slobin@cogsci.berkeley.edu, nhoiting@cogsci.berkeley.edu
THE BERKELEY TRANSCRIPTION SYSTEM (BTS)1
The Berkeley Sign Language Acquisition Project has developed a system for transcribing sign language videotapes at the morphological level. The system is based on efforts to transcribe adult-child interactions in American Sign Language and Sign Language of the Netherlands. The system was designed to be compatible with the conventions of CHILDES (Child Language Data Exchange System), and can be viewed on the web at http://childes.psy.cmu.edu/sign.pdf. The goal of the project is to provide a standard means of transcribing signed utterances, meeting the following objectives:
The system will be described, supported by sample transcriptions of adult-child signing in ASL and SLN. A presentation will also be made of the Berkeley-designed database tools for computer-aided logging and transcribing of videotapes, using Autolog and Microsoft Access (designed by Wes Reed).
The Berkeley Sign Language Acquisition Project has designed a system for transcribing videotapes of sign language interactions. The system was developed using data of adult-child interaction in American Sign Language (ASL) and Sign Language of the Netherlands (SLN); it is intended to be applicable to all sign languages and to all genres of signing. The Berkeley Transcription System (BTS) is modeled on extensive experience with computerized transcription and analysis of child speech, and is explicitly designed to be compatible with the accepted international standards in that field. At the same time, BTS strives to present a consistently morphological analysis of each sign language under study.
The full version of BTS is available for examination and use as part of CHILDES (the Child Language Data Exchange System), and can be accessed as Chapter 11 of the CHAT Manual on the web2. Here we present the rationale motivating the construction of BTS, with examples of the main features of the system. The Appendix to this paper summarizes all of the transcription conventions of BTS.
The goal of all transcription is to produce a permanent, written record of communicative events, allowing for analysis and re-analysis. The goals of analysis are critical in determining the appropriate level and scope of transcription (Ochs, 1979; Slobin, 1993). BTS is designed to capture children's emerging grasp of meaning components and combinatorial possibilities in the exposure language. At the same time, the multi-tiered format allows investigators to add tiers for the transcription and coding of other levels, such as phonology, syntax, and pragmatics.
The most basic aim of every system of notation of behavior is to help researchers see patterns in the data-that is, to facilitate their human pattern-recognition devices. The task of transcription and subsequent data summaries is to present information in various forms, so that one may identify regularities that may not be evident while directly observing the behavior in question. In the domain of communicative interaction, the data must be amenable to both qualitative and quantitative exploration.
Our inspiration is the CHILDES system, developed by Brian MacWhinney and many colleagues for the transcription, storage, and analysis of spontaneous speech corpora (Sokolov & Snow, 1994; see website [footnote 2] for details and references). CHILDES consists of two major components: a standardized format for computerized transcripts (CHAT), and an extensive set of programs for automatic analysis of transcripts in CHAT format (CLAN). Our goal is to provide a transcription format for signed languages that is fully compatible with CHAT guidelines, and thereby accessible to analysis using CLAN programs. At every point, we are working with Brian MacWhinney (Carnegie-Mellon University, Pittsburgh, Pennsylvania), in order to ensure that we have met these goals. MacWhinney has been supportive in guiding our efforts, placing the current version of BTS in the online CHAT manual (Chapter 11: Sign Language-UC Berkeley), and working towards the design of additional CLAN programs to deal with sign-specific issues. In the long run, we would hope that the CHILDES database will include corpora of child signing, eventually supported by digitized video archives. The current paper is an interim report of work in progress. We would appreciate feedback, especially with regard to the applicability of BTS to other signed languages and to a range of discourse types.
The CHILDES manual begins with a word of caution about "the dominance of the written word" (Section 2.2.1):
Perhaps the greatest danger facing the transcriber is the tendency to treat spoken language as if it were written language. The decision to write out stretches of vocal material using the forms of written language involves a major theoretical commitment. As Ochs (1979) showed so clearly, these decisions inevitably turn transcription into a theoretical enterprise.
