Professional Documents
Culture Documents
net/publication/249237828
CITATIONS READS
282 2,421
1 author:
Andrea Golato
Texas State University
35 PUBLICATIONS 1,177 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Andrea Golato on 13 February 2016.
1. INTRODUCTION
Investigations of phenomena related to talk-in-interaction, or conversation or
discourse features of any kind, are typically based upon an analysis of data. An
element of such investigations which has received relatively little attention
concerns the nature of the data upon which these studies are based.
Presumably, studies of talk-in-interaction use databases that have been
compiled using similar methods of data collection that adequately capture
those features of talk that are to be investigated. Yet one simply has to browse
through any of the journals devoted to the study of discourse in order to see
that there are a variety of data collection procedures and methodological
frameworks available to students of discourse and pragmatics. Even when
limiting oneself to reviewing the body of literature on a single speech event,
for instance compliments, which will be the focus of this paper, it becomes
immediately obvious that they have been studied within various methodo-
logical frameworks, using dierent tools and methods of data collection
including: (a) discourse completion tasks and questionnaires (e.g. Barnlund
and Akari 1985; Yuan 1996); (b) recall protocols (e.g. Knapp, Hopper, and
ANDREA GOLATO 91
Bell 1984); (c) role play (e.g. Saito and Beecken 1997); (d) ®eld observation
(e.g. Wolfson and Manes 1980; Herbert and Straight 1989, and many others);
and (e) recordings of naturally occurring talk-in-interaction (Pomerantz 1978;
Wieland 1995; Golato 2002). This may not be surprising since the research on
compliments does not constitute one uni®ed ®eld of research. Compliments
and compliment responses have been studied in detail in eleven dierent
languages (see Golato (2002) for a detailed list), and in six varieties of English
alone, and they have been studied from a variety of dierent perspectives as
well (e.g. pragmatics, second language analysis, conversation analysis,
discourse analysis, speech act theory, and speech communication).
Despite coming from dierent research directions, almost all of these studies
include a discussion section reporting on what people are actually saying/
doing when they are responding to a compliment. Signi®cantly, most of the
studies do not claim to address intuitions concerning compliments, the value
that is placed on them in a given society, etc. Instead, they purport to describe
actual language use: the forms and formats of a compliment response, the
comparison of compliment responses in dierent languages, and occasionally
the applications of their ®ndings to language teaching. Here, then, the
question of selecting an appropriate method of data collection becomes
crucial. I would argue that many such studies of compliments and
compliment responses do notÐin fact, cannotÐdescribe actual language
use, simply because their analyses are not based upon data that illustrate
actual language use with sucient granularity. In other words, due to an
inappropriate choice of data collection procedure, the claims of many
published articles on compliment and compliment responses may need to
be attenuated.
In this article, within the context of compliment responses, I compare two
methods of data collection in detail, namely the recording of naturally
occurring data and discourse completion tasks (DCTs). The naturally occurring
data for this study were ®rst recorded and analysed. Subsequently, the
discourse completion tasks were designed to mirror those situations in the
natural data in which compliments occurred. In other words, all the situations
described in the DCT actually occurred in the video-taped data. The
questionnaire provided the contextual information and the verbatim compli-
ment that was given in the naturally occurring data. DCT participants were
asked to respond to the compliment. As I will demonstrate, there are striking
dierences between DCTs and naturally occurring data. My results suggest
caution in using DCTs if one's goal is to describe actual language use.
This article is by no means an attempt to present one data collection
procedure (recordings of naturally occurring data) as being the most suitable
for any and all research questions; nor is it the intent of this article to claim
that DCTs are of dubious value. It is simply argued that compared with data
collection instruments routinely used in conversation analytic (CA) studies,
DCTs are inappropriate for studying actual language use. Moreover, the article
will show that DCTs and other methods of data collection can provide
92 STUDYING COMPLIMENT RESPONSES
interesting, informative results, and thus are legitimate in their own rightÐ
even if they are better suited to addressing dierent research questions since
they measure phenomena other than (or additional to) actual language use.
