they’ MASC) rather than hen (‘they’ FEM), which would be the member of the paradigm agreeing with meymiyot (‘canteens’,FEM PL). Keenan and Comrie 1977, as well as Ariel 1999, include in the category ‘resumptive pronoun’ also subject pronouns.

On the emergence of Hebrew adverbial connectives

303

do appear in the form of a subject pronoun (i.e., Keenan and Comrie’s ‘resumptive pronoun’) rather than as a bound subject morpheme on the predicate. 3.3.2 Direct object coreferents For direct object coreferents, 89 % appear with no resumptive pronoun. Only 2 cases throughout the corpus (11 %) exhibit a direct object resumptive pronoun. Thus, although according to prescriptive Hebrew grammar, direct object coreferents can be realized as either a resumptive pronoun or a Ø coreferent, naturally-occurring conversational data show that the latter option is highly preferred. 3.3.3 Indirect object coreferents and adverbial complement coreferents For indirect object coreferents, 33 % appear with no resumptive pronoun, whereas the majority – 67 % – consist of a resumptive pronoun. In the case of adverbial complements, 94 % appear with no resumptive pronoun. We found only a single case consisting of a resumptive pronoun. It is the second example considered in this study: Excerpt 1 (‘Pets’): 70

.. lakaxnu we took, ‘we took,’

71

le[vet hasefer to house [of] the book ‘to the open school,’

72

... sheyaron lamad bo? that Yaron studied at it (3P MASC SG PREP) ‘that Yaron studied at (where Yaron studied)?’

hapatuax], the open

These findings again contradict Ariel (1999: 257), who writes that “there is perhaps no default choice for object of preposition [coreferents] in current spoken Hebrew”. According to our findings, there is a strong default choice – Ø for adverbial complement (in 94 % of the cases), resumptive pronoun for oblique (in 67 % of the cases). The present findings document quantitatively for the first time the phenomenon long lamented by prescriptive Hebrew grammarians. The resumptive pronoun is not mentioned in altogether 77 % (17+4+15=36 out of 19+12+16=47) of the cases (excluding subject coreferents). However, non-

304

Yael Maschler

occurence of the resumptive pronoun is not equally distributed across the various syntactic roles. In the case of direct objects, the resumptive pronoun is not mentioned for 89 % of the cases, a possibilty allowed for by prescriptive Hebrew grammar. In the case of indirect objects, the resumptive pronoun is not mentioned for only 33 % of the cases. Thus, the majority of indirect object cases (67 %) actually adhere to prescriptive grammar. It is the adverbial complements that show a gross divergence from prescriptive grammar: 94 % of the cases (even more than in the case of direct objects) do not exhibit a resumptive pronoun.

4

The functions of relative clauses in discourse

In order to account for these patterns, let us consider the functions of relative clauses in discourse. In this corpus, relative clauses function in two main roles (a third role will be explored in Section 5.0). Most of the time (in about 66 % of all cases) they are of the ‘information-bearing type’ (Bernardo 1979, Beaman 1984), supplying new information concerning the antecedent. This is the case with all examples we have seen so far. For instance, in kavyot. she'atem natatem lahem. (‘guinea pigs. that you had given them.’, Excerpt 1, lines 42–44), the relative clause supplies new information concerning the guinea pigs. In 24 % of all cases, relative clauses are of the ‘identificatory type’ (Bernardo 1979, Beaman 1984). The speaker employs them in order to facilitate the hearer in identifying a previously mentioned antecedent. Examine, for instance, Excerpt 4: Excerpt 4 (‘Violence’): 111 Tom:

... ve'az ‘and then

ha--, the,’

112

.. [ze] this (DEMONST PRO MASC SG) ‘one who got the blows,’

113

.. hu haya ‘he was

114

.. dey mevugar kaze. pretty old sort of ‘a pretty old man sort of’.

shexatáf COMPL got

makot, blows

'ish, [a] man,’

Previously in the discourse, Tom described a fight he had witnessed between two men on a street. In line 112, he refers to one of them via the demon-

On the emergence of Hebrew adverbial connectives

305

strative ze (‘this’). In order to identify this referent further and distinguish him from the other man, Tom adds the relative clause shexatáf makot (‘that gót the blows’), with the main stress on the verb xatáf (got), (in contrast to ze shenatán (‘the one who gáve [the blows]’) found a few seconds later). This information is not new in the discourse but rather aids in deciphering the previously mentioned referent. As Beaman writes, the categorization of relative clauses into the information-bearing vs. identificatory types can only be understood within the context of the entire discourse (1984: 73). Restrictive relative clauses are sometimes defined as aiding in identification of the head (e.g., Quirk, Greenbaum, et al. 1985, Comrie 1989: 138) but not necessarily in relation to the discourse context. We have chosen to ignore the restrictive vs. non-restrictive distinction and examine relative clauses in terms of their discourse function. The restrictive vs. non-restrictive distinction is not commonly made in Hebrew, its structural prosodic characteristics (see below) often do not hold, and so it was not found useful for the present study. Excerpt 4 differs from all previous examples in yet another respect: it is the only one in which the antecedent, relativizer and relative clause all appear within the same intonation unit. It is tempting to correlate the identificatory type of relative clause with this prosodic pattern. Certainly prescriptivists using the term ‘restrictive relative clause’ for the identificatory type have claimed that such relative clauses are not to be separated from the antecedent (and main clause in general) by a comma in writing or a pause in speaking (e.g., Quirk, Greenbaum, et al. 1985: 1258). However, the distribution of relative clause constructions across intonation units shows that the pattern is more complex. We found that all three components (the antecedent, relativizer, and relative clause) appear in the same intonation unit in 50 % of the cases. Thus, information-bearing relative clauses as well are sometimes not separated from their head by an intonation unit boundary. Slightly less often (in 47 % of the cases) the components are spread across two separate intonation units, such that the antecedent appears separately. In the remaining 3 % of the cases, the antecedent and the relativizer appear in the same intonation unit (even though she- is a proclitic supposedly attached to the first element of the relative clause), whereas the relative clause appears separately:

306

Yael Maschler

Table 4: Distribution of relative clause constructions across intonation units antecedent + relativizer + relative clause in same intonation unit

1. antecedent 2. relativizer + relative clause (2 separate i.u.’s)

1. antecedent + relativizer 2. relative clause (2 separate i.u.’s)

Total

58 (50 %)

54 (47 %)

3 (3 %)

115 (100 %)

The factors influencing the distribution of the relative clause construction across intonation units include additional properties besides the function of the relative clause. In particular, Shaer 2007 found that for 94 % of the relative clause constructions throughout the data, the distribution across intonation units is such that an antecedent denoting new information (Chafe 1987, 1994) and an information-bearing relative clause do not appear within the same intonation unit. In this way, Chafe’s one-new-concept-at-a-time constraint (1994: 108–119) is maintained for the great majority of relative clause constructions. Fox and Thompson 2007 have studied the phenomenon of ‘optionality’ of the relativizer in English relative clauses. They have shown that the more the main clause and relative clause approach monoclausal status (i.e., the more intergrated with each other they are), the more likely we are to find no relativizer. Fox and Thompson show that for English relative clauses, the degree of their ‘monoclausality’ is a function of at least seven factors: – ‘emptiness’ of head NP – complexity of main clause (length of head NP, copular main clause) – unique head – length of object, oblique, and adverbial relative clause (ORC) verbal expression – subject of the ORC – breaks between head NP and ORC – rhythmic relationship between head NP and ORC. The ‘optionality’ in the case of spoken Hebrew relative clauses concerns not the relativizer but rather the resumptive pronoun. This ‘optionality’ can be correlated with some of the features in Fox and Thompson’s list (for ’emptiness of head NP, see 5.1.1, for ‘unique head’, see 5.1.2 below), but this requires further research. In either of the two functions of relative clauses – information-bearing or identificatory – the action accomplished by the relative clause accompanies another action with which the speaker is concurrently engaged (e.g., telling

