Mambila Corpus

David Zeitlyn
Institute of Social & Cultural Anthropology
University of Oxford


Participants: 14
Type of Study: naturalistic
Location: Cameroon
Media type: audio
DOI: doi:10.21415/T5N88R
Browsable transcripts
Download transcripts
Media folder

Citation information

Publications using these data should cite:
Wilson, A. J., & Zeitlyn, D. (1995). The distribution of person-referring terms in natural conversation. Research on Language and Social Interaction, 28, 61–92.

Additional relevant publications include:
Blum-Kulka, S., & Snow, C. (1992). Developing autonomy for tellers, tales, and the telling in family narrative events. Journal of Narrative and Life History, 2, 187–217.
Ochs, E. (1979). Transcription as theory. In E. Ochs & B. Schieffelin (Eds.), Developmental pragmatics. New York: Academic.
Stiles, W. B. (1992). Describing talk: A taxonomy of verbal response modes. Newbury Park: Sage.

In accordance with TalkBank rules, any use of data from this corpus must be accompanied by at least one of the above references.

Project Description

These data have been prepared as part of an ESRC-funded project “Kinship and language: a computer-aided study of social deixis in conversation” (grant no R000233311) which supported Andrew Wilson. Dr. David Zeitlyn, the project director, is a British Academy Research Fellow at the Institute of Social and Cultural Anthropology, Oxford and a Research Fellow at Wolfson College, Oxford.

Introduction to Mambila

The Mambila lie on either side of the Nigeria/Cameroon border, the bulk of them living on the Mambila Plateau in Nigeria. A smaller number (c. 12,000) are to be found in Cameroon, especially at the foot of Mambila Plateau escarpment, on the Tikar Plain. The fieldwork was restricted to these latter groups, and in particular to the village of Somie. Self-sufficient in food, the villagers have grown coffee as a cash crop since the early 1960s. Cameroonian Mambila on the Tikar Plain have adopted the Tikar institution of the chiefship, yet their social structure otherwise closely resembles that described for the Nigerian village of Warwar by Rehfisch (1972) based on fieldwork in 1953. Nigerian Mambila did not have the same type of institutionalized chiefship as is found in Cameroon. In Nigeria, villages were organized on gerontocratic principles, and largely lacked political offices. The system of exchange marriage described by Rehfisch (1960) has now vanished, and with it the two sorts of named group which recruited through different combinations of descent, marriage type (exchange or bridewealth), and residence. Marriage is viripatrilocal, and is increasingly on the basis of courtship although bridewealth is still a major factor. However, bridewealth may be paid in installments over a number of years. It is not cited as a reason for the failure of young men to marry. Most people in the village are members of either the Catholic or Protestant church. Zeitlyn (1993) gives some information about the kinship terminology. A short transcript from that paper plus digitized audio recording may be found at the following URL using gopher within the anthropology corner or at the following URL for the World Wide Web:

Introduction to the Data

The Mambila transcript that accompanies this file has been transcribed according to CHAT guidelines, with the following exceptions and constraints. First the utterances have been segmented according to the principles described in Stiles (1992). We note that this is controversial and for many other purposes we feel the turn or the phase may be safer albeit harder to define. A spoken phrase may be crudely understood to be a turn or an utterance beginning and ending either with a turn transition or a pause.

The transcript records a conversation in the house of Michel Sondue on 15-DEC-1990. The recording was made, in Zeitlyn’s absence, by Sondue and comments made during the course of the conversation show that those present were not unaware of the presence of the microphone. Neither Zeitlyn nor Sondue can ascertain any significant difference between this conversation and others which were not recorded. The transcription procedures deserve some mention. Soon after making the recording, Zeitlyn went through it with Sondue. At this stage, they made some contextual notes and copy of the recording. Sondue 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. In the course of making this second recording, Sondue explained various idioms and vocabulary items that were new to Zeitlyn. Zeitlyn subsequently transcribed the second recording in the UK and then returned to the original recording. The transcript was then coded in the UK following a scheme developed in a pilot study by Blum-Kulka and Snow (1992), elaborated 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 that were resolved during a further fieldtrip in May 1994. The result is a robust transcript. 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 nonparticipating 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 Somie includes children playing somewhere in the background!