American English, it becomes obvious that thanking is NOT the most frequent
response at all (Pomerantz 1978). It follows that if the intuitive notion of
`Americans say thank you' becomes the base line of comparison for a
sociocultural transfer study of compliments in American and Chinese (Liu
1995), interpreting the results of such a study becomes dicult.
Furthermore, interviews and discourse completion tasks do not show the
interactional facets of a speech event; for example, they do not capture
whether and how multi-turn sequences develop in order to ful®l a certain
speech function (Beebe and Cummings 1985). In addition, research has
shown that discourse completion tasks may be problematic in eliciting
appropriate data from speakers of non-western languages (Rose 1994; Rose
and Ono 1995; Hinkel 1997).
Despite these limitations, discourse completion tasks are widely used in the
®elds of pragmatics, intercultural communication, and second language
acquisition, mainly because their simplicity of use and high degree of control
over variables lead to easy replicability. In order to improve DCTs (particularly
in order to overcome insucient situational background information in the
DCT), researchers have called for enhancing the situational prompts by
providing more contextual and social clues (Billmyer and Varghese 2000).
Additionally, Yuan (1996, 2001) found that orally administering DCTs yields
more naturalistic speech features than the equivalent written DCT. In order to
make the situation more realistic, some researchers have included listener
responses in the prompt. However, it is yet not clear to which extent these
responses are bene®cial (in that they make the situation more realistic) or
harmful (in that they in¯uence the informants' responses; see Johnston et al.
(1998) and Hinkel (1997) for a detailed discussion).
they received or gave themselves. Obviously, carefully taken ®eld notes, most
widely used in ethnographic studies, allow for the collection of indispensable
contextual information (Kasper 2000). The biggest advantage of this method
of data collection is that it allows the investigator to collect a very large
database from a wide range of speakers and across various settings (Kasper
and Dahl 1991). Having a large enough database allows for statistical analysis,
which serves as support for claims made in the study.
However, there are several potential limitations in such studies. Since most
®eld workers do not audio-tape or video-tape any of their conversations, they
have to rely on their memory and observational skills. Trying to recall
linguistic data several hours after the event will lead to `data that is limited in
both quality and quantity' (Labov 1984: 30). Particularly, research has shown
that hedges, intensi®ers, conjunctions and modi®ers seem to be recalled
markedly less well than words tightly integrated into the clause (Lehrer 1989:
105). All such elements, however, are important in classifying compliment
responses into dierent categories, as are gestures and head movements (i.e.
nods, etc.). For example, if a ®eld worker observes someone giving a
compliment which is answered with a head nod and a smile, he or she
might classify the response as an acceptance of the compliment. If the ®eld
worker misses the non-verbal clues, however, the compliment response may
be classi®ed as a non-response. The analyst has no way of checking
observations collected this way, either for accuracy or for the presence of
discourse markers or gestures. Kasper (2000: 319) probably provides the best
summary of the immediate drawbacks of this method of data collection:
`There is thus a real danger that memorization and taking ®eld notes will
result in recording salient and expected (or particularly unexpected) facets of
the interaction, at the expense of less salient but perhaps decisive (often
indexical) material.'