On the emergence of Hebrew adverbial connectives

307

a story, making an assertion, performing a disagreement, etc.). There is evidence suggesting that accomplishing two actions simultaneously is a cognitively complicated enterprise. Previous research suggests that there are constraints on the amount of material that can be active in the mind of a speaker at a given moment in discourse (Pawley and Syder 1983, Chafe 1987, 1994). In the case of the relative clause construction, the need to attend to two syntactic structures simultaneously (that of the ‘main clause’ and of the ‘relative clause’) as well as the need to attend to the connection between them constitutes a relatively complex situation. In Modern Hebrew, an SVO language (Ravid 1977), the subject of the clause generally appears at its very beginning (or very close to it). Furthermore, the subject very often appears only in the form of an obligatory bound morpheme on the predicate. When a speaker initiates a relative clause, the confusion which might be caused by being engaged in two tasks and two structures simultaneously is only beginning to build up. This generally happens when the speaker is verbalizing the subject of the relative clause, which in Hebrew constitutes minimally an obligatory morpheme on the verb. It is thus not a surprise that subject coreferents are never omitted in Hebrew discourse. Further along in the verbalization of the relative clause, the cognitive overload on the speaker may build up more heavily. Thus, it is not surprising that towards the end of the verbalization of the relative clause – where in Hebrew we generally find the objects and adverbial complements – 77 % of resumptive pronouns are left out. The fact that direct object coreferents are left out more often than indirect ones may have to do with the fact that there is only one direct object, whereas the obliques are many. The avoidance of ambiguity may motivate retaining the oblique coreferent. In order to understand why adverbial complement coreferents are so often left out, we must investigate the antecedents more closely.

5

From relative clause to adverbial connective

Investigation of the types of antecedents that serve as heads of relative clauses will lead us to the grammaticization pattern we wish to explore here of adverbial connectives from relative clause constructions. 5.1 The semantics of the antecedent Table 5 details the distribution of relative clauses according to the semantic properties of the antecedent:

308

Yael Maschler

Table 5: Distribution of relative clauses according to semantics of antecedent Semantic type of antecedent Semantically empty NP/Pro

N (%)

Impersonal or generic NP/Pro

22 (19 %)

Numeral denoting impersonal

10 (9 %)

Demonstrative Pro

11 (9 %)

NP related to the concept of time

15 (13 %)

Non-empty NP, unrelated to the concept of time

57 (50 %)

Total

43 (37 %)

115 (100 %)

A relative clause modifies a nominal kernel. The kernels can be classified into (1) semantically empty antecedents, (2) antecedents denoting a concept related to time, and (3) all others. The last category is the largest, comprising 50 % of the cases. The antecedents kavyot (‘guinea pigs’), bet hasefer hapatuax (‘the open school’), shtey meymiyot (‘two canteens’), and gesher (‘bridge’) of Excerpts 1, 3 and 4 belong here. Of this last category of antecedents, 35 % appear with a Ø resumptive pronoun coreferent, a rate similar to the one found throughout the entire corpus (31 %, Table 2). 5.1.1 Semantically empty antecedents The second largest category – ‘semantically empty NP/Pro’ (37 %) – includes three subcategories. We have seen the third subcategory (9 % of all antecedents) – demonstrative pronoun – in ze shexatáf makot (‘this [or the one] who got the blows’) of Excerpt 2. The semantically empty antecedent can also be an impersonal or generic NP/Pro, such as mishehu (‘someone’) mashehu (‘something’), dvarim (‘things’), dvarim ’axerim (‘other things’), hahu (‘that one’) (19 % of all antecedents), as in the following excerpt: Excerpt 5 (‘Looking for a partner’): 17 Orna:

... 'ex ‘how

higat did you get

'elav? to him?’

18 Rina:

... 'e--h, ‘u--h,’

19

.. derex ['eyze mishehi], through some (MASC) somebody (FEM) ‘through somebody (FEM),’

309

On the emergence of Hebrew adverbial connectives shegam hayta ‘who also was (FEM)

20

reshuma registered

'etslo. with him.’

The semantically empty noun mishehi (‘somebody’ FEM) is modified by the pronoun 'eyze (‘some’ MASC7) and this entire NP is then modified by the following relative clause. Finally, the semantically empty antecedent can also be a numeral denoting an impersonal entity, such as 'exad (‘one’, MASC), 'axat (‘one’, FEM), shtayim (‘two’, FEM), hasheni (‘the second [one]’, MASC), harbe8 (‘lots’) (9 % of all antecedents), as in the following excerpt: Excerpt 6 (‘Looking for a partner’): 114 Orna:

.. hu ‘he’

115

hu ‘he

116

ke['exad] ‘as one (MASC)

117

.. leshadex. make matches ‘do matchmaking.’

manpik markets

'et 'atsmo himself she . . who ..

ke as’ yaxol can (MASC)

'eh, uh,’

We see here the numeral 'exad (‘one’, MASC) modified by the relative clause sheyaxol 'eh leshadex (‘who can uh do match-making’). We have already seen (section 4.0) that Fox and Thompson found that in English, monoclausality of the relative clause construction is a function of emptiness of the head, among six other factors. Fox and Thompson found also that the most frequent subtype of ORC (object, oblique, and adverbial relative clause) in their data is not only monoclausal, but tends to have a head which is empty, such as thing, something, one (2007: 312). In the case of our Hebrew data, semantically empty antecedents constitute 37 % of all antecedents (43 tokens). Only 6 of them (14 %) appear with a Ø resumptive pronoun. This rate is considerably lower compared to the one found throughout the entire corpus (31 %, Table 2). Thus, it is generally not the semantically empty antecedents that contribute to the Ø resumptive pronoun coreferent phenomenon in Hebrew. 7

8

As in Excerpt 3, line 105, again, gender agreement is not maintained and the masculine form of the pronoun is employed instead of the feminine 'eyzo. This is not technically a numeral, but was classified into this category because of the similarity in semantics.

310

Yael Maschler

5.1.2 Antecedents related to the concept of time The category that will help us understand why adverbial coreferents are not mentioned in 94 % of the cases is the second one – NPs denoting a concept related to time (13 % of all antecedents). Examine the following four excerpts: Excerpt 7 (‘Fallow year’): 117 Ron:

.. 'axshav, ‘now,’

118

... ma ‘what[’s]

119

.. ba[rega] sheyesh le'atsits in minute that there is to a planter ‘the minute a planter has a hole,’

120

velo ‘and [it] doesn’t

ha'inyan the deal

'im with

xashuv matter

haxor? the hole?’

'im whether

'omed babayit, stands in the house,’ 121

'o baxuts, ‘or outside’,

122

... yesh lo maga, there is to it contact ‘it has contact,’

123

'im ‘with

ha'adama. the ground.’

Excerpt 8 (‘Breaking up’): 109 Eran:

..... ba[yom] shehem yitxatnu, ‘on the day that they marry,’

110 Tal:

....

111 Eran:

.. ken. ‘yes,’

112 Tal:

.. nu? ‘go ahead?’