Paradoxically, this danger faces transcribers of signed languages as well. In this case, the danger is to treat the sign language in terms of the written language of the surrounding speech community. All too often, scientific studies of sign language are based on upper-case glosses (in English, Dutch, Italian, and so forth), supplemented by descriptions of handshapes or discursive notes when a spoken-language equivalent is not readily available. The result is a hybrid of different types of information: lexical, phonological, gestural, pragmatic. A transcript of this sort is not consistent with CHAT format, and is therefore not amenable to CLAN analysis. In addition, on the linguistic level, it all too often leads the analyst to treat the sign language in terms of the written language used in glossing, rather than in its own terms.
Sign language communication poses additional problems for transcription, due to the simultaneous presentation of manual and non-manual information, movement through space, and the use of gestures along with conventional signs. These problems have been approached in various ways in the literature, but all current solutions require the use of varying font sizes, diacritics, special characters, superscripts and subscripts, and horizontal lines drawn across sequences of elements to indicate the scope of non-manuals. None of these formats can be simply and directly reproduced on the keyboard, using only ASCII characters in a single font size and typeface. Nor are there any clear guidelines for using such notations, as indicated by inconsistency from publication to publication. An ASL classifier might be indicated as "VEHICLE" in one paper or as "3-CL" in another; a non-manual feature might be indicated with an English word or abbreviation if it corresponds to an English category (e.g., "neg," "nod," "rhet.q"), or by a phonological notation if it does not easily map onto a single English word (e.g., "th," "puff.cheeks," "tight lips"). Motion might be described geometrically (e.g., "move to lf") or discursively (e.g., "swerve to lf to miss rabbit"). Such heterogeneity of transcription impedes systematic linguistic comparison between reported data in various publications.
BTS requires a consistent notation for all manual, non-manual, and movement components of utterances. The notation must be consistently morphological, as discussed in detail below. Furthermore, an utterance must be represented in linear fashion, on a continuous typed line, using only ASCII characters in one font size and typeface. We will not present each of the BTS conventions here; they are readily available to the reader in the appended Manual or online in subsequent updated versions. Rather, we will point out the ways in which BTS is based on (1) linguistic analysis, and (2) discourse considerations.
All content elements (lexical items, specific meaning components) are given in capitals and bounded by spaces. A sign is represented by at least two upper-case letters. This convention makes it possible to search for and count lexical items. Grammatical categories are indicated in lower-case, immediately bound to the content item. For example, cl'VEH indicates a vehicle classifier. The upper/lower case distinction allows for separate searches for function and content elements. Within a complex item, components are separated by hyphens. When necessary, the underscore is used to keep all parts of an item together, as a unit bound by spaces, e.g., PNT_1 'point to self'. Information provided in parentheses is not included as part of an item, e.g., SIGN(1h) indicates that a normally two-handed sign was produced with one hand. Further conventions will be introduced in the discussion of specific topics, below.3 See the appended Manual for a full listing of transcription elements.
BTS uses the symbol PNT (point) for all instances of pointing to a physically present referent or person, whether serving as signer (PNT_1), addressee (PNT_2), another person or object (PNT_3). PNT_3 is followed by a parenthetical indication of the referent, e.g., PNT_3(visitor), PNT_3(ball), PNT_3(picture_of_ball). The symbol IX (index) is reserved for reference to a spatial locus that represents a person or object in signing space. Section 2 of the Manual lays out the many possibilities for pronominal usage (i.e., points to persons) in signed languages. It is clear that the system is rich, including indications of number (singular, plural, multiple, selective) and inclusive/exclusive reference.
