AUTHOR: Ruhlemann, Christoph
TITLE: Conversation in Context
SUBTITLE: A Corpus-Driven Approach
SERIES: Corpus and Discourse
PUBLISHER: Continuum International Publishing Group Ltd
YEAR: 2007
David Schlangen, Department of Linguistics, University of Potsdam, Germany
SUMMARY
This interesting and readable book offers both detailed, focused investigations
of very specific conversation phenomena (as you would expect from a revised PhD
dissertation), and a comprehensive literature overview, and the beginnings of a
general framework for studying conversation phenomena in general. As the
subtitle ''A Corpus-driven Approach'' indicates, the book comes from the tradition
of and is mainly aimed at corpus linguistics, but it should be of interest to
anyone who studies conversation, including computational linguists, philosophers
of language, sociologists, and cognitive scientists.
The goals of the book are succinctly stated in the introduction (p.2): 1. to
provide an overview of characteristic functions of British conversation; 2. to
describe neglected features of British conversation in detail; 3. to explore
social differentiation in the use of selected features of British conversation;
4. to test the hypothesis that a situation-based description of conversation can
show how conversational language is adapted to certain needs arising from
specific types of constraints on speakers in conversational situations
('adaptedness hypothesis'). Of these aims, the first is approached through a
review of extant literature, while 2-4 are dealt with by looking at
conversational data from the British National Corpus (BNC). Goals 2 and 3 are,
as it were, in the service of 4, the central aim of the book. The 'adaptedness
hypothesis' serves as the conceptual center of the book, towards which the
detailed studies of particular phenomena are oriented.
The first three chapters of the book lay the foundation for the empirical
studies in chapters 4 to 8. Chapter 1 opens the book with an attempt to bring
out the distinctive features of 'conversation', as compared to other arenas of
language use. The route that is taken here is to describe conversation as a
register, following Systemic Functional Linguistics (e.g., Halliday 1978). The
quotes from Halliday (1978) given by Rühlemann (p. 5) already stress the role of
situatedness in the view of conversation as a register: ''[register is] 'variety
according to use' (p.35) and [...] 'what a person is speaking, determined by
what he is doing at the time' (p.110)''. Thus, it's both features of the language
and of the situation that distinguish registers: ''In short, registers may be
defined by the distinctive features of both situation and language.'' (p.7).
Rühlemann follows a large number of researchers (e.g., Clark, Schegloff) in
assuming that conversation has a privileged status among situations of language
use; in the terms used in the book, it is a 'core register'. After providing
this basic definition, the chapter closes with a discussion of the state of the
art in corpus linguistic work on conversation.
The short Chapter 2 briefly describes the empirical basis for the studies - the
BNC - and the methods that are used. The parts of the BNC that are used in this
book are subcorpus C, the demographically sampled part of the spoken language
data in the BNC, and, for contrast, CG, the 'context-governed' part of spoken
language data in BNC, as well as occasionally the written part. As Rühlemann
writes (p.24): ''given the relatively clear delineation of the texts in C as
conversation as well as the enormous size of C, generalizations of the
distinctive features in C to the register of British conversation do seem
permissible.'' As for methodology, as Rühlemann is interested in detecting
distinctive features of the register 'conversation', the strategy is to compare
frequency of occurrence across genres: ''to be distinctive of a register it [the
feature that is investigated, D.S.] should be clearly more frequent in this
register than in other registers'' (p.28). Once distinctness is established,
association patterns are established (collocation and colligation), and
discourse and sociological factors are explored. This methodology is followed in
the studies in the later chapters. (See below for a discussion of the
generalizability claim and the methodology.)
Chapter 3 introduces the conceptual core of the book, the ''situational framework
for conversation'' to which the conversational features studied in the later
chapters are related functionally. Rühlemann identifies five factors ''jointly
determining the conversational situation'' (p. 35). First, in a conversational
situation there will be a shared context, which for Rühlemann comprises
non-verbal context (of given utterances), social context and situational
context. (The linguistic context of a given utterance does not seem to fall
under 'shared context' for Rühlemann; the category seems to be reserved for
context that is not preserved in a transcript.) Second, the conversational
situation is marked by the necessity for co-construction - the conversation is
constructed interactively. This factor shows up in the orientation of
participants towards sequential organization of utterances (i.e., the utterances
always also are replies), and in the rotation of roles (like speaker and
hearer). Third, conversation is ''fully 'process''' (p. 43), that is, inherently
subject to real-time processing constraints. ''Conversationalists are fully
exposed to the pressures of planning and processing in real time'' (p. 43). (To
anticipate the discussion below, to me it is one of the major virtues of the
book to put equal emphasis on this aspect of language use, which is often
ignored in linguistics.) Fourth, the requirement of co-construction in real time
leads to a need for discourse management to maintain coherence. (To which for
Rühlemann also structural coherence with respect to role assignment / turn
taking seems to belong.) Fifth, there is a requirement for relation management,
especially so in conversation: ''we converse because we want to establish bond of
communion with one another'' (p48). (As becomes clear at this point, Rühlemann
reserves the term 'conversation' for interactions with no task-oriented
component at all.)