113 Eran:

.. 'en, there isn’t ‘nothing doing,’

shehem ‘that they

yitxatnu? marry?’

hu it

xor, hole

311

On the emergence of Hebrew adverbial connectives 114

..

gamarnu 'im ha'atsma'ut shelo. we’re done with the independence his ‘that’s the end of his independence.’

Excerpt 9 (‘Grandma’s remedies’): 102 Ayala:

.. [kol ‘every

103

shehayu ‘that would

104

'e--h, ‘uh,’

105

ze, ‘this {i.e. ‘weird things such as grandmother-type remedies’},’

106 Hanna:

... hayu ‘they used

107

pa'am] time

shehayiti that I would

nir'im seem

li to me

'ovdim to work

'osa do

dvarim, things,’

ktsat, a bit,’

'alayix. for you.’

(laughter)

108 Ayala:

... hayu ‘they used

'ovdim, to work,’

109

.. 'en xoxmot. they’re no ‘tricks’ ‘without fail.’

Excerpt 10 (‘Obituaries’): 1

...

[shavua] she'ava--r, week that passed (3P MASC SG) ‘last week,’

2

...

'aba sheli, father mine ‘my father,’

3

.. kara ‘[was] read[ing]

'iton. [the] paper.’

The 15 cases in this category share four features: 1) Semantically, they all consist of antecedents having to do with the concept of time. They are: rega (‘minute’, 4 tokens), yom (‘day’, 3 tokens), yom rishon (‘Sunday’ (lit. ‘first day’), 1 token), shavua (‘week’, 2), kol pa'am (‘every

312

Yael Maschler

time’, 1), hapa'am hakodemet (‘the previous time’, 1), pa'am 'axrona (‘last time’, 1), pa'am rishona (‘first time’, 1), hapa'am harishona (‘the first time’, 1).9 2) The coreferents are all realized by a Ø. The category is divided into two subcatories: those with a Ø resumptive pronoun coreferent (12 cases, as in Excerpts 7–9) and those with a subject coreferent (3 cases, as in Excerpt 10). The subject coreferent always consists of the ø third person masculine singular subject morpheme on the verb: shavua she'avar-ø (‘week that passed’, (last week)), yom rishon she'avar-ø (’Sunday that passed, (last Sunday)). 3) For 75 % of these time antecedents, the antecedent, relativizer, and the relative clause, and sometimes the preceding preposition as well, all appear within the same intonation unit. 4) In all of these cases, the antecedent, relativizer, and sometimes the preceding preposition and following relative clause as well, have all come to function as a single unit, connecting what follows to the main clause. For instance, in Excerpt 7, we find the time antecedent rega (‘minute’). The coreferent is a Ø resumptive pronoun. The ‘correct’ form according to prescriptive grammar should have been: ba[rega] sheyesh le'atsits xor bo, in minute that there is to planter hole in it (3P MASC SG PREP) ‘in the minute in which a planter has a hole,’ ‘the minute a planter has a hole,’ yesh lo maga 'im ha'adama. ‘there is to it contact with the ground.’ ‘it has contact with the ground.’

This is because the coreferent of rega (‘minute’) functions as adverbial complement in the relative clause (the planter has a hole at that minute). The antecedent, relativizer, and relative clause (as well as the preceding preposition) all appear within the same intonation unit. The definite noun harega (‘the minute’), along with the preposition be- (‘in/at’) preceding it, have merged with the relativizer, so that barega she- (‘in the minute that’) no longer constitutes a preposition + antecedent in the main clause, modified by a relative clause, but rather, they have become a single unit – a temporal connective – connecting the entire clause following it (yesh le'atsits xor ‘a planter has a hole’) to the main clause (yesh lo maga 'im ha'adama, ‘it has contact 9

The last three items are reminiscent of Fox and Thompson’s category ‘unique head’, which is another one of the properties of relative clauses influencing their monoclausality (and lack of relativizer). Fox and Thompson’s unique head category includes superlatives and nouns with only and first, such as the only place, the ugliest set of shoes, the first compliment (2007: 301).

On the emergence of Hebrew adverbial connectives

313

with the ground’), much like the temporal connectives lifney she- (‘before’) and 'axarey she- (‘after’). Schematically, this can be diagrammed as follows: Subj

Pred

Adv Compl

as Antecedent + she- + Relative Clause

(main clause)

Subj

Pred

+ Temporal Adverbial Clause

(main clause) Fig. 1

Note that the English translation of this ‘relative clause’ (‘the minute a planter has a hole’) shows omission not only of the coreferential element (the preposition in + case marking on the relativizer which), but of the relativizer altogether.10 However, in Hebrew, the symmetry with other adverbial connectives (not only of time, but also of cause (e.g., mipney she- ‘because’), concession (e.g., lamrot she- ‘although’) as well as other adverbial relations, see below) resulting from the ‘subordinator’ she-, which is never deleted in spoken discourse, shows the new category of connective more clearly.11

10

11

Indeed, Tottie, studying relative clauses in the British National Corpus, finds the following English constructions with no relativizer as “border[ing] on grammaticalization (cf. the way Ø you look tonight, the time Ø he left, the reason Ø he did it)” (1997: 5). Tottie and Lehmann 1999 find that the antecedents way and time account for fully 82 % of all relative clauses with Ø relativizer and that they account for most of the “zero constructions” in their British National Corpus data. Another variant exists for this temporal connective -- berega she-, without the definite article ha- fused with the preposition be- to form ba- (note the definite English ‘equivalent’ -- ‘the minute that’). The berega she- variant, then, could be said to exhibit further bonding within the unit -- deletion of the definite article ha-. Cf. Schwarzwald (2006), on the phenomenon of deletion of the Hebrew definite article ha- preceding the N in a definite NP whose modifier is definite or constitutes a relative clause, in temporal expressions such as beshana she'avra (‘in the year that passed’, i.e., ‘last year’) or beshana haba'a (‘in the year that will come’, i.e., ‘next

314

Yael Maschler

Another argument supporting this analysis is the fact that without the following ‘relative clause’, the utterances would sound peculiar: Excerpt 7: ba[rega] yesh lo maga 'im ha'adama in the minute there is to it contact with the ground ‘in the minute it has contact with the ground’

Excerpt 8: ba[yom] ‘in the day

gamarnu that’s the end

'im ha'atsma'ut shelo. of his independence.’

Excerpt 9: [kol pa'am] ‘every time

hayu they used to

'ovdim, work,’

Semantically, the antecedents rega (‘minute’), yom (‘day’), and kol pa'am (‘every time’) are somewhat bleached when they are followed by the relative clause. By contrast, in Excerpt 1: hayu ‘they

lahem dagim, had fish,

ve'et ma ‘and those

[kavyot]. guinea pigs.’

shenish'ar, that were left,

lakaxnu we took

le[vet hasefer hapatuax] to the open school.’

Excerpt 2: bekitser, yesh [gesher], ‘anyhow, there’s a bridge’

the antecedents ‘guinea pigs’, ‘the open school’, and ‘bridge’ remain unbleached semantically. The relative clause that follows only sharpens their meaning by supplying further information concerning the antecedent. Finally, unlike the non-empty antecedents unrelated to time, which tend to appear at the end of the main clause, time antecedents tend to precede the main clause. Ford has shown that English temporal adverbial clauses which precede the main clause function to determine or change the temporal frame of the discourse (1993: 28–42). This function indeed characterizes also the year’), rather than bashana haba'a (Hebrew nouns must agree with their modifiers also in definiteness) (see below).