A major strength of BTS is its attempt to represent each of the several components of a sign. However, we attend only to those components that can be productively used to create meaningful complex signs in the language. Consider three types of ASL examples: (1) A "plain" verb, such as 'love', is formed by crossed arms moving against the chest with two S-hands, palms inward. These components are fixed and do not vary to change the meaning of 'love'; that is, this sort of verb has no morphological components. Signs of this sort are represented in the usual manner in BTS, using upper-case letters: LOVE. (2) Another sort of verb demonstrates "agreement" by moving in space, but, like LOVE, has no internal morphology. Consider ASL 'help', which consists of an A-hand resting on a palm-up base hand, moving between the 'helper' and the 'helpee'. The movement component is clearly meaningful, in that a variety of person-loci can fill the two roles. However, the configuration of the two hands consists of a single unit in the BTS analysis, because neither handshape can be substituted for another while still retaining the meaning 'help'. Again, an upper-case gloss for the lexical element is sufficient, but it will have affixed path morphemes, as discussed below: HELP-. (3) Verbs of object transfer and manipulation (e.g., 'give', 'put') have path movement components and handshape "classifiers" that can be substituted to specify the transfered or manipulated object. Such verbs are richly polycomponential ("polymorphemic" in Engberg-Pedersen's [1993] terms), and it would be grossly misleading to transcribe them as GIVE, PUT, and the like. For such verbs, BTS provides an expanded analysis, as discussed in detail below.4
BTS has a uniform analysis of verbs that move from one spatial locus to another, whether that locus is conceptualized as a place (locative verb, verb of motion) or a person (agreement verb, dative case, recipient). Path is one of many potential verb components, each of which is indicated by an initial hyphen. A verb that is not monomorphemic (in contrast to a verb such as LOVE), consists of a collection of components. The format for all such components is a lower-case indication of the component, followed by an apostrophe and an upper-case indication of the particular instantiation of the component. Thus, for example, -pth'Z indicates a zigzag path.5 Section 3.1. of the Manual lays out the full set of path elements (shape and directions). The elements can be combined-e.g., -pth'ZUF ('zigzag up forward'), but any combination still represents a single path component and is treated as a single content element.
There are four types of path of movement:
Verbs of motion (self-movement, caused-movement, object transfer) are polycomponential, including handshapes or body parts that indicate the figure and/or ground involved in the motion event. BTS considers signed languages in typological perspective, treating them as head-marked and polymorphemic. The transcription of "classifier constructions" is richly detailed in BTS.
The element that specifies ("classifies") figure/ground is always indicated in semantic terms. That is, an "inverted V" handshape is transcribed as cl'TL (two-legged animate being), and never as "V-CL," "inverted V," or the like. If both figure and ground are part of a verb, the order of notation is always ground before figure, following the logic of manual representation of such events.
In essence, verbs of motion in signed languages (at least in ASL and SLN, the languages we have worked with in detail), consist of components of ground, figure, path, and various additional movement elements indicating features such as aspect and manner. Such verbs cannot be directly glossed in English, or the other Indo-European, dependent-marking languages that are characteristic of the surrounding speech communities that have been most extensively studied. Consider, for example, an ASL verb with the following components: the non-dominant hand is held vertically, with flat palm, fingers extended forward (cl'VP 'vertical plane'); the dominant hand is in an inverted-V position (cl'TL 'two-legged animate being') and it moves to the top of the non-dominant hand (gol'VP_TOP 'move to top of vertical plane') to straddle the hand (pst'STR 'posture straddle'). This verb could refer to a range of events, such as a cowboy mounting a horse or a boy sitting up on a fence. It can be represented as a verb with four meaning components ("morphemes"), as indicated by four hyphens:
This is, in fact, a sufficient transcription linguistically, but it lacks legibility-at least for hearing readers. We would like to be reminded of a comparable English verb, but we do not want such a gloss to influence our transcription or analysis. To solve this problem, BTS allows the transcriber to begin a verb with a parenthetical, lower-case possible equivalent. Thus one might type:
The parenthetical gloss is not a conventional part of the system, and each transcriber can provide a suitable equivalent. For example, this verb could also be glossed as (get_up_on_horse) or (mount_straddling), or whatever seems useful to the transcriber. The parenthetical glosses stand outside of the analyses, and function only to facilitate reading.
If more contextual detail is needed, it can be provided on a dependent tier, under the utterance line. For example, one could add a "gloss" tier (%gls). Note that the utterance line begins with an asterisk and an identifying code for the speaker in three upper-case letters, while dependent tiers begin with a percent sign and lower-case ID:
The transcription is thus based on linguistic analysis, often resulting in initially non-obvious decomposition of complex signs. This work cannot be done without the active participation of native signers6. At almost every point in the development of BTS, the native signers have helped us to discover contrasts, nuances, and possibilities that may not have been evident to second-language signers.
