esrc.html
Kinship and language: a computer-aided study of social deixis
in conversation. R000 23 3311
Summary
of Aims
The proposed research has both descriptive and theoretical objectives
(unchanged from original application):
1) Theoretical.
It is hoped to develop a method of studying kinship employing theory
and techniques deriving from pragmatics and conversation analysis,
instead of more conventional anthropological methods. It will then
be possible by examining Dr Zeitlyn's Mambila data to demonstrate
the power of analysing kin terms as part of a set which also includes
proper names, pronouns and titles. This will constitute a significant
re-casting of the anthropological theory of kinship.
The validity of this approach to the study of kinship will be further
tested by the computer-assisted study of linguistic corpora from a
range of other societies to show that the method is capable of far
wider application. It will also demonstrate the benefits of this
style of analysis to a wider audience than may be expected to read
studies of a single African group.
2) Descriptive.
The descriptive objective of the project is to undertake an analysis
of Mambila kinship, both demographic and linguistic. Much relevant
information has already been collected and is currently recorded in
unpublished fieldnotes. This data will now be collated into a form
which may be checked and completed in the course of further fieldwork.
The fieldwork will also produce new data for analysis in accordance
with of the theoretical position outlined below.
A wide range of data, particularly in the form of transcripts of naturally
occurring conversation, will be made available for other researchers.
3. Significant Achievements
1 The establishment of a systematic cross-cultural approach to the
use of social deictic expressions in naturally occurring conversation
2 The production and archiving of high quality natural language data
from an 'obscure' African language providing data for hard tests of
'universal' theories developed only on well documented European languages.
Such languages cannot be used legitimately as the sole basis for
universal claims.
3 The development and testing of a coding scheme allowing comparisons
to be made between strategies of person reference from different languages
and cultures
4 The demonstration that anthropological research on kinship terms
can be clearly related to empirical evidence of language in use.
5 The development of systematic tests for identifying different speakers
and transisitions between topics on the basis of analysis of transcript
alone. Although still under developement this stands to have important
implications for machine understanding of texts.
4 Dissemination
Publications that have already resulted from the project are listed
below, others are in preparation.
- Further papers in draft (at point of submission to major journals)
- Introducing Mambila Social Deixis
- Summing Up Kin talk
Asking about people, not about kin terms: an essay on method and interpretation
The main data has already been archived as part of the Child Language
Interchange (CHILDES) project and is being distributed by them (with
full acknowledgement of the ESRC) both online and on CD.
It is planned to present some of the results at major international
conferences over the next two years as the final results are fully
tested and completed.
Some of the theoretical background to the project has been used in
briefings given to Nokia Mobile Phones, UK.
5 Nominated Publications
- Wilson, A.J. & D. Zeitlyn. 1995. The Distribution of Person Referring Terms in Natural
- Conversation. Research on Language and Social Interaction
28(1), 61-92.
- Zeitlyn, D. 1993. Reconstructing Kinship or the pragmatics of kin talk. Man 28(2),
- 199-224.
7 Major Difficulties
The systematic
coding of the Mambila data took somewhat longer than anticipated.
We decided to concentrate on producing less very-high-quality transcript
rather than more lower-quality data, and succeeded in coding, and
double checking with the original speakers a conversation containing
some 1302 utterances which is sufficient to allow statistically significance
results to be achieved. In the light of the current ESRC funded research
of Dr Bruce Connell no dictionary or lexicon was produced as had been
originally intended but the source data is being collaboratively worked
on by Zeitlyn and Connell as part of Connell's continuing research
on the Mambila language.
8 Other Issues and Unexpected Outcomes
In the course of developing the coding scheme we realised how the
theory could be further expanded and tested by coding extra data from
which some of the social context had been removed. If this were successful
it would both vindicate the theory and point to ways in which the
methods could have application to the automatic processing of speech.
Sadly the ESRC declined to fund the extra work. This led Mr Wilson
to leave to the project to work as a marketing consultant. He was
the research assistant who could have undertaken this work speedily
and efficiently without further (expensive) training had his contract
been extended. Zeitlyn has concentrated in finishing the main project
but hopes, if further funding can be obtained to pursue this extension
of the work.