315

On the emergence of Hebrew adverbial connectives

Hebrew time antecedents of the present corpus and it is a function different from the previous two functions of relative clauses discussed in section 4 (the information-bearing and identificatory functions). For example, in shavua she'avar (‘last week’) of Excerpt 10, the relative clause she'avar (‘that passed’) functions not so much to supply information concerning the antecedent shavua (‘week’) nor to indentify a previously mentioned antecedent (this antecedent does not appear previously in the discourse). Mainly, this relative clause appears at the opening of a narrative, setting the time frame of the events told and differentiating them from the present time of the telling12. Thus, time antecedents differ from other antecedents not only semantically and syntactically, but pragmatically as well. 5.2 Grammaticization of adverbial complement constructions We have seen (Figure 1) that from a situation of an adverbial complement in the main clause which is modified by a relative clause, there has been movement towards a situation of a predicate in the main clause modified by a temporal adverbial clause. Hopper and Traugott define grammaticalization13 as “the change whereby lexical items and constructions come in certain linguistic contexts to serve grammatical functions and, once grammaticalized, continue to develop new grammatical functions” (2003: xv). In the cases studied here, a content word (lexical item) (e.g., rega ‘minute’, yom ‘day’, kol pa'am (‘every time’), shavua (‘week’)) has connected to a function word (the ‘subordinator’ she-), and the two elements together now function as a function word – i.e., a temporal connective: content word rega (‘minute’)

+

function word she-

function word connective

In other words, grammaticization has occurred for these particular relative clause constructions. And so the merging of the time antecedent with the

12

13

For a detailed study of the entire narrative from which Excerpt 10 is taken, see Maschler 1998. Hopper and Traugott use the term ‘grammaticalization’ rather than ‘grammaticization’. I employ here Hopper’s term, ‘grammaticization’ (1987, 1988), so as not to invoke a possible understanding of “entering the grammar of a language” (Hopper and Traugott 1993: xvi) and in order to focus on the continually changing categories and meanings from a discourse-pragmatic, rather than a historicaldiachronic, perspective. For more on this, see Maschler (2009: 33–34).

316

Yael Maschler

relativizer has come to be realized as a temporal connective. Since the remainder of the utterance is now interpreted as a temporal adverbial clause, and such clauses do not require a resumptive pronoun, it is therefore not surprising that so many resumptive pronouns are not verbalized. This grammaticization in turn, then, serves as further motivation for not verbalizing the resumptive pronoun.14 The grammaticization of the time antecedent and relativizer into a temporal connective has not occurred to the same extent for all time antecedents. Of the time antecedents found throughout our corpus, it is those involving the nouns rega (‘minute’) and pa'am (‘time’) which can be characterized as full-fledged Hebrew temporal connectives – barega she- (‘the minute that’), kol pa'am she- (‘every time’, or ‘whenever’). These two nouns are also the ones found most frequently among the time antecedents in our corpus (rega, 4 tokens, pa'am (in various NPs) 5 tokens). Indeed, as much corpus-based research has shown (e.g., Thompson and Mulac 1991, the various studies in Bybee and Hopper 2001, Thompson 2002, Bybee 2003, Tao 2003), it is those constructions most commonly repeated in discourse that are more likely to undergo grammaticization.15 Since the great majority of adverbial complement coreferents are temporal in our database (12 out of 16, 75 %), and since some other adverbial complement coreferents have undergone similar grammaticization (see below), we now understand why 94 % of the adverbial complement coreferents in the corpus are realized as a Ø resumptive pronoun. Furthermore, we have seen that 12 out of 15 (80 %) of all time antecedents appear with a Ø resumptive pronoun (i.e., 100 % of the cases in which the coreferent of the time antecedent functions as adverbial complement in the relative clause). This rate is by far greater compared to the one found throughout the entire corpus (31 %, Table 2). We have seen that the rate of a Ø resumptive pronoun for semantically empty antecedents is only 14 % 14

15

Ariel relates to these adverbials in footnote 17 of her article: “Another category of cases where OPrep’s are coded by zero is when the head is an optional time or place NP and serves as an adverbial in the RC [… T]hese examples deserve a careful examination, because I think they are the source for creating time and place adverbials which take complements. […] Such cases have already grammaticized as noun sentential complements, I believe, and were not counted in this study” (1999: 261). The present study examines precisely those cases not counted in Ariel’s study and shows that exploring them actually sheds light on the motivation for Ø resumptive pronoun coreferents. However, there is another factor contributing to this grammaticization; namely, the syntactic role of the autecedent in the main clause; see Maschler 2011.

On the emergence of Hebrew adverbial connectives

317

whereas that for non-empty antecedents unrelated to the concept of time is 35 %. Thus, it is the time antecedents that contribute, more than any other semantic type of antecedent, to the phenomenon of a Ø resumptive pronoun coreferent in this corpus. 5.2.1 Adverbial complements of place The lack of resumptive pronoun is found in the corpus not only for temporal adverbial complements, but for place and manner adverbial complements as well (see Table 1). Regarding adverbial complements of place (3 tokens), examine the following utterance which is found immediately preceding a narrative. Orna responds here to her sister’s request to tell a story about the marriage proposal she had just accepted: Excerpt 11 (‘Marriage proposal’): 12 Orna:

.. lehatxil ba[keta] shehiganu lirushalayim? to begin at the part that we arrived in Jerusalem? ‘should I begin at the part where we arrived in Jerusalem?’

The antecedent keta (‘part’, lit. ‘segment’) is followed by a relative clause in which the coreferential element functions as adverbial complement of place – ‘we arrived in Jerusalem at that part (of the story)’. The correct form, according to prescriptive grammar, should have been: lehatxil [baketa]

shebo higanu lirushalayim? that at it to begin at the part we arrived in Jerusalem? (3P MASC SG PREP) ‘to begin at the part at which we arrived in Jerusalem?’

As in the case of temporal adverbial complements, the resumptive pronoun bo (‘at it’, 3P MASC SG) is not verbalized. The noun keta (lit. ‘a segment’) has become semantically bleached.16 No longer are we concerned with the geometrical concept of a ‘segment’ but rather with a ‘segment’, or part, of a story. However, in the context of a story, it is no longer completely clear whether ‘part’ concerns the semantic category of place or time.17 16

17

This bleaching can be seen also in grammaticization processes involving the noun keta (‘segment, part’) in metalingual constructions which have come to function as discourse markers (see Maschler 1998). The possible lack of resumptive pronoun in the case of adverbial complements of time and place was documented by Peretz (1967: 92) already for Biblical Hebrew. As an example of a place adverbial complement in Biblical Hebrew, Peretz cites e.g.: vayatsev ya'akov matseva bamakom 'asher diber 'ito [bo/sham] (‘Jacob erected a

318

Yael Maschler

The construction baketa she- (‘at the part that’) hasn’t grammaticized into an adverbial connective to the same extent as time antecedents such as barega she- (‘the minute’) and kol pa'am she- (‘every time’, or ‘whenever’) have. Table 1 shows that time adverbial coreferents are also four times more common throughout the data than place coreferents. Nevertheless, we will argue that the lower extent of grammaticization here is due not only to frequency, but also to the unique nature of the semantics of time (see section 5.2.2 below).18 5.2.2 Adverbial complements of manner Table 1 shows a single manner adverbial coreferent throughout the data. It is shown in Excerpt 12. Ora constructs here the dialogue of a woman describing the state of her closet following a burglary. She then reformulates following the discourse marker ke'ilu (‘like’) (Maschler 2002), emphasizing that the point is not only what the burglars took (described in the immediately preceding lines), but also the way in which they did it – leaving total chaos in the apartment:

18

monument at the place that he talked with him [at it/there]’, Genesis 35: 14). As an example of a time adverbial complement, he gives e.g.: lemin hayom 'asher yatsata [bo] me'erets mitsrayim (‘from the day that you left the country of Egypt [at it]’, Deuteronomy 9: 7). However, Peretz does not provide a quantitative account of the phenomenon and does not tie it to processes of grammaticization. The change in the construction baketa she- (‘at the part that’) is related to another type of grammaticization. The extreme bleaching of the noun keta could result in replacing it with semantically empty antecedents such as the demonstrative ze (‘this’) or the noun 'uvda (‘fact’), as in: lehatxil

be[ze]

shehiganu

lirushalayim?

to begin

with this

that we arrived

in Jerusalem?