Segmentation of a sign into meaning components depends on the availability of contrasts in the language. For example, our analysis of 'mount' is based on the possibilities of contrasting the ground (e.g., by use of a horizontal plane to indicate movement onto a different sort of ground), the figure (e.g., by reference to an animal, such as a cat, mounting a horse), and the posture (e.g., by contrast with a person standing on a horse's back). The search for contrasts is essential to the analysis, and contrasts are not always obvious without careful examination of a range of potential scenarios and their signed descriptions.
To continue the demonstration of this method, note that 'mount' is part of a collection of verbs that have a derivational relationship with one another, as revealed by the addition or removal of a meaning component.
Again, the parenthetical gloss is not part of the analysis. This verb could describe a man seated on a horse, a boy seated astraddle on a fence, etc.
Note that these relationships are not evident in the English glosses for each of the verbs discussed above. That is, if one relied on glosses as the central element of transcription, there would be no reason to see the regular relationships that hold between three verbs describing a human being mounting, straddling, and riding a horse: GET_ON, BE_LOCATED, and RIDE.
This sort of detailed morphological analysis is familiar to linguists who deal with comparable spoken languages, such as many American Indian languages. Consider, for example, Leonard Talmy's (1985) work on Atsugewi, a Hokan language of northern California. The verb roots in this type of language designate figures of particular shapes, postures, and consistencies, e.g.:
The roots take locative/directional suffixes, such as:
Polymorphemic combinations are similar to those of ASL; for example, consider the Atsugewi construction: s-´-w-it-mic. The first three morphemes indicate a first person subject in factual mood. The last two identify the postural figure and movement:
| it | -mic |
| linear_object_in_lying_posture | -move_down_onto_ground |
As in the BTS transcription, this would be sufficient to indicate the morphological components of the verb. In addition, following linguist practice, Talmy provides a gloss:
GLOSS:
s-w-´-it-mic
'I lay down onto the ground'
A transcription in BTS is not a gloss in the linguistic sense; that is, it does not correspond to an English paraphrase, such as 'I lay down onto the ground'. Rather, it is intended to be the equivalent of a morpheme-by-morpheme analysis, with a collection of abbreviations designed for signed languages.
The abbreviations of BTS are based on English, and equivalents will have to be found for transcriptions into other written languages. This is normal in linguistics, both for signed and spoken languages. For example, the dictionary of Argentinian Sign Language (Massone, 1993) uses CM (configuración de la mano) for 'handshape', RNM (rasgo no-manual) for 'non-manual feature', etc.; and upper-case representations of lexical items are in Spanish. The same is true, of course, for linguistic descriptions of spoken languages, according to the language of the publication. For example, 'singular/plural' can be represented as sg/pl in papers written in English, as??????? in papers written in Russian, and as tekil/ço_ul in papers written in Turkish. Users of BTS in other countries will face the challenge of devising abbreviated equivalents for each of the abbreviations of our system. Although many linguists will be able to work with the English-based system, it will be essential to prepare equivalents in other languages for the use of bilingual Deaf co-workers who do not know English. Once this has been accomplished, it should be possible to design simple translation programs that would make it easy for researchers to study each others' transcripts. For example, a list of Dutch equivalents of BTS notation symbols, along with a glossary of Dutch lexical items used in a transcript, would be used to prepare an English-based transcript of an SLN corpus.
Sign language researchers with experience in typological linguistics should not be surprised by the elaborateness of BTS transcriptions of polycomponential verbs. Such relatively opaque morpheme-by-morpheme glosses are familiar in papers dealing with a wide range of agglutinative and polysynthetic languages. Consider the following example from Inuktitut, spoken by a child of 2;5. Here we have an entire sentence in one polycomponential utterance (Crago & Allen, in press).
| suna | -tuq | -juq | -viniq | -u | -vunga |
| what | -consume | -NOM | -former | -be | -IND.1SG |
Note that the morpheme-by-morpheme gloss (what-consume-NOM-former-be-IND.1SG) is uninterpretable without knowledge of Inuktitut, just as BTS utterance-line transcriptions are uninterpretable without knowledge of the particular sign language. Because BTS is designed for investigators who know the sign language, however, the utterance line should generally be sufficient. The %gls line, like the line in single quotes above, is always available for clarification.