Header Tiers and Background Information

The conversation is taken from a household in Somie where a taperecorder has been left with the father of the household (MIS — Michel Sondue) to minimize any effects due to the presence of the investigator (DZ — David Zeitlyn). The participants are predominantly family of MIS and his wife TBL, with the exception of two visitors, DAN and MBM. Most of the family members are co-resident with MIS and TBL, except for their eldest daughter ANG and her young child NKB. Throughout the conversation there is a general procession of participants in and out of the house. These tend to consist of the younger children who are playing outside. This causes two problems: a) it is often unclear who is speaking or what is being said when the voice comes from outside the range of the microphone, and b) without a visual record of what is happening, it has proved difficult to track the whereabouts of the participants, hence causing difficulties with the address tier. In particular cases of doubt or confusion, the file has been checked and re-checked with the participants themselves.

Main Tiers

Many turns in this conversation have been segmented into several distinct utterances that are coded individually. The motivation for this segmentation comes from our own research interests and the need to have a conceptually viable unit of conversation over which to score frequencies over certain linguistic and illocutionary items. The criteria by which we performed this segmentation comes from the work of Stiles (1992) and his concept of speech act. Stiles has presented a taxonomy that he claims is an improvement on the traditional attempts in being based on principles of classification whereas “most other systems have been developed empirically-by examining samples of a particular domain of discourse” (p.31). This will not only ensure that the taxonomy is both mutually exclusive and exhaustive (every possible utterance is categorized uniquely) but is more likely to be applicable to all languages, a feature of great interest to anthropologists doing cross-cultural comparisons, so long as the principles themselves are universally applicable. This will depend on the theoretical underpinnings of the taxonomy.

Stiles (1992, p. 14), a clinical psychologist, conceives every utterance (1) to concern either the speaker’s or the other’s experience, with experience understood broadly to include thoughts, feelings, perceptions, and intentional actions, (2) to either make presumptions about the other’s experience or not to presume anything of the other’s experience, and (3) to represent the experience either from his or her own personal viewpoint or from a viewpoint that is shared or held in common with the other. These three principles of classification he calls “source of experience,” “presumption about experience,” and “frame of reference,” respectively. These principles are dichotomous in having the value “speaker” or “other.” Hence we have a possible eight categories (2 X 2 X 2), or Verbal Response Modes (V.R.M.s for short), which he labels Disclosure (D), Edification (E), Advisement (A), Confirmation (C), Acknowledgment (K), Interpretation (I), and Reflection (R). He is careful to make clear that these names are only for convenience and the category classification should not be confused with their everyday connotations, although he uses the considerable overlap with natural categories as evidence that his principles are salient. This is further discussed in our review of Stiles (Wilson & Zeitlyn, 1995). Applying these criteria often requires that what seem to be normal sentences are split into two or more phrases that constitute separate utterances. On the main tiers we have paid little attention to tonal and prosodic information, but have incorporated information such as interruptions, pauses, overlaps and retraces. Retraces are further discussed below.

Dependent Tiers

%eng: Each main line has a free English translation. These tiers look like main tiers in that they preserve the main line information in CHAT format (except for the ID code). This will permit certain analyses to be completed on the basis of the English translation alone. In particular, the translation attempts to preserve the person referring expressions (i.e., pronouns, names, kin term) used.

%spa: All utterances are coded for their speech act. The taxonomy we have adopted for this purpose is the VRM taxonomy of Stiles (1992), which is easily applied to natural conversations. See Wilson and Zeitlyn (1995) for a full review.

%xadd: Addressee. This conversation is a multi-party conversation, so most utterances have several candidates for addressee, as there are several people whom can reasonably be expected to hear the utterance. Coding this aspect is achieved by a series of hierarchical cues that, when applied, will cut down the set of potential addressees to the set of actual addresses. The cues are as follows (Wilson & Zeitlyn, 1995):

%xtop: This tier records any changes in the topic or content of the conversation. Notoriously hard to pin down, we have adopted a rather intuitive approach to this aspect, coding on each utterance a change ($new) or reversal ($rev) of topic as it occurs.

%xfta: Face threatening act. An attempt to capture the intuitions of face work and linguistic politeness (Brown & Levinson, 1987) within this natural conversation has led to our coding of any utterance that is reckoned to threaten the face of others (addressee or not, we have not included face threats, e.g., insults, to absent third parties). Each code is presented in the following order: the face threatener, usually the utterer, though not always, the threatened, the imposition of the face threat, and whether the face threatened was positive or negative (Brown & Levinson, 1987; Wilson & Zeitlyn, 1995). A weak threat is coded as 1 and a strong threat is coded as 3. Thus the code $mis:gun:2+ve tells us that in that utterance, MIS has threatened GUN’s positive face to a value of 2.