{{Page X}}
Summary of Research Results
The aim of this project was to study the usage of person referring
expressions in corpora of naturally occurring language in order to
provide substantive evidence for the theoretical constructions about
the meaning of kinship terms that have characterised a century of
anthropological theorising. The methodological inspiration for our
approach may be found in ethnomethodology and in particular in the
fields of conversational analysis and pragmatics.
To take an example from home: speakers of English ³know²
that children use kin terms to address their parents and receive names.
Like other conversational maxims this may not always be the case
but variants from the norm will be marked. For example, a child may
shift pattern in imitation of a political act on the part of an elder
sibling. The irony is that the intention to be like one's big brother
commits one to the political implications of his acts. One of the
strengths of anthropology is its sensitive analysis of such unintended
consequences.
By coding transcripts of 'actual' (unprompted) conversation we were
able to study systematically the choices made by speakers between
the many different possibilities open to them: to refer to X as e.g.
Cousin, Cousin Jean, Jean, Shorty etc. etc. The coding scheme allows
us to correlate the choice of term with the relationship between the
speakers (and the referent) as well as the sort of speech event occuring.
The application of this approach to English and Mambila language data
has permitted some systematic comparative analyses (which are still
underway) and has generated a significant data set for the use of
other researchers.
{{Page X}}Full Report of Research Activities and Results
Background
This project concerns the distribution of kin terms and other referring
expressions as they occur in natural conversation. Some cross cultural
comparisons are being made between data drawn from American sources
and from my research with the Mambila in Cameroon. Some of the successes
and difficulties that have been encountered are outlined below.
Objectives
The original proposal included the following objectives, which were
satisfactorily achieved.
1) Theoretical.
It is hoped to develop a method of studying kinship employing theory
and techniques deriving from pragmatics and conversation analysis,
instead of more conventional anthropological methods. It will then
be possible by examining Dr Zeitlyn's Mambila data to demonstrate
the power of analysing kin terms as part of a set which also includes
proper names, pronouns and titles. This will constitute a significant
re-casting of the anthropological theory of kinship.
The validity of this approach to the study of kinship will be further
tested by the computer-assisted study of linguistic corpora from a
range of other societies to show that the method is capable of far
wider application. It will also demonstrate the benefits of this
style of analysis to a wider audience than may be expected to read
studies of a single African group.
2) Descriptive.
The descriptive objective of the project is to undertake an analysis
of Mambila kinship, both demographic and linguistic. Much relevant
information has already been collected and is currently recorded in
unpublished fieldnotes. This data will now be collated into a form
which may be checked and completed in the course of further fieldwork.
The fieldwork will also produce new data for analysis in accordance
with of the theoretical position developed as part of the project.
A wide range of data, particularly in the form of transcripts of naturally
occurring conversation, will be made available for other researchers.
Methods
The
transcription procedures deserve some mention. Soon after making
the recording I went through it with an informant in the village in
Cameroon. At this stage as well as making some contextual notes we
made a second recording - the informant repeated each utterance into
a second tape recorder, speaking slowly and clearly. To do this he
used both his understanding as a native speaker and the fact that
he was an actor in the conversation to understand parts of the recording
that were (and remain) extremely indistinct to my foreign ears. In
the course of making this second recording he explained various idioms
and vocabulary items that were new to me. I transcribed the second
recording in the UK and then returned to. I checked this transcript
against the original recording since the slow repetitions often corrected
infelicities of grammar or removed elisions resulting from fast speech.
During a subsequent fieldtrip I checked some passages whose transcription
was uncertain and prepared a free English translation. The transcript
was then coded in UK (following a scheme developed in a pilot study
(described in Wilson and Zeitlyn 1995)). In the course of the coding
the English translation was revised so that the use of pronouns and
names was parallel to their use in the Mambila original - although
there are obvious problems in this such as a gender neutral third
person and some (rare) compound pronouns. The coding process turned
up some further problems which were resolved during a further fieldtrip
in May 1994. The result is a transcript that is of unparalleled quality
for its type. This is not to say that it is not theory laden (Ochs
1979) and inevitably it could be improved, in particular the absence
of a visual channel combined with the free passage of children (and
adults) in and out of the house makes it uncertain just who the non-participating
audience is at any one time. In addition there are often the voices
of children at play in the background. Most of the time these have
proved too indistinct to be able to transcribe. Almost any recording
one makes in Somié will have the voices of children playing
somewhere in the background!