It should be evident to the reader that BTS relies heavily on criteria of morphological productivity for the analysis of a sign into components. To the extent that we have succeeded for a particular sign, this is a contribution to linguistic description. We are well aware, however, that children who are learning a language may not yet have carried out the adult analyses reflected in the transcription. This problem is a familiar one in child language, where it is well known that children's early forms may be "amalgams"-that is, unanalyzed Gestalten that correspond to more complex adult forms. The only way to determine if a particular morphological analysis is productive for a given child is to seek evidence of productivity. Such evidence is available in two forms: (1) One searches the corpus for uses of a given morpheme across lexical items and contexts, especially for overgeneralizations. That is, when an English-speaking child says "breaked" one has evidence for the productivity of the past-tense inflection. (2) One presents the child with new lexical items (often nonsense, or "nonce" terms made up for experimental purposes), and puts them in contexts that should elicit the form. Thus if an English-speaking child is presented with a nonce verb, wug, and says that someone "wugged" yesterday, again, one has evidence for productivity.
The same issue of rote-learned versus productive forms applies to the acquisition of signed languages. The purpose of the detailed componential analysis embodied in BTS is to make it possible to discover, for a particular child, when there is sufficient evidence to credit the child with control of various components of signs. The advantage of detailed analysis is that it pushes us to describe the language carefully, and makes us sensitive to critical dimensions of acquisition.
Analyses of adult signing, however, often treat complex lexical items as if they were frozen or unananlyzed. Again, we appeal to our criterion of substitutability in a lexical frame. A simple example may help to clarify this central issue in the design and use of BTS. Consider the ASL sign 'give', which is similar to signs for 'give' in other sign languages. In the "frozen" or "generic" or "non-specific" form, the handshape (flattened-O, palm up) does not specify the nature of the transferred object. The path components (-src- and -gol-) indicates the giver and recipient. One might conclude that the handshape is not a "classifier," because it does not "classify." However, in the BTS analysis, the generic handshape is just that: a generic classifier, corresponding to generic classifiers in many spoken languages. It is fully a meaning component of the verb. Furthermore, it contrasts with flattened-O, palm down, which means 'put' in ASL, using the same path component. In BTS such generic elements are labeled LEX-that is, their function is to indicate the particular lexical category (e.g., 'give' vs. 'put'). The transcription of generic 'give', in the frame 'from me to you', is therefore:
LEX(give) should be read as "the generic handshape that specifies the lexical item as 'give'." Replacing cl'LEX with a specific "classifier" handshape adds specification. For example, if the transferred object is a glass, cl'LEX is replaced by cl'CYL (cylinder). Thus the notation would be:
We will not give further examples of the use of LEX here. But we wish to underscore the importance of searching for contrasting examples, with native-signing co-workers, in order to determine if a sign should be transcribed componentially or simply with an upper-case word in the corresponding written language. In our experience, this is often a difficult issue, requiring a good deal of exploration and discussion before a satisfactory analysis is reached for a particular lexical item or set of related items.
The temporal dimension is central to signed languages, thanks to the use of two moving articulators and simultaneous information conveyed by parts of the face, body parts, and shifts in gaze and posture. All such components are treated as meaning components in BTS (Section 4 of the Manual), without deciding on the ultimate linguistic status of each dimension of signed communication. (Thus we prefer the term "component" or "meaning element" to "morpheme.")
The two hands can do different things at the same time. For example, a child signs CANDY with the dominant hand, while pointing on a picture in a book with the non-dominant hand (nh). BTS indicates temporally co-occurring elements by enclosing them between curly brackets. This utterance would have the form:
(For purposes of economy, the dominant hand is not explicitly indicated unless its use is exceptional in some way.)
A central feature of signed languages is the use of the face and/or body to add meaning to ongoing manual signing. BTS marks four distinct types of non-manual components. These can occur simultaneously with a single sign, or can have duration (scope) over several signs. The carat (^) is used to indicate onset and offset of a non-manual feature which has scope (corresponding to the horizontal line drawn across glosses in most sign language transcriptions).
A grammatical operator has scope over a phrase or clause (negation, question, topic, relative clause, conditional, etc.). The notation is ^opr'X ... ^ . For example, the following indicates negation of a proposition:
Note that opr counts as a grammatical element, similar to cl, pth, and so forth; NEG counts as a content element-that is, a lexical or meaning-bearing element, along with other upper-case elements. The same is true of other non-manual components.