‘to begin with the fact that we arrived in Jerusalem?’ lehatxil

ba['uvda]

shehiganu

lirushalayim?

to begin

with the fact

that we arrived

in Jerusalem?

‘to begin with the fact that we arrived in Jerusalem?’

Such constructions are no longer considered relative clauses according to prescriptive Hebrew grammar but rather are known as attributive content clauses (Zewi 2008), since there is no longer an element in the subordinate clause coreferential with the antecedent, either explicitly or one that could be added (Tsadka 1989). As in the case of movement from a relative clause construction to an adverbial clause construction in the case of time antecedents, here we see grammaticization of attributive content clauses from relative clauses. For more on this, see Maschler 2011.

319

On the emergence of Hebrew adverbial connectives

Excerpt 12 (‘Burglary’): 152 Ora:

....

hi 'omeret, ‘she says,’

153

....

kol ha'aron haya 'al haritspa. all the closet was on the floor ‘the entire closet was on the floor.’

154

... ke'ilu, ‘like,’

155

... [hatsura] shehem ‘the way that they

156

...

'asu 'et ze did it

gam, also,’

hakol ! ‘everything’!

The antecedent tsura (‘way, manner’) is followed by a relative clause in which the coreferential element functions as adverbial complement of manner – ‘they did it in such a way’. The correct form, according to prescriptive gram19 mar, should have been: [hatsura] the way ‘the way ‘the way19

sheba that in it (3P FEM SG PREP) in which they did it’ they did it’

hem they

'asu did

'et ze it

We see that, again, the resumptive pronoun ba (‘in it’, 3P FEM SG) is not verbalized. Fox and Thompson list the noun way as one of the most frequent heads (along with thing, something, time, and one) employed in the most common type of monoclausal ORCs (object, oblique, and adverbial relative clauses) in their English database (e.g., that’s [the way] it is, is there [any way] he could meet us? (2007: 312)). Fox and Thompson classify the semantically empty antecedents along with the temporal and manner antecedents here. However, the Hebrew data suggest classifying them separately, since it is specifically not empty antecedents, but rather antecedents denoting concepts which can serve as adverbial complements (e.g., time, place, manner, cause, condition (see below)) that may grammaticize into adverbial connectives.20 As in the case of baketa she- (‘at the part that’), the construction hatsura she(‘the way that’) has not grammaticized into an adverbial connective quite to 19 20

Cf. Tottie 1997, Tottie and Lehmann 1999 (footnote 10). In Hebrew, semantically empty antecedents, such as the demonstrative ze (‘this’, MASC SG), may be involved in another type of grammaticization; namely, the grammaticization of relative clause constructions into attributive content clauses (see footnote 18). See also Maschler 2011.

320

Yael Maschler

the same extent as some time antecedents have. This is not strictly a matter of frequency. Some adverbial complements are of a semantic nature more likely to become grammaticized into adverbial connectives in a particular culture than others. This has to do with the centrality of certain semantic categories to speakers of that culture. Becker (1979) has shown that whereas in many ‘Western’ cultures, time is a concept more central to the discourse activity of relating events (Labov and Waletzky’s definition of narrative (1967) certainly confirms this); in Javanese, it is rather the semantic category of place that is more central as an organizing category. In Indo-European as well as Semitic languages, for instance, time is the semantic category that has become grammaticized as a bound morpheme on the verb, most likely because of a discourse need to compare the time of the narrated event to that of the telling. It is not surprising, then, that speakers of such a language will tend to relate more to the temporal aspects of the events they talk about (rather than to their location or manner, e.g.) also via adverbial complements. This might explain why in Hebrew, of all adverbial coreferents, it is the temporal ones which have grammaticized most often into adverbial connectives. Fox and Thompson also suggest that frequency is perhaps not the only factor involved in the grammaticization they describe for English relative clauses. However, the other factor they mention, “mergedness” of the two clauses in the relative clause construction, also involves frequency, as they, too, admit (2007: 318–319). Here we have a case of a clearly different type of factor; namely, the centrality to speakers of a particular culture of the semantic category of the antecedent involved. As Bybee writes, Repetition alone, however, cannot account for the universals of grammaticization. […] It is not just the fact of repetition that is important, but in addition what is repeated that determines the universal paths. The explanation for the content of what is repeated requires reference to the kinds of things human beings talk about and the way they choose to structure their communications (2003: 622).

5.2.3 Reanalysis as a single processing chunk Haiman (1994) suggests regarding the process of grammaticization as ritualization resulting from repetition. Bybee cites several properties characterizing this ritualization, according to Haiman, some of which are of particular interest to us: [R]epetition leads to the automatization of a sequence of units, and the reanalysis of the sequence as a single processing chunk, with formerly separate units losing their individual meaning; repetition also leads to the reduction of form through the weaken-

321

On the emergence of Hebrew adverbial connectives

ing of the individual gestures comprising the act, and through the reorganization of a series of formerly separate gestures into one automated unit (Bybee 2003: 603).

Hebrew time antecedents often show an additional reduction of form besides the lack of resumptive pronoun. Examine again Excerpt 9: Excerpt 9 (‘Grandma’s remedies’): 102 Ayala:

.. [kol ‘every

103

shehayu ‘that would

104

'e--h, ‘uh,’

105

ze, ‘this {i.e. ‘weird things such as grandmother-type remedies’},’

106 Hanna:

... hayu ‘they used

107

pa'am] time

shehayiti that I would

nir'im seem

li to me

'ovdim to work

'osa do

dvarim, things,’

ktsat, a bit,’

'alayix. for you.’

(laughter)

108 Ayala:

... hayu ‘they used

'ovdim, to work,’

109

.. 'en xoxmot. they’re no ‘tricks’ ‘without fail.’

Besides the lack of the resumptive pronoun, we also find a lack of the preposition be- (‘in/at’) connecting the temporal adverbial complement to the predicate of the main clause. The correct form, according to prescriptive grammar, should have been: be [kol pa'am] shehayiti 'osa ba at every time that I would do at it ((3P FEM SG PREP) ‘at every time at which I would do things …’

dvarim …, things …,

Thus, the time antecedent kol pa’am (‘every time’) participated in the follow21 ing process: be[kol pa'am] she … ba > be[kol pa'am] she- > [kol pa'am] sheat every time that … at it at every time that every time that ‘at every time at which’ ‘whenever’21 21

Note the temporal connective whenever in the English translation. We see that grammaticization of adverbial connectives from relative clause constructions can also be found in certain environments in English.