Results
The research has gone broadly according to schedule although the timing
of field research in Cameroon was altered due to political disturbances
in that country. However, the administration are familiar with me
and I have had no problem obtaining the necessary research permits.
The main delay has occurred in the preparation of transcripts which
proved even more time consuming than originally estimated. Rather
than produce more low quality material it was decided to produce less
material of high quality, and this has now been done. The resulting
transcript is of extremely high quality, having been checked in the
field in three successive stages of transcription and translation.
It represents a unique source of data that will be of great value
to those studying African languages. Since the demands of time in
the field were greater than originally planned the extra work on the
dictionary and comparative study of kinship terminologies has not
been undertaken. A draft dictionary exists but this needs further
work before it is in a form suitable for circulation. This will not
occur within the duration of the project. Although regrettable this
was never one of the major priorities of the original proposal. It
was decided that it was best to concentrate on the core research in
order to be able to complete that on time. The data is being shared
with Dr Connell, a linguist working on the Mambila language with ESRC
funding.
In the initial stages of the project a comprehensive literature search
was undertaken. This covered several disciplines ranging from psychology
through anthropology to conversation analysis. Following this a pilot
project was undertaken to establish an adequate coding protocol.
At this stage of research four papers were written by Dr Zeitlyn and
Mr Andrew Wilson, the RA1B employed as part of this project. Zeitlyn
(1993) presents a theoretical overview of the project. Wilson 
(1995)
is a survey of literature research techniques for social anthropologists.
Wilson, & Zeitlyn (1994) is a detailed
discussion of the work of William Stiles and his attempt to code spoken
discourse. Despite some reservations about Stiles' theoretical position
we feel that his taxonomy has practical utility, and have used it
as part of our own coding scheme. Zeitlyn and Wilson 
(forthcoming)
is a discussion of a set of computational tools (CLAN) designed for
the study of talk. Originally developed for developmental psychologists
we outline how it can be used as an ethnographic tool. It provides
us with the means to analyse our own material.
Having developed the coding scheme and tested for inter-coder agreement
scores, we then applied it to two American conversations, using data
made available by American colleagues. These conversations are about
forty-five minutes in length and were selected to compare with the
Mambila data, all the conversations occurring as food was prepared
and then eaten. The analysis of the first of the conversations, along
with a comprehensive discussion of the methodology of the project
are presented in Wilson, & Zeitlyn (1995). The scheme has now
been applied to a corpus consisting of a conversation in a similar
setting but spoken by a Mambila family. The results of this analysis,
and the comparison with the American data are being computed at this
moment.
Outputs
- Wilson, A.J. 1995. A field-guide to bibliographic research. Journal of the Anthropological
- Society of Oxford
25(2), 179-83.
- Wilson, A. J., & Zeitlyn, D. 1994. Speech acts and Stiles. Linguistics and Education
- 6(1), 91-98.
- Wilson, A.J. & D. Zeitlyn. 1995. The Distribution of Person Referring Terms in Natural
- Conversation. Research on Language and Social Interaction
28(1), 61-92.
- Zeitlyn, D. 1993. Reconstructing Kinship or the pragmatics of kin talk. Man 28(2),
- 199-224.
- Zeitlyn, D., & Wilson, A. J. forthcoming 1995. The Childes Project: an anthropological
- resource. CAM. The Cultural Anthropology Methods Journal
.
It is planned to present some of the results at major international
conferences over the next two years as the final results are fully
tested and completed. A world wide web site (The Virtual Institute
of Mambila Studies) is also planned. This will be hosted at University
of Kent, Canterbury, and will contain reports (such as this) and bibliographic
data as well as links to digitised sound recordings that I have already
made available on WWW.
Some of the theoretical background to the project has been used in
briefings given to Nokia Mobile Phones, UK.