Non-manual components can also be employed to modify the referential meaning of a lexical item or proposition by adding a dimension-through the articulation of the sign and/or accompanying facial expression-such as augmented or diminished size or rate, intensity, and so forth. The notation is ^mod'X ... ^ . To give a simple example, an SLN-signing 2-year-old wanted her mother to draw a big house. She greatly extended the sign HOUSE (AUG=augmented):
In this example, the non-manual component is part of a word. (There is no offset carat because such a non-manual ends with the end of the word sign.) Modification, of course, can extend over longer stretches of signing as well.
Another use of mouth, face, and body is to provide affective accompaniment to an utterance, indicating the signer's attitudinal stance towards the situation being communicated (e.g., disgust, surprise, excitement). The notation is ^aff'X ... ^ . For example, in SLN, a teacher asks a child to do something and the child agrees, though with some worried concern:
While it may seem unorthodox to treat such affective coloring on a par with other components of the language, we believe that such concerns are based on a narrow and conventional definition of what is "linguistic." In comparable situations in spoken languages, affective prosody has a marginal role in most linguistic analyses, while affective particles and inflections are considered more "linguistic." We do not have a reasoned argument to exclude some types of non-manual components from our transcriptions; we leave it to individual researchers to decide how to present different types of components in their descriptions and theoretical accounts of the language and its course of acquisition.
Non-manual markers are also used to regulate the flow of discourse, corresponding to discourse particles and intonation contours in spoken languages. Such markers have interpersonal functions such as checking if the addressee has comprehended, has agreed, and so forth. The format is ^dis'X ... ^ . In the following example, a Deaf SLN-signing mother responds to her 2-year-old's labeling of the lights on a picture of an ambulance. Note that there are two types of non-manual elements in this utterance. The first is an operator, indicating confirmation (YES); the second is a discourse marker checking whether the child agrees (CONF = confirmation check). The operator (repeated head-nodding) extends throughout the utterance, including the discourse marker (a sort of questioning facial expression). The offset timing of the two non-manuals coincides ( ^ ^ ).
A pervasive aspect of sign language communication are the subtle shifts of gaze and posture that allow the signer to convey the utterances, thoughts, or actions of other people. This part of sign language needs much more careful study, and BTS does not present a fine-grained analysis of role shift at this time. However, we do consider it to be a meaningful element, and one that follows conventional, linguistic patterns. At this point, we simply indicate role shift by RS. Note that we use capital letters for this element, treating role shift as a meaning component in an utterance. We do not use the carat (^) to indicate onset and offset of role shift, because we want to search separately for non-manual features and role-shifting. Instead, we use the reverse apostrophe (left single quote, grave accent,) for this function: 'RS ... ' . For example, in a book-reading activity, a Deaf ASL-signing mother points out a picture of a dog, and then role shifts into the dog to indicate that the dog is excited. She signs EXCITE with an accompanying non-manual indicating the dog's affect. The notation 'RS(dog) indicates that she has taken on the role of the dog. Note that ^aff can co-occur with role shift. (Section 4.2. provides notation for cooccurrences of various types of non-manual components, with various onset/offset timings.)
Gaze direction is another aspect of signing that needs more careful analysis. BTS allows for indication of gaze direction, leaving it up to the transcriber to decide when this is relevant to the particular study at hand. Because we are concerned with the development of signing, and with the nature of caregiver-child interaction, we have developed means of noting direction and timing of gaze (Section 4.2.6. of the Manual).
Again, because we are studying the development of signing within the context of ongoing communication, BTS provides means of noting attention-getting devices (Section 5.2. of the Manual) and gestures and actions that are relevant to communicative events. As indicated in Section 5.1. of the Manual, gestures (%ges) and actions (%act) can be entered as part of the utterance line, or on a dependent tier, at the discretion of the transcriber. In our preliminary transcriptions of parent-child interaction with 2-year-olds, we have often found it useful to include such information on the utterance line. For example, an SLN-signing 2-year-old is looking for a pen:
Finally, BTS provides means of indicating factors that are relevant to analysis of child signing. Section 6 of the Manual describes notation conventions for errors, interruptions, and for the creation of dependent tiers for additional information and coding.