322

Yael Maschler

This further reduction in form can be accompanied by the entire relative clause construction grammaticizing into a single unit. We can see this particularly in constructions with a subject coreferent consisting of the ø third person masculine singular subject morpheme on the verb: shavua she'avar-ø (‘week that passed’, (last week)), yom rishon she'avar-ø (‘Sunday that passed, (last Sunday)’, as in Excerpt 10: Excerpt 10 (‘Obituaries’): 1

... [shavua] she'ava--r, week that passed ‘last week,’

2

... 'aba sheli, father mine ‘my father,’

3

.. kara ‘[was] read[ing]

'iton. [the] paper.’

The coreferent here is the ø subject morpheme on the verb 'avar (‘passed’), which cannot be omitted. However, the preposition be- (‘in/at’) connecting this adverbial complement to the main verb kara (‘was reading’) has been omitted.22 In addition, the entire relative clause construction [antecedent + relativizer + relative clause] has become reanalyzed as a single processing chunk functioning as adverbial complement of the main verb. This can be ascertained from the fact that the construction shavua she'avar without the verb of this so-called ‘relative clause’ is not a grammatical utterance: *shavua 'aba sheli kara 'iton. week father my read newspaper *‘week my father was reading the newspaper.’

Interestingly, neither is the equivalent English utterance grammatical, although the construction last week does not involve a relative clause construction.

22

Cf. Schwarzwald 2006, who studied the deletion of the preposition be- in the fixed temporal expressions (not necessarily involving relative clause constructions) shana she'avra / haba'a (‘last / next year’), shavua she'avar / haba (‘last / next week’), shi'ur she'avar / haba (‘last / next lesson’), basing her study on written internet discourse (employing Google to locate all instances). Schwarzwald found that the more frequently employed temporal expressions manifest a greater extent of deletion of the preposition be-.

On the emergence of Hebrew adverbial connectives

6

323

Conclusion

We have seen that time antecedents tend to merge with the relativizer and to form a new construction in Hebrew – a temporal connective. What follows then gets reinterpreted as a temporal adverbial clause, rather than as a relative clause, and therefore does not require a resumptive pronoun. In the same way, we can explain the emergence of Hebrew conditional connectives from relative clauses modifying place antecedents, such as bemikre she- (‘in the event that’, i.e., ‘if ’), (cf. bemirkim she- (‘in cases in which’) from the newspaper example cited in section 1), as well as the emergence of Hebrew causal connectives, such as misiba she- (‘from the reason that’, i.e., ‘because’), which is not found in our data, but has become a common causal connective in Hebrew (cf. Ariel 1999: 261). The notion of emergent construction is thus useful for understanding one aspect of the elusive nature of Hebrew resumptive pronouns. The grammaticization of adverbial connectives from relative clause constructions shows many of the common features characterizing grammaticization phenomena. We have seen, first of all, that the process involves lexical elements (nouns denoting concepts related to time, place, manner, etc. …) merging with a functional element (the relativizer she-) and becoming functional elements (connectives). Thus, the principle of unidirectionality in grammaticization (i.e., the general direction content f function) (Hopper and Traugott 2003: 106–109) is observed. Second, this particular grammaticization pattern was seen to exhibit many semantic characteristics common to grammaticization, such as semantic bleaching of the nouns serving as heads of these relative clause constructions, but also persistence of the original semantic meaning, as older meanings of a form “glimmer through” (Bybee and Pagliuca 1987), so that, e.g., antecedents associated with the concept of time become temporal connectives. Generalization of meaning (Traugott 1995) can also be observed, as the heads move from being more to less referential. The constructions show pragmatic strengthening (ibid.) in their “movement away from [their] original specific and concrete reference […] toward increasingly general and abstract reference” (Pagliuca 1994: ix), such as setting the stage for a narrative (in the case of shavua she'avar (‘week that passed’, (last week), Excerpt 10, for instance). Third, many of the structural properties characterizing grammaticization are observed. We have seen morphological reduction as the resumptive pronoun, and occasionally the preposition connecting the head to the main clause, are not verbalized, and we have seen much bonding within the phrase, so that the antecedent, relativizer, and sometimes the preceding preposition

324

Yael Maschler

and following relative clause as well, all come to function as a single processing chunk, connecting what follows to the main clause, i.e., functioning as a connective. As a result of this grammaticization, the syntactic projectability of these constructions (Auer 2005, Maschler 2009: 225–9) also changes, from one projecting a relative clause modifying one of the arguments of the verb, to one projecting an adverbial clause modifying the verb. The process is embedded within other grammaticization phenomena in Hebrew, so that other types of relative clause constructions, those involving certain semantically empty antecedents, e.g., have become attributive content clauses (see footnote 18, also Maschler 2011). In much earlier periods of the language, other grammaticization processes resulting in connectives (not involving relative clause constructions necessarily) have occurred. For instance, nouns referring to body parts (e.g. panim ‘face’) have become parts of other connectives (lifney she- lit. ‘to the face that’, ‘before’; and mipney she- lit. ‘from the face that’, ‘because’). These body part nouns often became also parts of prepositions, as in the case of lifney (lit. ‘to the face of ’, before (PREP)). Similarly, e.g., yad (‘hand’) became part of the preposition 'al yedey (lit. ‘on the hand of ’, ‘by’) and gav (‘back’) – part of the preposition 'al gabey (lit. ‘on the back of ’, ‘on’)). Thus, the grammaticization observed here is only one among many grammaticization processes which have occurred in Hebrew over the years. Throughout the study, we have seen some parallels with English, as in the case of, e.g., the connective at every time at which f whenever (see footnote 21). We have also seen some similarities with the various monoclausal relative clause constructions studied in Fox and Thompson (2007) involving heads such as time, minute, way, and the semantically empty heads. However, the English grammaticization paths are of course not identical and often less clear because of differences between the structures of the two languages. Most importantly, although some English adverbial connectives, such as when, do include traces of the relativizer in the form of the wh- morpheme, this is not necessarily the case for most adverbial connectives (e.g., English before, after, because, although). In Hebrew, on the other hand, the relativizer and general ‘subordinator’ she- (‘that’) constitute the same morpheme (at least synchronically), and this morpheme is a component of the great majority of adverbial connectives (e.g., mipney she-, mishum she-, mikeyvan she-, heyot she-, biglal she- (all of them variants of ‘because’), 'al 'af she-, 'af 'al pi she-, lamrot she- (‘although’), lifney she-, (‘before’), 'axarey she- (‘after’), etc. …), and so there is opportunity for symmetry within the grammatical system when the antecedent + relativizer grammaticize into an adverbial connective.