A second drawback of many of the studies in which compliments were
collected ethnographically (i.e. via ®eld observations) is that they included
data from very dierent settings in their analyses. For example, Herbert (1990:
202) states: `The data for the present study were collected ethnographically by
students . . . , who were instructed to collect compliments and responses
within the student community, in places such as classrooms, dining hall, the
student union and so on.' There is a growing body of literature (see for
example the collection by Drew and Heritage (1992); here in particular
Levinson (1992) ) suggesting that task-related interactions in institutional
settings such as the classroom can be sequentially and functionally very
dierent from interactions in ordinary conversation (i.e. dinner table, dining
halls, etc). Task-related compliments from teachers to students thus may (and
often do) have very dierent functions from compliments given among
friends (Streeck 1979; Manes 1983). A similar study, although with much
younger children, showed that compliments in the classroom are often used
by the compliment giver in order to encourage a behaviour of the person who
is addressed and/or by those who can hear the compliment (i.e. the other
96 STUDYING COMPLIMENT RESPONSES
of data elicitation also have to contend with the problem facing researchers
working with discourse completion tasks: recall protocols also fail to yield the
interactional features related to a particular speech event, and thus lack a
correspondence to natural data. Another problem related to this and other
methodologies is that subjects are selected based on convenience, and not on
principles of random selection (Yuan 2001: 275) which is more easily
achieved with role plays or discourse completion questionnaires.
Having given this overview of advantages and disadvantages of four data
collection procedures routinely adopted for studying language use, I would
like to focus on a comparison of two of these methods in the following
sections, namely DCTs and video-taped naturally occurring data. DCTs are
frequently used in (interlanguage) pragmatics, second language acquisition
research, and to some degree in communication studies. Video-tapings and
audio-tapings are most frequently used in discourse and conversation
analysis. The following databased comparison of these two collection
procedures are meant to provide the reader with insights about the strengths
and weaknesses of the two procedures in gathering information about normal
discourse. To this end, I will ®rst present a taxonomy of compliment responses
based on an analysis of video-taped naturally occurring conversation (Section
3). In Section 4, I will then contrast this taxonomy with one obtained from a
discourse completion questionnaire. As stated above, this questionnaire was
designed based upon the compliment sequences presented in Section 3. In
other words, the situations and compliments on the DCT were drawn from
naturally occurring talk.
I. Acceptances:
1. Agreements/assessment of compl. A: aber heute abend hier war's schoÈn bei
A compliments B euch
but it was nice this evening here at your
place
B positively assesses compliment B: schoÈn.
that's nice.
2. Agreements/con®rmation
A compliments B A: ich sach ihr habt ja so en schoÈnes zwie-
belmuster hier,
I said you have such a nice onion pattern
here,
B con®rms compliment assertion B: joa:
ye:s
3. Assessment + agreement-pursuit
A compliments B A: uÈbrigens (.) das ¯eisch exzel[lent
by the way (.) the meat excel [lent
[
B uses same strength adj. + tag B: [super ne?
[super
right?
A compliments again A: exzellent
excellent
B accepts compliment B: joa.
yeah.
II. Rejections:
1. Disagreements
A compliments B A: robert hat es .hhh ( (sni) )eh- ich hm
w- wie sagt man
robert is .hhh ( (sni) ) eh- i hm h- how do
you say
das beste g'muÈt und du hoast es beste
feingefuÈhl
the most even-tempered and you are the
most sensitive
B disagrees with compl. assertion B: ach nee: komm,
oh no:: come on,
III. Solution types for 2 con¯icting con-
straints
1. Questions ± neutral stance
A compliments B A: m::m lecker,
m::m tasty,
B questions compliment assertion B: ja:a? ( (high pitch) )
yeah? ( (high pitch) )
A con®rms question/compliment A: uh uhm
uh uhm
2. Evaluation shift (downgrade)
A compliments B A: ahh da sin so viele schoÈne ecken bei
euch.
ahh there are so many nice places where you
live.