Impacts
This project has demonstrated by example that systematic research
using data from naturally occurring conversations can yield results
of central importance for the development of anthropological theories
of kinship, as well as being relevant to those working in applied
arenas such as communication technology. It should be stressed that
this was an unexpected benefit from the research which would not have
occurred had it been constrained by local and applied foci from the
outset.
Future research priorities
It is hoped that funding will be forthcoming for the further development
and testing of the methods that have been developed in this project.
In particular we would like to explore hard testing of the theory
by coding partially and fully anonymised transcripts - from which
information about speakers has been removed. If the intuitions of
native speakers lead them to segment a transcript correctly and if
our analysis can replicate this then a (very) hard test will have
been performed. We have so far not succeeded in raising the funding
for this.
Another way in which the project may be extended is to consider analysing
other languages to make the cross-cultural test more exacting. An
obvious case to consider would be Japanese which is notorious for
the scarcity of pronominal usages, and the complexity of its system
of honorifics.
General report on results obtained to date.
Coding and counting. The broad brush and the fine grain.
Counting words may produce some interesting results particularly if
accompanied by other information about the sociology of the speakers
and the speech act in question. Take mother-in-law avoidance as a
case in point. What people notice and orient to are some regularities
in the styles of talk and some of these regularities are fairly crude
in kind. For instance, different words are used, or NOT used, when
talking to or about, mothers-in-law. What we gloss as mother-in-law
avoidance is, inter alia, that a son or daughter-in-law will not address
their mother-in-law. If we take a transcript of a group of kin chatting
we should find no (or relatively few) instances of a son or daughter-in-law
addressing their mother-in-law. I hope that is not problematic and
controversial but rather that it demonstrates how the method can be
linked to some mainstream social anthropological concerns.
By contrast to the scale of much anthropology I am taking a fine grain
approach. The systematic and detailed analysis of naturally occurring
conversation poses many problems in its own right, both of how the
analysis should proceed and how the results relate to conventional
anthropological theorising. By examining in detail the way in which
speakers use small words such as names, pronouns and kin terms we
are asking questions at a level where informants perceptions are not
very acute. Speakers are concerned generally with the issue at stake
rather than their choice of words. This is no bad thing, particularly
since we are in no wise suggesting that inability to describe what
is going on means that speakers are not (at some level) aware of the
patterns in their speech. The ease with which one can give insult
and the inappropriateness of saying, for example, 'Professor old cock'
in the seminar room or (even worse) in the PhD viva clearly demonstrates
this. Recognition of the possibility of such inappropriate speech
implies that there are social norms of proper usage widely disseminated
among a community of speakers which we as students of human society
can legitimately study. The patterns of usage that may emerge are
indicative of underlying social/cultural phenomena.
Coding the data
Arguments for Crudity. The Lowest common denominator.
The manner or style of speech is affected by a variety of social factors
such as: the relative ages, sexes and intimacy of conversationalists.
We focus on one particular element of speech: the choice of expression
used to talk about people (we talk of Person Referring Expressions:
PREs for short). To the straightforward counting of the frequency
of linguistic items (such as PREs) we add some simple social variables,
relying on the simplicity of the coding categories to warrant their
application. The simple assessments required by our coding scheme
are those of a lowest common denominator. This seems legitimate as
a first approximation, in order to begin the analysis of the data.
It is definitely not, and must be taken as being, in any way exhaustive.
That caveat aside it seems prima facie plausible that,
for example, parents are more powerful than their young children.
The simplicity of the coding permits us to be confident that we have
correctly assessed what is going on and that comparability may thereby
be achieved. That we have achieved acceptable inter-coder reliability
scores attests to this.
One of the surprises in the results was the lack of statistically
significant differences which could be found between the conversations.
To some extent this is a product of our relatively small sample size.
Small numbers restricts the validity of the tests making it hard
to obtain statistically significant results. However, the project
sought to pioneer new research techniques and therefore can be regarded
as a precursor to further more extended research on larger datasets
in which small numbers will not restrict the statistical analyses.