Errors are noted by [*], with further information on a dependent tier (%err). For example, an ASL-signing child of 1;9 signs HORSE with a handshape error. For this analysis, the transcriber is not concerned with the phonology of the error and simply notes on the error tier that there was a handshape error ($hs); however, another transcriber might have added a dependent tier for phonological notation (%pho).
Dependent tiers (Section 6.6) provide for more extended commentary, including situational comments, error analysis, phonology, and so forth. For example, a hearing ASL-signing mother to a 3-year-old; the comment tier (%com) provides the transcriber's clarifications:
We hope that this paper, in conjunction with the Manual, will provide the reader with sufficient information (and motivation) to try out BTS. We expect to receive many questions, objections, and constructive suggestions. Our aim is to continue to refine BTS, in international collaboration, to the point that it can serve as a standard, along with the CHILDES standard for transcription and analysis of spoken languages. We also hope that the practice of extending BTS will contribute to the morphological analysis of various sign languages. Please send comments to Dan Slobin at slobin@cogsci.berkeley.edu or to Department of Psychology, 3210 Tolman #1650, University of California, Berkeley, CA 94720-1650, USA. Comments will be shared with the BTS group, and a discussion should ensue.
Crago, M. B., & Allen, S. E. M. (in press). Acquiring Inuktitut. In L. Leonard & O. Taylor (Eds.), Language acquisition in North America: Cross-cultural and cross-linguistic perspectives. San Diego, CA: Singular Publishing.
Massone, M. I. (1993). Lengua de Señas Argentina: Diccionario bilingüe. Argentine Sign Language: Bilingual dictionary. Buenos Aires: Editorial Sopena Argentina.
Ochs, E. (1979). Transcription as theory. In E. Ochs & B. Schieffelin (Eds.), Developmental pragmatics (pp. 43-72). New York: Academic Press.
Slobin, D. I. (1993). Coding child language data for crosslinguistic analysis. In J. A. Edwards & M. D. Lampert (Eds.), Talking data: Transcription and coding in discourse research (pp. 207-219) . Hillsdale, NJ: Lawrence Erlbaum Associates.
Sokolov, J. L., & Snow, C. (Eds.). (1994). Handbook of research in language development using CHILDES. Hillsdale, NJ: Lawrence Erlbaum Associates.
Talmy, L. (1985). Lexicalization patterns: Semantic structure in lexical forms. In T. Shopen (Ed.), Language typology and semantic description: Vol. 3. Grammatical categories and the lexicon (pp. 36-149). Cambridge, UK: Cambridge University Press.
1 The Berkeley Transcription System (BTS) has been developed in the Child Language Research Laboratory, Institute of Human Development, University of California, Berkeley (UCB). Support is provided by the Linguistics Program of the National Science Foundation under grant SBR-97-27050, "Can a Deaf Child Learn to Sign from Hearing Parents?” to Dan I. Slobin, PI and Nini Hoiting, co-PI. Additional support has been provided by the Institute of Human Development and the Institute of Cognitive Studies, UCB, and by the Royal Institute for the Deaf "H. D. Guyot”, Haren, The Netherlands. The system has been developed on the basis of parent-child videotapes in American Sign Language (ASL), collected by Reyna Lindert, and Sign Language of the Netherlands (SLN), collected by Nini Hoiting.(back to text)
2 The North American site, organized by Brian MacWhinney, is at Carnegie-Mellon University: http://childes.psy.cmu.edu/chat.pdf. There is a European mirroring site organized by Steven Gillis in Antwerp, Belgium: http://atila-www.uia.ac.be/childes/chat.pdf. There is a Japanese mirroring site organized by Hidetosi Sirai in Chukyo: http://jchat.sccs.chukyo-u.ac.jp/CHILDES/chat.pdf. (backt to text)
3 BTS is still under revision, as new issues are encountered in the process of transcription. A website will be established for ongoing updates and discussion across research groups and sign languages.(back to text)
4 We prefer to use the term "component” or "meaning component,” rather than "morpheme,” thus leaving us free to include a range of meaning components without prejudging their formal linguistic status. As indicated in examples below, and in the Manual, non-manual components of affect, perspective, and the like, function as systematic meaning components in polycomponential verbs. We are also uncomfortable with the term "classifier,” preferring to think of handshapes and body parts that specify characteristics of referents as "specifiers.” However, we’ll keep the familiar term, "classifier”—at least until the NSF Classifier Conference to be held in San Diego in April 2000. (back to text)
5 We are limited in our choice of diacritics by pre-assigned functions of elements in the CHILDES system. Thus, while -pth:Z might be a more familiar or transparent notation, users of BTS must get used to an apostrophe for specification of content elements. (back to text)
6 There are three in our group: Marlon Kuntze, who is Deaf, and Jennie Pyers and Helen Thumann, who are CoDA’s. Their insights have been essential to the creation of BTS. (back to text)
7 See the CHILDES website [footnote 2] for the full set of required conventions.(back to text)
The following conventions are intended to be consistent with CHAT and CLAN (with some necessary additions, given the nature of sign languages).1 The goal is to represent utterances in a consistent morphological and semantic notation, following the grammars of ASL and SLN (and, potentially, other sign languages). We have avoided any sort of phonological transcription of utterances in the basic representation of turns. These conventions are thus intended for the speaker tier (*). Dependent tiers (%) will be addressed later. Manual and non-manual elements are represented in a single line, using ASCII characters only.