B scales down compliment assertion B: jo jo sin schoÈne ecken
yea yea nice places
ANDREA GOLATO 101
Table 1 (cont.):
Response type Example
B: [ja ja ( ) ( (non-committal) )
[yes yes ( ) ( (non-committal) )
[`yeah yeah' ( ) ( (non-committal) )
A: von dir behandeln zu lassen
by you treat to let
`you treated me'
B: ja:: is aber noch en weiter weg
ye::s is but still a long way
`ye::s but there's still a long way to go'
Germans can also be seen to de¯ect the credit away from themselves either to
an object or, as in data sample 8, to a third party:
(8) A: lecker ( (talking about barbecued meat) )
yummy
`yummy'
B: das ¯eisch hat sie gekauft ich hab's nur gegrillt
the meat has she bought I have it only barbecued
`she bought the meat I only barbecued it'
Since sie (`she') refers to B's wife, B appears to be de¯ecting the compliment to
her, or to at least be sharing it with her. Compliments may also be returned to
the ®rst speaker, as in the following data sample:
(9) A: schmeckt lecker ( (talking about cake) )
tastes yummy
`tastes yummy'
(1.5)8
B: der is aus dem kochbuch was du mir mal zu weihnachtn
it is out of the cookbook what you to me once for christmas
`it is from the cookbook that you once gave me for christmas'
geschenkt has
given have
Another German response type is a non-evaluative comment on, or
history of, the object/feature/talent that has been complimented. Often,
compliment recipients ®rst con®rm the assertion with ja (`yes') before
giving the comment or history. This is the case in the following example,
in which A compliments B on a dish B has made. After con®rming the
compliment, B explained how he made the salad dressing that he has
been complimented on:
(10) A: das klingt gut du. is ja dann lecker
that sounds good you. is m.p.9 then yummy
`hey that sounds good. is really yummy'
ANDREA GOLATO 105
B: ja:a. und dann aÈhm hab ich so:: .hh aÈhm diesmal mal so
ye:es. and then uhm have I like:: .hh uhm this time m.p. like
`ye:ah. and then uhm I like .hh uhm this time I mixed like
mirakel wip und majonese gemischt, . . .
miracle whip and mayonnaise mix, . . .
miracle whip and mayonnaise, . . . '
Compliments can also be reinterpreted as requests, which can lead to the
recipient's extending an oer as a next preferred action:
(11) A: lecker
yummy
`yummy'
B: da is noch mehr = du kanns gerne noch en stuÈck essen
there is still more = you can welcome another piece eat
`there is more' = `you are welcome to eat another piece'
Data sample 12 shows that compliment recipients can also choose to ignore a
compliment entirely. Without any verbal or non-verbal reaction, speaker B
simply turned her back on A and began to set the table:
(12) A: sieht aber lecker aus
looks m.p.10 yummy
`looks real yummy'
(30.0 +)
1. Acceptances
Appreciation token (`thank you0 ) 0 0 27 12.4
Agreements/assessment of compl. 4 8 33 15.2
Agreement / upgrade 0 0 4 1.8
Agreement with compliment 1 2 45 20.7
Agreements/con®rmation 20 40 31 14.3
Assessment + agreement-pursuit 6 12 1 0.5
II. Rejections
Disagreements 1 2 3 1.4
(Some turns consisted of several components, thus the percentage exceeds 100).
more frequently than in the CA data (see Appendix B). Speci®cally, while in
the naturally occurring data at most two response types were combined
(usually a con®rmation marker with one other response type and/or a second
assessment with a response pursuit marker), the DCT data yielded far more
combinations. Consider the following responses, all given by dierent
informants:
(13) Ja, stimmt. Das Service ist ein ErbstuÈck und mein ganzer Stolz.
`Yes, correct. The china is an heirloom and my pride and joy.'
108 STUDYING COMPLIMENT RESPONSES
(14) Ja, das ®nde ich auch, dass er gut eingekauft hat.
`Yes, I think so too, that he shopped well.'