For, on the face of it, there should be a world of difference between
North American families sitting round a table after a day at work
or school and a Mambila family preparing beer and food in the run
up to a major ritual in rural Cameroon. By aiming at rigour we may
have isolated ourselves from common-sense. Surely we know there are
important differences between Mambila and North America. However,
I would rather use this as a check on our common-sense intuitions
which we should recognise as being dangerously unreliable (as was
noted in the case of code switching above). If we can demonstrate
that some of those intuitions are correct then that is not
a trivial result. One of the major lessons of anthropology is that
native speaker intuitions, even of the educated English, cannot be
elevated to universal propositions about the human population! It
is no small task to try and test the sense of similarity and difference
that underlies so much of anthropology. Hence, I would claim it as
a non-trivial result that in our pilot study of north American English
we were able to demonstrate that parents use names for their children
and receive kin terms back. In other words not only is there a norm
but it is demonstrably practiced, unlike other norms we may care to
think of which are honoured as much in the breach (both the prohibition
of murder and Grice's maxims for conversation (Levinson 101/2?) have
been suggested as candidates for such norms).
Comparing across languages
If we want to be able to effect some meaningful comparisons between
languages and cultures we need be able to allow for variation in the
linguistic factors. In particular, some languages have a greater
frequency of pronouns than others. In Japanese, for example, pronouns
are notoriously rare. (Note that this implies that when they do occur
they are likely to be marked and hence of interest to sociologists).
What we have attempted to do is to normalise the figures we use in
order to take account of such factors.
If we sum over all speakers and conversations (within one language)
we get figures for the relative frequency of names, kinterms and pronouns.
We can use those figures to normalise the frequencies of individual
speakers. A speaker may use many names relative to the norm, or far
fewer. The variations from the language specific averages may be
socially conditioned by factors such as the social role of the speaker.
The averages themselves reveal more of the effect of the grammar
of the language in question.
Mambila (one conversation)
PRO KIN NAME DES COMBINATIONS Total
1166 35 222 83 55 1561
Normalized
0.747 0.022 0.142 0.053 0.035 1
English (summing over two conversations)
PRO KIN NAME (TITLE and NAME) DES COMBINATIONS Total
1377 84 192 116 43 1812
Normalized
0.760 0.046 0.106 0.064 0.024 1
The data considered
The English language data analyzed here consists of two conversations
each made by a nuclear family with a visitor in the setting of an
evening meal. The dinner-time setting determines the speech event
under study, with its specific scene, participants, and rules of interaction
(see Blum-Kulka, 1990). The speakers are white, middle-class, Jewish-Americans
with a guest/visitor, all from the east coast of the U.S.. The Mambila
data is the transcript of a conversation in the house of one of my
principal informants, his wife and some of their children, with two
adult visitors at different parts of the conversation. The recording
was made, in my absence and comments made during the conversation
show that those present were not unaware of the presence of the tape
recorder. Neither I nor Mambila people who have heard the tape can
ascertain any significant difference between this conversation and
others which were not recorded. I have explained above how the transcript
was prepared. I am in the process of writing up what may be called
the 'thick ethnography' of this conversation in the form of an annotated
translation and a set of ethnographic diversions to explain the significance
of particular topics which occur in the conversation, such as a description
of a set of marriage gifts.
Two statistical results:
I shall now briefly present two different statistical results.
1) A significant difference: explicit vocative use
One statistically significant difference between the US English and
the Mambila data is found in the distribution of explicit vocatives.
We class as vocatives all PREs that refer to the
addressee, including pronouns kin terms, titles, and names. When
the distinction is made between pronouns and other PREs, such as kin
terms, names and titles, we call this latter class ³explicit
PREs² because they uniquely specify the addressee in the vast
majority of cases, whereas the denotation of a pronoun is inherently
contextual. Hence ³explicit vocatives² are all vocatives
excepting pronouns.
Distribution of Explicit Vocatives
We examined both the frequency of use of vocatives and
the social role of the person referred to. This led us to construct
a standardised measure of net vocative use, that is, a measure of
the extent to which a participant gives or receives more vocatives.
This figure is the number of vocatives used by someone minus the
number of times that person has been uniquely1
referred to with a vocative. Both are normalised to be the frequency
per one thousand utterances. In this respect, it seems from §1
that there is a clear difference between the cultures considered.