As Rühlemann himself is quick to point out, most distinctive features of
language use in conversation are the result of more than one (if not all) of
these factors working together, or, more precisely, of the language user
attending to the situational requirements. Nevertheless, the subsequent chapters
are organized around individual factors, each providing an overview of phenomena
that can be related to a single dominant situational factor, and giving one or
more analyses of such phenomena as they occur in the BNC data.
For reasons of space, I will only briefly review one of these chapters, the one
on shared context phenomena (Chapter 4). The other chapters are devoted to
co-construction phenomena (Chapter 5, with example analyses of co-constructed
sentence relatives), discourse management phenomena (Chapter 6, analysis of
''like'' as discourse marker), real-time processing phenomena (Chapter 7, analysis
of grammatical reduction in 'I says'), and relation management phenomena
(Chapter 8, third person singular 'don't').
The overview part of Chapter 4 (on shared context phenomena) looks at a
phenomenon that provides the clearest, most explicit link in a corpus between
language and its situational context, namely deixis. It begins with a quick
review of terminology on deixis, settling on the notion of deictic centers to
guide the presentation. Person deixis is then further explored using the BNC
data (as this shows, the separation in each chapter between overview and case
study is not always strict), and found to be much more prevalent than in
writing: ''the fact that person deictics 'I' and 'you' are the two most frequent
words in conversation while they occupy two-digit ranks in writing clearly
suggest that person deixis is much less important in writing than in
conversation, where its centrality cannot be overstated. [...] [A] concomitant
of turn-taking is the deictic-centre switch and, hence, the need in each new
turn to re-establish 'the rest of the deictic systems' (Levinson 1983: 68), or
deictic context, including, first and foremost, the person-deictics 'I' and
'You''' (p.75). The case study in the chapter finally looks at laughter as
non-verbal context, finding that ''the typical 'laugher' in British
conversation, hence, seems to be a young, female, white-collar worker'' (p. 86).
(An observation that however is not further analyzed.)
The book closes with a brief Conclusions chapter that, again, relates the
observations to the central thesis of the book, the 'adaptedness hypothesis'.
EVALUATION
While offering a wealth of information, the book sometimes appears a bit
breathless in its attempt to be both a comprehensive literature review and a
comprehensive corpus study. Also, at some points, one could perhaps wish for a
more thorough analysis, with slightly more sophisticated statistical methods.
Typically, only frequencies (raw or normalized) are given, and conclusions are
drawn on the basis of comparisons of such frequencies in subcorpora, without
performing statistical tests of significance. Some analyses use rather small
data bases. E.g., some of the dialectal subcorpora contain as little as 5000
utterances; one would assume that idiosyncrasies of even only one speaker can
influence the relative frequencies within such subcorpora drastically. It would
also have been interesting to see more of a discussion of the question of how
one can generalize from the observations in the corpus to statements about
English conversation in general. (See for example (Evert 2006) for a discussion
of typical problems in corpus linguistic analyses.)
Coming to the study of conversation from a different background (not corpus but
computational linguistics and cognitive science), it was interesting for me to
see how large the overlap in interest is - and how small the overlap in common
literature. I certainly took away from the book useful pointers to research on
conversation that I was unaware of. But similarly, I was surprised to see no
mention, for example, of the work of Herb Clark (e.g., Clark 1996), who has been
developing for many years now a rather similar framework for describing the
conversational situation, which however is grounded in deeper, more principled
considerations of cooperation and coordination. Also, there's a rich literature
on experimental investigations into the influence of contextual factors on the
shape of conversations (if understood as including task-oriented dialogue); the
MapTask corpus work (e.g., Anderson et al. 1991) shall suffice as one example here.
But all this aside, this book offers a wealth of interesting and useful
observations, analyses, references, and methodological inspirations and is to be
recommended to everyone interested in (data-oriented) analysis of conversation.
REFERENCES
Anderson, A., Bader, M., Bard, E., Boyle, E., Doherty, G. M., Garrod, S., Isard,
S., Kowtko, J., McAllister, J., Miller, J., Sotillo, C., Thompson, H. S. and
Weinert, R. (1991). The HCRC Map Task Corpus. _Language and Speech_, 34, pp.
351-366.
Clark, Herb (1996). _Using Language_. Cambridge, UK: Cambridge University Press.
Evert, Stefan (2006). How random is a corpus? The library metaphor. _Zeitschrift
für Anglistik und Amerikanistik_, 54(2), 177 - 190.
Halliday, M.A.K. (1978). _Language as Social Semiotic: The Social Interpretation
of Language and Meaning_. London: Edward Arnold.
Levinson, Stephen (1983). _Pragmaticsh_. Cambridge, UK: Cambridge University Press.
ABOUT THE REVIEWER
David Schlangen is a Senior Researcher in Computational Linguistics at the
University of Potsdam. He has worked on error handling in human/human and
human/machine dialogue, and is now mainly interested in modeling natural timing
of conversational actions in spoken dialogue systems.
|