(15) Ja, das ®nde ich auch. Ich habe mir viel MuÈhe gegeben bei der
`Yes, I think so too. I was very careful in selecting the
nature, DCTs are poorly suited: they are in a sense `noisy' tasks in that by
design, they also unavoidably yield o-line, metapragmatic data. Put
dierently, DCTs are better suited to the study of `what people think they
would say' than to the study of `what people actually do say' in a given speech
setting. If the subject of investigation concerns o-line or emic issues (Kasper
2000), for example people's beliefs or values with respect to culture, then
DCTs may very well be a good starting point, followed by detailed interviews
with subjects and informants. However, in the context of examining real
conversation, there is an inevitable diculty with DCT data which cannot be
resolvedÐthat is, not unless the data are compared with audio- and video-
taped data collected during actual speech events. It should be clear at this
point that DCTs do not cleanly and reliably inform us as to how talk-in-
interaction is organized and realized in natural settings (for a similar
conclusion, see Hinkel 1997). As Moerman (1974: 55) put it when discussing
language use, `folk beliefs have honorable status but they are not the same
intellectual object as scienti®c analysis'.
Possible questions with regard to CA methodology and, more generally, to
®ndings obtained using data from recordings of naturally occurring talk-in-
interaction, concern the generalizability of their results. As I will argue, such
questions are appropriate, but only when they are asked in their proper
context. For instance, such questions are legitimate when asked with respect
to ®ndings from language-related behavioural studies conducted using
experimental methods, for example psycholinguistic studies. By design, such
studies are interested not so much in the performance of the participants
being tested, but to whether the ®ndings obtained are reproducible; that is,
whether the behaviour observed in a controlled setting would resemble the
behaviour of comparable samples taken from the same or similar populations.
In such studies, behaviour is very often measured in experiments in which
participants (grouped according to, for instance, their ®rst language) are
instructed to perform various kinds of tasks which require a response (often a
button-push) upon the detection of a visually or auditorily presented stimulus
item (often a linguistic construct such as a morpheme, syllable, phoneme,
clausal boundary, etc.). Response latencies to successful stimulus detections
are analysed using inferential statistics (e.g. t-tests and analyses of variance).
Assuming appropriate methodology (including but not limited to: adequate
sample size, proper instrumentation, use of controls or blinding, behavioural
measures which yield interval-ratio data, clearly speci®ed procedures for the
identi®cation and handling of statistical outliers, etc.), such studies allow
psycholinguists to make inferences, or probability-based assertions, about the
eect of various participant- or stimulus item-related variables (here, a
participant's ®rst language, or the linguistic status of a stimulus item,
respectively) upon a task-related measure of performance (in the present
example, response latency). Put dierently, such studies allow researchers to
make general statements about the language processing abilities of a
population (e.g. the population of bilingual German±English speaking
112 STUDYING COMPLIMENT RESPONSES
adults) based upon a small sample of that population (e.g. 30 adult German±
English bilingual participants).
In contrast to psycholinguistic and other psychometrically oriented studies,
CA methodology is similar to other qualitative research agendas in that its
primary objective is not to make probability-based assertions. Rather, its goal
is to comprehend the (sequential) organization of talk-in-interaction (Schegl-
o 1993, Markee 2002). Accordingly, within CA methodology, data are
collected not through having participants perform experimental tasks under
controlled conditions, but through careful observation of interactants as they
freely engage in conversation within natural settings. As a result, and in
further contrast to psychometric studies, CA data consist of detailed records of
conversational interactions in the form of video and audio tapes and
transcripts thereof. Crucially, the `behaviours' noted in the transcripts bear
no resemblance to the task-related, interval-ratio data collected by
psycholinguists under laboratory conditions. On the contrary, in CA and in
other qualitatively oriented research, interactants are not given instructions
on, for instance, how they should interact, whom they should interact with,
for how long, what they should say, how they should phrase what they wish
to say, etc. Were they to be used, such manipulations would only serve to
ensure that the data collected did not represent naturally occurring talk-in-
interaction!13 Therefore, in experimental methodological terms, there are no
controlled conditions within CA, and no manipulations of participant- or
item-related variables. In fact, CA methodologists have yet to determine
what, if anything, constitutes a theoretically relevant `variable' in the
psychometric sense within instances of talk-in-interaction (for a detailed
discussion of this point, see Scheglo 1993). Additionally, within CA there
are no tasks to perform, and thus no measures of behaviour to be collected
and analysed using inferential statistical procedures. Consequently, with
respect to CA and other qualitatively oriented work, questions concerning
generalizability of results are entirely misplaced since they mistakenly
presuppose an approach to data collection and analysis used neither within
CA, nor within many other qualitative research methodologies.