On the emergence of Hebrew adverbial connectives

325

The notion of emergent construction is useful not only for understanding one aspect of the elusive nature of Hebrew resumptive pronouns, but also for rethinking our current syntactic models. Towards the end of their study of English relative clauses, Fox and Thompson write: Traditionally, noun phrases containing Relative Clauses have been assumed to be “constructions” with their Main Clauses. […] Our findings suggest, however, that the notion of a single “construction” for Relative Clauses is problematic. [… T]here is a great deal of small, local patterning, with varying degrees of lexical specificity, within the large domain we think of as Relative Clause. [… T]he concept of grammatical organization underlying the notion “Relative Clause Construction” might be usefully replaced by a view of grammatical organization that has small-domain, even sometimes lexically specific, formats which exist in a dynamic, family-resemblance relationship to one another that can be modeled in tems of a continuum. (2007: 315–318)

In our Hebrew data, the continuum stretches from the lexically-specific “relative clauses” which have grammaticized into temporal connectives such as kol pa'am she- (‘every time’, ‘whenever’), or into temporal adverbial fixed expressions such as shavua she'avar (‘week that passed’, (‘last week’)), to the non lexically-specific, bi-clausal, instantiating no low-level format cases, such as ve'et ma shenish'ar, lakaxnu, levet hasefer hapatuax, sheyaron lamad bo (‘and those that were left, we took, to the open school, that Yaron studied at’), which is much more along the lines of the traditional “Relative Clause Construction”. At other points along this continuum, closer to the lexically-specific end, we find constructions with less frequently employed time antecedents, as well as adverbial complement constructions involving antecedents of a different semantic nature (e.g., place, manner, cause, condition). Some of the constructions have grammaticized into attributive content clauses (see footnote 18). And on yet other points along this continuum, closer to the non lexically-specific end, will be located constructions with semantically empty antecedents of various kinds – impersonal or generic NP/Pros, numerals denoting impersonals, and demonstrative Pros – again, with a different degree of fixedness each. “[S]peakers make use of a wide range of practices – some entirely prestored, others partially pre-stored and partially composed based on low-level formats, others not at all pre-stored” (Fox and Thompson 2007: 319). Our syntactic theories must allow room to account for this diversity rather than focus on one particular end of the continuum. The notion of emergent construction allows this. As Hopper has written about constructions in general, we have seen that the members of the relative clause construction “are related not as central and marginal instances but rather as a democratic chain whose members are linked family resemblance style” (2004).

326

Yael Maschler

Appendix: Transcription method Each line denotes an intonation unit and is followed by an English gloss. In the cases in which this gloss is not close enough to an English utterance, it is followed by a third line supplying a usually literal (but sometimes functional) translation. Utterances under consideration (usually relative clauses) are given in boldface. The antecedent is marked also by square brackets. Transcription basically follows Chafe 1994, with a few additions. Conventions are as follows: ... half second pause (each extra dot = another 1/2 second) .. perceptible pause of less than half a second (3.56) measured pause of 3.56 seconds á primary accent à secondary accent , comma at end of line -- continuing intonation (‘more to come’) . period at end of line -- sentence final falling intonation ? question mark at end of line -- sentence final rising intonation, ‘appeal intonation’ (Du Bois et al. 1992) ! exclamation mark at end of line -- sentence final exclamatory intonation ø lack of punctuation at end of line -- a fragmentary intonation unit, one which never reached completion -two hyphens -- elongation of preceding vowel sound square bracket to the left of two consecutive lines indicates overlapping speech, two speakers talking at once alignment such that the right of the top line is placed over the left of the bottom line indicates latching, no interturn pause /??????/ transcription impossible /words/ within slashes indicate uncertain transcription pp pianissimo, very softly ff fortissimo, very loudly [xxxxx] material within square brackets in the gloss indicates exuberances of translation (what is not there in the original) {curly brackets} transcriber’s comments concerning paralinguistics and prosody, which do not have an agreed upon symbol in this transcription system. boldface relative clause along with its main clause, if there is one [xxxxx] material within square brackets in the transliterated text indicates head of relative clause

On the emergence of Hebrew adverbial connectives

327

List of Abbreviations DIR OBJ MRK 3P SG PL MASC FEM PREP COMPL PRO SUBJ PRO DEMONST INF

direct object marker 3rd person singular plural masculine feminine preposition complementizer / relativizer pronoun subject pronoun demonstrative infinitive

References Ariel, M. 1999 Cognitive universals and linguistic conventions: The case of resumptive pronouns. Studies in Language 23, 217–269. Auer, P. 2005 Projection in interaction and projection in grammar, Text 25, 7–36. Azar, M. 1995 taxbir leshon hamishna (‘Syntax of Mishnaic Hebrew’), Jerusalem, 214–222. Beaman, K. 1984 Coordination and subordination revisited. In: D. Tannen (ed.), Coherence in Spoken and Written Discourse, Norwood, 45–80. Becker, A. L. 1979 Text-building, epistemology, and esthetics in Javanese shadow theater. In: A. L. Becker & A. Yengoyan (eds.), The Imagination of Reality, Norwood, 211–43. 1995 Beyond Translation: Essays Toward a Modern Philology, Ann Arbor. Bernardo, R. 1979 The function and content of relative clauses in spontaneous oral narratives. Proceedings of the Fifth Annual Meeting of the Berkeley Linguistics Society 5, Berkeley, 539–551. Blau, Y. 1977 zutot mitxum mishpetey hazika ba'ivrit hamikra 'it (‘Issues relating to relative clauses in Biblical Hebrew’). shnaton lamikra ulexeker hamizrax hakadum (‘Annual Studies of Bible and Ancient Near East’) 2, 50–53. Borer, H. 1984 Restrictive relatives in Modern Hebrew. Natural Language and Linguistic Theory 2, 219–160. Bybee, J. 2003 Mechanisms of change in grammaticization: The role of frequency. In: B. D. Joseph & R. D. Janda (eds..), The Handbook of Historical Linguistics, Oxford, 602–623.

328

Yael Maschler

Bybee, J. L. & P. J.Hopper (eds.) 2001 Frequency and the Emergence of Linguistic Structure, Amsterdam/Philadelphia. Bybee, J. & W. Pagliuca 1987 The evolution of future meaning. In: A. G. Ramat, O. Carruba, & G. Bernini, (eds.), Papers from the 7th International Conference on Historical Linguistics, Amsterdam, 109–122. Chafe, W. 1987 Cognitive constraints on information flow. In: R. S. Tomlin (ed.), Coherence and Grounding in Discourse, Amsterdam, 21–51. 1994 Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing, Chicago. Comrie, B. 1989 Language Universals and Linguistic Typology, 2nd edition, Chicago. Doron, E. 1982 On the syntax and semantics of resumptive pronouns. Texas Linguistic Forum 19, 1–48. Erteschik-Shir, N. 1992 Resumptive pronouns in islands. In: H. Goodluck & M. Rochemont (eds.), Island Constrains, Dordrecht, 89–108. Fassberg, S. 1996 haktiv sheha- lexinuy hazika (‘The written form sheha- of the relativizer’). mexkarim belashon (‘Studies in Language’) 7, 109–118. (In Hebrew). Ford, C. E. 1993 Grammar in Interaction, Cambridge. Fox, B. A. & S. A. Thompson 2007 Relative clauses in English conversation: Relativizers, frequency, and the notion of construction. Studies in Language 31, 293–326. Givón, T. 1973 Complex NPs, word order and resumptive pronouns in Hebrew. In: C. Corum, T. C. Smith-Stark & A. Weiser (eds.), You’ll take the high node and I’ll take the low node: Papers from the comparative syntax festival: The difference between main and subordinate clauses. A paravolume to the Papers from the ninth regional meeting of the Chicago Linguistic Society, Chicago, 135–146. Givón, T. 1975 On the role of perceptual clues in Hebrew reletivization. Afroasiatic Linguistics 2, 1–17. Goldenberg, G. 1995 Attribution in semitic languages. Langues Orientales Anciennes Philologie et Linguistique 5–6, 1–20. Reprinted in Goldenberg 1998, 46–65. 1998 Studies in Semitic Linguistics, Jerusalem. Haiman, J. 1994 Ritualization and the development of language. In: W. Pagliuca (ed.), Perspectives on Grammaticalization, Amsterdam, 3–28. Huehnergard, J. 2006 On etymology of the Hebrew relative she-. In: S. E. Fassberg & A. Hurvitz (eds.), Biblical Hebrew in Its Northwest Semitic Setting, Typological and Historical Perspectives, Jerusalem & Winona Lake, 103–125.