§1:
It appears from this that Mambila parents use more explicit vocatives
than their American counterparts. Unfortunately I suspect that this
is a product of the data. There are far more children speaking in
the Mambila transcript and so the problem of specification becomes
more acute - sending a child or a younger sibling out to fetch water
is straightforward when there is only one such person, more complex
when there are several possible candidates.
When we examined the total number of explicit vocatives used or received
by a speaker, their vocative involvement as it were, we found it to
be strongly correlated with involvement in what Brown and Levinson
call 'face-threatening acts' (FTA) (that is, ftas given and received).
See §2.
§2:
The likely explanation is that if there is an underlying factor that
has a role in determining the number of times one addresses, and is
addressed with, an explicit vocative, then this factor is also responsible
for the number of ftas one is involved with. Later we will return
to this suggestion in the discussion of the factor analysis.
We next considered the role of social relationships in explicit vocative
usage. Having established that there are differences between the
Mambila and English data, we then considered if there were any differences
regarding social relationship. To do this we examined the speaker-addressee
pairs and the relationships between them. Results may be seen in
§3:
§3
We can see from the figure and the table that in the Mambila sample
there is a significant difference between the categories of child
and parent. Mambila children and parents have significantly different
patterns of gross explicit vocative involvement. In other words,
Mambila parents produce and receive more explicit vocatives than their
children. The American sample, however, shows no significant differences,
although the visitor is involved in fewer vocatives.
Let us examine this more closely by asking whether particular categories
use or receive more or fewer kin terms and names? §4 shows the
difference between the frequency of vocatives uttered for each social
category in the two cultures. It appears that children utter more
frequently than parents, who, in turn, utter more frequently than
visitors. This is true in both cultures, though it appears that Mambila
enjoys a higher overall frequency. The cultural differences are not
significant, but there is a significance between children and visitor
as shown in §5.
§4:
§5:
Looking at the production of explicit vocatives we can see (§5)
that children are more prone to using vocatives than their parents,
in both Mambila and American cultures. However, in Mambila there
is a significant difference between parent and child receiving explicit
vocatives, whilst there is no such difference between parent and
child in the American samples.
§6
{{Page X}}2) Introducing a factor analysis
The second statistical analysis that will be discussed here took all
the data together, pooling the English and Mambila results suitably
normalised, looking for factors to account for some of the variation
within the conversations.
As an initial step, we further simplified the data by considering
only the utterances that had unique addressees. Since there are many
possible permutations of conversational dyads we then restricted ourselves
to the more frequently occurring dyads, choosing to ignore dyadic
pairs that occurred less than ten times (which as we have already
noted may itself be a significant list - It does not need Sherlock
Holmes to point to the mother-in-law who is not addressed...)
We considered five variables within in the utterances:
1) Use of explicit vocatives
2) Address Pronoun
3) Involvement in (either giving or receiving ) FTAs.
4) Topic Change
5) SSA - measure of ³Presumptuous² Intent
The last variable uses the work of William Stiles to provide a means
to investigate the link between PRE use and Brown and Levinson's analysis
of politeness phenomena. The intent of each utterance is to coded
according to three principles (source of experience, presumption about
the experiences of others, and the frame of reference). These create
eight possible categories that represent different types of speaker-addressee
micro-relationship. One of Stiles' guiding assumptions is that each
category carries with it an associated set of grammatical features.
This enables the linguistic form of the utterance to be coded separately
according to grammatical rules with the same eight categories. Thus
a declarative sentence in the first person ('I' or 'we' that does
not include the hearer) may be coded as a 'Disclosure' form, whereas
a sentence whose subject is first person plural ('we') and includes
the hearer is to be coded as a 'Confirmation' form. In this way,
an utterance may share the same code for intent and form or have different
codes, as in the following examples (Stiles 1992, p. 10):
(1) ³Sit down² (pure Advisement).
(2) ³Would you like to sit down?² (Question form with Advisement
intent).
(3) ³I'd like you to sit down² (Disclosure form with Advisement
intent).
From the preceding, it is clear that by reformulating
one's intent in a grammatical form that is typically reserved for
other types of intent one is performing a strategy akin to facework
(see Goffman as elaborated by Brown and Levinson). Indeed, Stiles
identified (1981) 'presumptuousness' as a connection between his
classification scheme and the model of politeness strategies of P.