A further clari®cation to be made regarding experimental methodology
and conversation analytic methodology concerns the treatment of outliers
(or `deviant cases', in CA parlance). While performing an experimental task,
a participant can be observed to generate responses which are demonstrably
dierent from those of others in the same group (i.e. which are statistical
outliers), or to commit errors (i.e. to generate responses in the form of misses
and false alarms, or fail to follow task instructions). While procedures in
experimental studies vary concerning the treatment of statistical outliers,
data representing errors is never analysed together with data representing
successful trials, but is instead removed prior to any statistical analyses. Error
removal is a methodologically sound practice: errors represent instances in
which a participant was not performing the task of interest to the researcher.
By contrast, within the context of naturally occurring talk-in-interaction, no
ANDREA GOLATO 113
ACKNOWLEDGEMENTS
I would like to express my thanks to Irene Koshik for inspiring me to write this article, and to
Brigitte Vlatten, Tobias Barske, and Carsten Wilmes for their help with data collection. I am
grateful to my husband, Peter Golato, for extensive comments and feedback. I owe much of the
discussion on `generalizability' to our numerous conversations which served to sharpen my
appreciation of issues related to psycholinguistics, experimental design, and data analysis. My
thanks also to the Language and Culture reading group at the University of Illinois for feedback on
an earlier version of this article. Finally, I would like to thank three anonymous reviewers and the
editors for their helpful and insightful comments. All remaining errors are, of course, my own.
114 STUDYING COMPLIMENT RESPONSES
APPENDIX A
Fragebogen
Anweisungen: Bitte reagieren Sie auf die folgenden Situationen. Bei einigen der
folgenden Situationen kann es sein, dass Sie mehrere Antworten/Reaktionen als
passend emp®nden. In diesem Fall schreiben Sie bitte alle Ihre Reaktionen auf die
vorgegebenen Zeilen. WaÈhrend Sie den Fragebogen ausfuÈllen, diskutieren Sie ihn bitte
nicht mit anderen Personen.
1. Sie haben Freunde zu einem gemuÈtlichen Abendessen eingeladen. Bei der
Verabschiedung sagt einer Ihrer Freunde: ,,Heute abend hier war's schoÈn bei euch!''
Sie antworten darauf:
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
2. Sie sind gerade in ein Eigenheim umgezogen. Ihr Haus liegt in einer wunderbaren
Gegend.
Eine Verwandte sagt zu Ihnen ,,Ahh, da sind so viele schoÈne Ecken bei euch!''
Sie antworten darauf:
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
3. Eine Bekannte hilft Ihnen beim AufraÈumen nach einem Fest. Sie stellt unbenutztes
Geschirr in den Schrank und sieht dort ein altes Kaeeservice und sagt: ,,Ihr habt ja so
ein schoÈnes Zwiebelmuster hier!''
Sie antworten darauf:
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
4. Sie und Ihr Partner/ Ihre Partnerin haben Freunde zum Grillen eingeladen. Als alle
anfangen das Fleisch zu essen, sagt ein Freund: ,,UÈbrigens, das FleischÐexzellent!''
Sie antworten darauf:
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
5. Freunde und Verwandte sind zum Geburtstagskaee bei Ihnen. Alle essen Kuchen
und trinken Kaee und jemand sagt uÈber den Kuchen, den Sie gebacken haben:
,,Schmeckt lecker!''
Sie antworten darauf:
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
6. Sie haben Freunde zum Mittagessen eingeladen. Sie sind dabei, das Essen
vorzubereiten bzw. aufzutragen und eine Freundin sagt: ,,Sieht aber lecker aus!''