On the emergence of Hebrew adverbial connectives

329

Hopper, P. J. 1987 Emergent grammar. In: J. Aske, N. Beery, L. Michaelis, & H. Filip (eds.), Proceedings of the Thirteenth Annual Meeting of the Berkeley Linguistics Society 13, Berkeley, 139–157. 1988 Emergent grammar and the a priori grammar postulate. In: D. Tannen (ed.), Linguistics in Context: Connecting Observation and Understanding (Lectures from the 1985 LSA/TESOL and NEH Institutes), Norwood, 117–134. 1998 Emergent grammar. In: M. Tomasello (ed.), The New Psychology of Language. Cognitive and Functional Approaches to Language Structure, Mahwah, N.J., 155– 175. 2001 Grammatical constructions and their discourse origins: prototype or family resemblance? In: M. Pütz, S. Neimeier, & R. Dirven (eds.), Applied Cognitive Linguistics I: Theory and Language Acquisition, Berlin / New York, 109–129. 2004 The openness of grammatical constructions. 40th Annual Meeting of the Chicago Linguistic Society, Chicago, 153–175. Hopper, P. J. & E. C. Traugott 2003 Grammaticalization. Second edition, Cambridge. Keenan, E. & B. Comrie 1977 Noun phrase accessibility and universal grammar. Linguistic Inquiry 8, 63–99. Labov, W. 1972 Language in the Inner City, Philadelphia. Labov, W. & J. Waletzky 1967 Narrative analysis: Oral versions of personal experience. In J. Helm (ed.), Essays on the Verbal and Visual Arts. Seattle, WA. 12–44. Reprinted in The Journal of Narrative and Life History, 1997, 7, 3–39. Landau, R. 1975 mishpat hazika umishpat levay hatoxen lesugav ba'ivrit shel yameynu (‘The relative clause and the attributive content clause in Modern Hebrew’). bikoret Ufarshanut (‘Criticism and Interpretation’) 7–8, 132–136. Maschler, Y. 1998 rotse lishmoa keta? ‘wanna hear something weird/funny [lit. ‘a segment’]?’: The discourse markers segmenting Israeli Hebrew talk-in-interaction. In: A. H. Jucker & Y. Ziv (eds.), Discourse Markers: Descriptions and Theory, Amsterdam, 13–59. 2002 On the grammaticization of ke'ilu (‘like’, lit. ‘as if ’) in Hebrew talk-in-interaction. Language in Society 31, 243–276. 2009 Metalanguage in Interaction: Hebrew Discourse Markers, Amsterdam/Philadelphia. 2011 'al hithavutan shel tavniyot min hasiax: mishpaxat psukiyot hazika (‘On the emergence of constructions from discourse: The family of relative clauses’). le ˇsonenu 73: 167–207. Maschler, Y. & R. Estlein 2008 Stance-taking in Israeli Hebrew casual conversation via be'emet (‘really, actually, indeed’, lit. ‘in truth’). Discourse Studies 10, 3, 283–316. Ornan, O. 1979 hayidua umishpat hazika (‘Definiteness and the relative clause’). balshanut 'ivrit xofshit (‘Hebrew Computational, Formal, and Applied Linguistics’), 14. Pagliuca, W. (ed.) 1994 Perspectives on Grammaticalization, Amsterdam/Philadelphia.

330

Yael Maschler

Pawley, A. & F. Syder 1983 Two puzzles for linguistic theory: Nativelike selection and nativelike fluency. In: J. C. Richards & R. W. Schmidt (eds.), Language and Communication, London, 191–267. Peretz, Y. 1967 mishpat hazika ba'vrit lexol tekufoteha (‘The relative clause throughout the history of Hebrew’), Tel Aviv. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik 1985 A Comprehensive Grammar of the English Language, London. Ravid, D. 1977 mispar heybetim shel be'ayat seder hamarkivim be'ivrit yisra'elit modernit (‘Several aspects of the problem of element order in general Israeli Hebrew’). balshanut 'ivrit xofshit (‘Hebrew Computational, Formal, and Applied Linguistics’) 11, 1–45. Schegloff, E. A. 1996 Turn organization: One intersection of grammar and interaction. In: E. Ochs, E. A. Schegloff, & S. A. Thompson (eds.), Interaction and Grammar, Cambridge, 52–133. Schwarzchild, R. 1990 The Syntax of ['sr] in Biblical Hebrew with special reference to Qoheleth. Hebrew Studies 31, 7–40. Schwarzwald, O. 2006 ben hashana she'avra leshana haba'a: he'eder yidua vehashmatat bet (‘Between the last year and next year: lack of definiteness and deletion of bet' ’). xelkat Lashon 37–38, 165–174. Shaer, S. 2007 mishpetey zika basiax ha'ivri hayomyomi hadavur. (‘Relative Clauses in Casual Spoken Hebrew’). M.A. Thesis, Department of Hebrew Language, University of Haifa. Shlezinger, Y. 1999 tefutsat kinuyey hazika she- ve-'asher besefer kohelet (‘ The distribution of the relativizers she- and 'asher in the book of Ecclesiastes (Quoheleth)’). In: S. Sharvit (ed.), mexkarim balashon ha'ivrit ha'atika vehaxadasha (‘Studies in Ancient and Modern Hebrew’), Ramat Gan, 91–112. Tao, H. 2003 A usage-based approach to argument structure: ‘Remember’ and ‘forget’ in spoken English. International Journal of Corpus Linguistics 8, 75–95. Thompson, S. A. 2002 “Object complements” and conversation: Towards a realistic account. Studies in Language 26, 125–163. Thompson, S. A. & A. Mulac 1991 A quantitative perspective on the grammaticization of epistemic parentheticals in English. In: E. C. Traugott & B. Heine (eds.), Approaches to Grammaticalization, vol. II, Amsterdam, 313–329. Tottie, G. 1997 Relatively speaking: Relativizer marker usage in the British National Corpus. In: T. Nevaliainen & L. Kahlas-Tarkka (eds.), To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissan, Helsinki, 465–481.

On the emergence of Hebrew adverbial connectives

331

Tottie, G. & H. M. Lehmann, 1999 The realization of relative adverbials in present-day British English. In: H. Hasselgård & S. Oksefjell (eds.), Out of Corpora: Studies in Honour of Stig Johansson, Amsterdam, 137–157. Traugott, E. C. 1995a The role of the development of discourse markers in a theory of grammaticalization. 12th International Conference on Historical Linguistics. Manchester, August 1995 (version of 11/97). Retrieved 22. 1. 2004 from Stanford University. http://www.stanford.edu/~traugott/papers/subject2intersubject.pd f. Tsadka, Y. 1989 taxbir veshixbur. (‘Syntax’), Tel Aviv. Tsarfati, G. B. 1996 haktiv sheha- lexinuy hazika (‘The written form sheha- of the relativizer’). le ˇsonenu (‘Our Language’) 59, 267–268. Waltke, B. K. & M. O’Connor 1990 An Introduction to Biblical Hebrew Syntax, Winona Lake. Zewi, T. 2008 mishpetey toxen be'ivrit. (‘Content clauses in Hebrew’). le ˇsonenu (‘Our Language’) 70, 627–657.