Brown and Levinson. The four intent categories that are considered
'presumptuousness' may be said to be intrinsically face-threatening:
'Advisements' are those utterances that threaten negative face, and
'Confirmations', 'Interpretations' and 'Reflections' threaten positive
face. The intrinsic threat can be dissipated by using a non presumptuous
linguistic form such as examples (2) and (3).
Having selected these five variables an initial statistical analysis
was performed. In this we asked to what extent the distributions
of these different variables can be seen as resulting from fewer underlying
factors. If such factors are identifiable there is the possibility
of interpreting them as being basic principles that are instrumental
in the choices people make of the PREs they use.
Having conducted the factor analysis for the variables above we then
continued to explore how the factors we have identified may relate
to other, more sociological variables. To do this we calculated the
values of the factors for each dyad and then sought to relate this
to the coding of each dyad for the social distance between the speaker
and addressee (intimate, distant), and the gradient of status difference
(i.e. whether they are talking 'up', 'down' or 'sideways').
Two factors emerged, onto which the five variables loaded quite well
(a variable loads with a value of +-0.3 onto one factor and hardly
at all onto the other). Pronouns/1000, total ftas/1000, presumptuousness
load onto factor 2, whilst vocatives/1000 and topic changing / 1000
load onto factor 1. In order to see where each dyad falls relative
to the others we returned to the data and calculated standardised
scores for each of the factors. This requires calculating the difference
of each dyad compared to the mean of the sample, all divided by the
standard deviation of the sample. Thus giving a standard variation
of mean zero, and standard deviation one. We then used graphs of
the dyads to investigate the role of status and intimacy in the variables
used for the factor analysis.
Plotting the dyads against the two factors but marking them for the
gradient of status, in other words, whether the addressee is of greater,
equal or lower social status than the speaker, produces the following
graph:
Note how the dyads where the speaker
was of lower status than the addressee (i.e. child to parent) are
to the left, and those where the speaker was of higher status than
the addressee (i.e. parent to child) to the right. In other words,
factor 1 appears to discriminate between the relative social status
of the conversants. The clustering of eight dyads in the same place
shows the limits of the analysis.
In the next graph we relate a simple measure of intimacy to the factors
already identified.
This shows the problems that face us
in seeking to interpret the factors we have identified. At first
sight factor 2 seems independent from intimacy since the dyads are
evenly distributed over the whole range of factor 2, but closer inspection
shows a slight tendency of the distant dyads to have low factor 2
values (i.e. to occur on the bottom half of the graph), yet the intimate
dyads occur for all values of factor 2. Factor 1 shows a slight articulation
with intimacy, a higher value of factor 1 seems connected to greater
intimacy but it is far from clear.
To help interpret these results consider first the results of extending
this factor analysis by including positive and negative politeness
in our list of variables. The factors then increase to three. In
the table that follows I have omitted the factor scores when they
are less than 0.3 to aid clarity.
What is revealing about this is the
way that the third factor seems mainly implicated in topic change
and use of explicit vocative. Note also the spreading of total FTA
involvement over all three factors. Brown and Levinson could take
heart from this for one factor that they identify which we have been
unable to take into account is the ranking of imposition. Asking
for a glass of water is different for asking for a thousand pounds.
Consider again the smaller analysis.
Now the classic models to account for choice of terms are first those
proposed in Brown and Gilman's ³The pronouns of power and solidarity²
which we may gloss as 'status' and 'intimacy'. This approach was
generalised by Brown and Levinson who identified three factors (status,
intimacy and ranking of imposition) as determining the choice of politeness
strategy to be employed, choice of term being a critical such strategy.
The factors that these theories deploy relate interestingly to Mary
Douglas's grid and group or even Durkheim's integration - regulation
distinction.