ANDREA GOLATO 115
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
6. Some friends are over at your house for lunch. As you are bringing it to the table, a
female friend says: `This looks great!'
You answer:
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
7. Friends are over at your house for a barbecue. The meat is especially tender and
juicy, and someone says: `Yum, yummy!'
You answer:
a) _________________________________ b) _________________________________
c) _________________________________ d) _________________________________
APPENDIX B
NOTES
1 I am grateful to one of the editors for this 7 Providing the history of the object of the
observation. compliment (i.e. where or how one
2 For a clear example of this principle, see obtained it) would be an example of a
Silverman's account (Silverman 1998: 110± compliment response that is neutral in its
11) of Scheglo's work (Scheglo 1972) on stance.
telephone openings. Alternatively, see 8 B has her mouth full and is chewing the
Egbert's study on the role of eye-gaze and cake.
the repair initiator bitte? (`pardon?') (Egbert 9 m.p. = modal particle.
1996). Egbert found that Germans use this 10 m.p. = modal particle.
particular general repair initiator when 11 Approximately 6 months to one year had
there was no eye-contact between the elapsed between the original recording and
participants during the production of the the distribution of the questionnaire, thus it
repair construction. The production of the seems unlikely that subjects remembered
repair initiator re-establishes repair the speci®cs of the situation in which they
between the participants. In her study, the were given the original compliment.
single instance in which the repair 12 Interestingly, in all of these studies, a longer
sequence between a non-native speaker of response is always depicted as a better
German and a German native speaker was response, without any critical re¯ection of
dierent from all other instances in her why that should be so.
collection led Egbert to consider the role of 13 There is potentially a more serious problem
eye-gaze (Egbert, personal commun- related to adopting an experimental meth-
ication). odological approach to analysing transcript
3 m.p. = modal particle. or DCT data. If such data are in the form of
4 For a comparison of German and American frequencies (i.e. counts or tallies of instances
English compliment responses, see Golato of behaviour; e.g. the number of compli-
(2002). ments given of a particular turn shape, the
5 Note the dierence between data segments number of instances in which someone
3 and 5: in data segment 5, speaker B gives responded `das Fleisch, exzellent', etc.),
an assessment of the compliment assertion then statistically speaking they lack true
while in data segment 3, speaker B gives an numerical status. One of the assumptions
assessment of the compliment. For a more of inferential statistical analyses (e.g. t-tests
detailed discussion of this dierence, see and analyses of variance) is that the data
Golato (2002). being analysed are true numbers of an
6 I would like to thank the editors for interval nature (e.g. ounces, seconds,
pointing out the following alternative read- temperature, voltage, etc.). Analysing raw
ing of this compliment response: the re- frequency data with a statistical procedure
sponse can be seen as minimizing the designed for interval data is methodo-
degree of imposition of the compliment by logically unsound, since it leads to uninter-
transforming the compliment into a state- pretable (not to mention ungeneralizable)
ment of fact, to which the recipient simply results. For a clear discussion of these and
acquiesces with another factual statement. related issues with speci®c reference to
One could say that the response is an research in applied linguistics, see Hatch
avoidance of self-praise by changing the and Lazaraton 1994.
nature of the original act through its
perlocutionary eect.
ANDREA GOLATO 119
REFERENCES
Anderson, J. R. 1974. `Verbatim and proposi- between American English and Chinese
tional representation of sentences in speakers.' Journal of Pragmatics 20/1: 49±75.
immediate and long-term memory.' Journal Drew, P. and J. Heritage. (eds) 1992. Talk at
of Verbal Learning and Verbal Behavior 13: Work. Cambridge: Cambridge University
149±62. Press.