%com: This tier provides additional information that might be of help in understanding the significance or meaning of certain parts of the conversation. For example, reference to third parties are elaborated upon, along with some general ethnographic information, as well as any breaks in the tape recording.

%xpre: This tier provides a code for all person referring expressions that occur in any one utterance. Three aspects of person referring expressions (PREs) have been coded for the category of expression, the actual linguistic form, and the status of the referent relative to the utterance. These three aspects are coded together and separated with colons.

We have distinguished the following five categories of expression: pronoun, kin terms, names, titles, and descriptive expressions. The distinction can be operationalism both on semantic grounds and on syntactic grounds (Wilson & Zeitlyn, 1994). For each category, the linguistic form is coded in a different way as follows:
1. Pronouns ($pro): The Mambila system of pronouns is roughly comparable to the English system, that is there are three persons (first, second and third) that can be either singular or plural. We have marked these by a number followed by “s” (singular) or “p” (plural). In addition we noted if the pronoun is a possessive (e.g., $pro:2spos). In addition there are other pronouns that do not have simple translations in English. For example, “Bubu” is a compound pronoun that refers to two persons, and “nyi” is an anaphoric pronoun used to refer to the speaker of reported speech. In these cases the Mambila form has been maintained in the coding tier.
2. Names ($nam): With names the actual linguistic realization is maintained on the tier to allow immediate inspection without reference to the main tier from which it came. There are some cases in which the form of the name is often varied, possibly to act as a mitigator (e.g., “Celistine” and “Celi”). These changes have been coded in the following way. The “unmarked” version of the name is determined by examining which is the most common version. Thereafter any variation of that name (“marked” forms) are coded with a (+) or (-) depending on whether they appear to be marked in a positive, more intimate, direction, or a negative direction. For instance, DAN is sometimes addressed as “Dan-e” This has been coded as $nam:dan-e(+).
3. Kin terms ($kin): The linguistic form of kin terms are coded in the same way as names, that is by preserving the original form and adding any intimacy marker where necessary.
4. Titles ($tit): The linguistic form of titles is maintained on the coding ti= er.
5. Descriptive expressions ($des): These expressions too are preserved on the coding tier. Often a descriptive phrase will be made up of two or more words. In these cases the words are linked on the coding tier with a _ (e.g.$des:b`¿_ntaar_dÏ). This preserves the expressions when conducting any FREQ searches. Some linguistic expressions are made up from combinations of these simple expressions. These can either consist of a series of simple expressions that refer to the some person which we term compound expressions (e.g., Aunt Sally, or Sir Brian) or a series of simple expressions which achieve ultimate reference by referring to others, termed oblique expressions (e.g., my brother’s daughter). These combinations are coded by coding the simple expressions connected with an & sign (e.g., $kin:tele:abs&$pro:1s:utt). In the case of compound expressions, where only one person, or set of persons, are referred to, the conversational status (explained below) is included only once.

The third aspect of this %pre tier is the conversational status of the referent, and requires coding the referent according to how he or she stands in relation to the utterance. This code rests on the following distinctions:
1. Utterer (utt): the expression can refer to self (e.g., “me”)
2. Addressee (adr): the expression can refer to the addressee (e.g., you or an explicit vocative).
3. Participant (cnv): the expression can refer to someone in the conversation who is not being addressed with that utterance.
4. Overhearer (aud): the expression can refer to someone within earshot, though not participating in the conversation.
5. Baby (bby): the utterance can refer to someone or something who does not have the ability to comprehend or reply. This category includes pets or small babies.
6. Absentee (abs): the expression can refer to someone who is absent (or dead).
7. Rhetorical (rhe): the expression can refer only in a rhetorical sense.

It is, of course, possible to refer to more than one individual with any simple expression (e.g., “they”, “sisters” etc.). Thus it is possible for the referents to have different conversational statuses. For instance, “we” will often refer to utterer and addressee or utterer and absentee. This outcome is coded by combining the status codes in the order given above, separated with a + sign. For instance, $pro:1p:utt+cnv+abs.

One final piece of information required when coding the person-referring expressions is to whether the expression was uttered as part of a retrace or not. If any p.r.e. is then retraced or coded, it is flagged with a dash at the end of the code. This allows the analyst the choice of considering attempts to refer (an illocutionary concept) or with linguistic data exactly as uttered. For instance:
*TBL: <Dan-o ke ka> [/] Dan-o, ke ka!
%eng: <Dan-o Look> [/] Dan-o Look!
%pre: $nam:dan-o:adr- $nam:dan-o:adr