We are then posed the challenge of attempting the interpretation of
the factors identified by the statistical analysis. How can these
factors be linked to the sort of solid, sociologically revealing variables
that conventional analysis deploys? In short my question is whether
Brown and Gilman should take heart from an interpretation of factor
1 as status and factor 2 as intimacy (or vice versa)? I think that
what these results show is that such a simplistic reading is not possible
- we have shown that there are more things in heaven and earth than
are dreamed of in Brown and Gilman's (or even Brown and Levinson's)
philosophy! Among these are semantics, at least in the form of the
importance of topic shift. This may itself be related to power, and
hence to social status. For one manifestation of power in a conversation
is the control over the topic of conversation and how it shifts.
Very stark examples of this may be found in television and radio interviews,
but it has also been analyzed in the context of family conversation
by Richard Watts (1991). To a dimension involving semantics we appear
to require a further one that inter-relates status and intimacy.
This sort of complexity may be the appropriate way to further expand
Brown and Levinson's model to try and take account of some of the
recent criticisms that have been levelled at it.
This is a starting point for further research. For the present I conclude
with the claim that, counter Brown and Gilman and Brown and Levinson,
status and intimacy cannot be satisfactorily separated in the analysis
of naturally occurring conversation.
Bibliography
- Blum-Kulka, Shoshana 1990. 'You don't touch lettuce with your fingers: parental
- politeness in family discourse.', Journal of pragmatics ,
Vol. 14, pp. 259-288.
- Brown, P.B., and S.C. Levinson 1978. 'Universals in Language Use: Politeness Phenomena',
- in E.N. Goody (ed.) Questions and Politeness , Cambridge:
Cambridge University Press, pp. 56-290.
- Brown, R, and A Gilman 1960. 'The pronouns of power and solidarity', in T.A. Sebeok
- (ed.) Style in Language , Cambridge, Mass: MIT Press,
pp. 253-76.
- Ochs, E. 1979. 'Transcription as theory', in E. Ochs and B. Schieffelin (eds.) Developmental
- pragmatics
, New York: Academic.
- Stiles, W.B. 1981. 'Classification of intersubjective illocutionary acts.', Language
- in Society
, Vol. 10, pp. 227-249.
- Stiles, William B. 1992. Describing Talk: a Taxonomy of Verbal Response Modes (Interpersonal
- Communication 12), Newbury Park: Sage.
- Watts, Richard J 1991. Power in family discourse. Contributions to the sociology
- of language 63
, Berlin: Mouton de Gruyter.
Part C Publications for inclusion in RAPID
- copy
of electronic submission already received by RAPID
award number R000 23 3311
Article in Journal
%A Andrew J. Wilson
%D 1995
%T A Field-Guide to Bibliographic Research
%J Journal of the Anthropological Society of Oxford
%V25
%N 2
%P 179-83
Refereed? No
award number R000 23 3311
Article in Journal
%A Andrew J. Wilson
%A David Zeitlyn
%D 1994
%T Speech acts and Stiles
%J Linguistics and Education
%V 6
%N 1
%P 91-98
Refereed? Don't know
award number R000 23 3311
Article in Journal
%A Andrew J. Wilson
%A David Zeitlyn
%D 1995
%T The Distribution of Person Referring Terms in Natural Conversation
%J Research on Language and Social Interaction
%V 28
%N 1
%P 61-92
Refereed? Yes
award number R000 23 3311
Article in Journal
%A David Zeitlyn
%D 1993
%T Reconstructing Kinship or the pragmatics of kin talk
%J Man
%V 28
%N 2
%P 199-224
Refereed? Yes
award number R000 23 3311
Article in Journal
%A David Zeitlyn
%A Andrew J. Wilson
%D forthcoming 1995
%T The Childes Project: an anthropological resource
%J CAM. The Cultural Anthropology Methods Journal
Refereed? No
award number R000 23 3311
Dataset
Mambila transcript
archived in the CHILDES (Child Language Data Exchange System) database.
FTP poppy.psy.cmu.edu (128.2.248.42).
The files are in /childes/noneng
mambila.tar
mamfonts.sea.hqx
award number R000 23 3311
Dataset
Mambila transcript
Submitted to ESRC Data Archive July 1995
award number R000 23 3311
Consultancy
David Zeitlyn
Consultancy to Nokia Mobile Phones UK
Sociolinguistic aspects of communication: mediated and direct
throughout 1995