Anderson, J. R. and G. Bower. 1974. `A Duranti, A. 1997. Linguistic Anthropology. Cam-
propositional theory of recognition bridge: Cambridge University Press.
memory.' Memory and Cognition 2: 406±12. Egbert, M. M. 1996. `Context-sensitivity in
Aston, G. 1995. `Say ``Thank you'': Some conversation: Eye gaze and the German
pragmatic constraints in conversational clos- repair initiator ``bitte''?' Language in Society
ings.' Applied Linguistics 16/1: 57±86. 25/4: 587±612.
Atkinson, J. M. and J. Heritage. (eds) 1984. Golato, A. 2000. `Pragmatic interference and
Structures of Social Action. Studies in Conversa- compliments.' Paper presented at the AAAL
tion Analysis. Cambridge: Cambridge Univer- Annual Convention: Crossing Boundaries, in
sity Press. Vancouver, British Columbia, 11±14 March
Auer, P. and S. Uhmann. 1982. `Aspekte der 2000.
konversationellen Organisation von Bewer- Golato, A. 2002. `German compliment
tungen [Aspects of the conversational organ- responses.' Journal of Pragmatics 34: 547±71.
isation of assessments].' Deutsche Sprache 10: Graesser, A. C. and G. Mandler. 1975.
1±32. `Recognition memory for the meaning and
Barnlund, D. C. and S. Araki. 1985. `Inter- surface structure of sentences.' Journal of
cultural encounters: the management of Experimental Psychology: Human Learning and
compliments by Japanese and Americans.' Memory 104: 238±48.
Journal of Cross-Cultural Psychology 16/1: 9±26. Gumperz, J. J. 1982. Discourse Strategies. Cam-
Beebe, L. M. and M. C. Cummings. 1985. bridge: Cambridge University Press.
`Speech act performance: A function of the Hanson, V. L. and U. Bellugi. 1982. `On the
data collection procedure?' Paper presented role of sign order and morphological struc-
at the TESOL Convention, New York. ture in memory for American Sign Language
Beebe, L. M. and M. C. Cummings. 1996. sentences.' Journal of Verbal Learning and
`Natural speech act data versus written Verbal Behavior 21: 621±33.
questionnaire data: How data collection Hartford, B. and K. Bardovi-Harlig. 1992.
method affects speech act performance' in `Experimental and observational data in the
S. M. Gass and N. Joyce (eds): Speech Act study of interlanguage pragmatics' in
Across Cultures: Challenges to Communication in L. Bouton and Y. Kachru (eds): Pragmatics
a Second Language. Berlin: Mouton de and Language Learning, Monograph 3. Urbana-
Gruyter. pp. 65±86. Champaign, IL: DEIL. 33±52.
Billmyer, K. and M. Varghese. 2000. `Investi- Hatch, E. M. and A. Lazaraton. 1994. The
gating instrument-based pragmatic vari- Research Manual: Design and Statistics for
ability: Effects of enhancing discourse Applied Linguistics. Boston: Heinle & Heinle.
completion tests.' Applied Linguistics 21/4: Herbert, R. K. 1990. `Sex-based differences in
517±52. compliment behavior.' Language in Society 19:
Blum-Kulka, S., J. House, and G. Kasper. 201±24.
1989. Cross-Cultural Pragmatics: Requests and Herbert, R. K. and H. S. Straight. 1989.
Apologies. Norwood, NJ: Ablex. `Compliment-rejection versus compliment-
Bodman, J. and M. Eisenstein. 1988. `May avoidance: listener-based versus speaker-
God increase your bounty: The expression of based pragmatic strategies.' Language and
gratitude in English by native and non- Communication 9/1: 35±47.
native speakers.' Cross Currents 15/1: 1±21. Heritage, J. 1984. Garfinkel and Ethnomethodol-
Chen, R. 1993. `Responding to compliments. A ogy. Cambridge: Polity Press, in association
contrastive study of politeness strategies with Basil Blackwell, Oxford.
120 STUDYING COMPLIMENT RESPONSES