Lexical items are written in capitals and bounded by spaces. Because of this distinctive use of capitalization, searches in CLAN must use the +k switch to recognize the distinction between upper/lower case. A sign is represented by at least two upper-case letters. There can be no spaces within a lexical item: The components of polycomponential lexical items are separated by hyphens (as discussed in Section 3, Polycomponential Signs); other elements are joined by underscore or parentheses without spaces. An utterance line ends with a period or question mark, preceded by a space.
| SIGN # SIGN | pause between SIGNs |
| SIGN(*2) | sign repeated rapidly, twice |
| SIGN(*N) | sign repeated rapidly, multiple times |
| SIGN_SIGN | contracted SIGNs, treated as a single new word |
| SIGN-SIGN | contracted SIGNs, treated as two words |
| &SIGN | uncompleted SIGN |
| <SIGN> [?] | uncertain transcription |
| XX | unintelligible but definite sign, to be included in word counts |
| xxx | unintelligible sign or gesture, to be excluded from word counts |
| SIGN(fs) | SIGN is a fingerspelled loan sign |
| S_I_G_N(fs) | SIGN is fingerspelled, not a loan sign |
| SIGN: | SIGN is held |
| SIGN(prx) | SIGN directed to close/proximate location |
| SIGN(mid) | SIGN directed to intermediate location |
| SIGN(dis) | SIGN directed to distant location |
| SIGN(1h) | one-handed SIGN (if usually 2h) |
| SIGN(2h) | two-handed SIGN (if usually 1h) |
| SIGN(nh) | non-dominant-handed SIGN (if anomalous) |
| SIGN(dh) | dominant-handed SIGN (if anomalous) |
Note: If marking both number of hands and which hand(s), the number of hands comes first. e.g., SIGN(1h)(nh). If SIGN is (1h), only mark which hand if the non-dominant hand is used.
| SIGN(v) | SIGN is a verb (if ambiguous) |
| SIGN(n) | SIGN is a noun (if ambiguous) |
| SIGN | citation form |
| SIGN2 | alternative form (e.g., WHERE, WHERE2) |
| X@ns | name sign (with X handshape) |
| X@is | idiosyncratic/invented sign (with X handshape) |
| X@hs | home sign (with X handshape) |
| PNT_1 | point to self |
| PNT_2 | point to interlocutor |
| PNT_3(person) | point to third person, if present |
| PNT_3(obj) | point to object, if present |
| IX_3(person/object) | index a person or object in signing space, if not present |
| PNT_1_2 | 1st & 2nd pers sing ('me and you') |
| PNT_1_3 | 1st & 3rd pers sing ('me and him/her', 'two of us') |
| PNT_1+ | 1st pers pl ('me and somebody', generic 'we') |
| PNT_S | Selective: Specific points to each of the people or objects being referenced. This is used either to emphasize the individual referents, or if the people being referenced are not physically near each other. |
| PNT_M | Multiple: The referents are indicated using a "1" handshape (index finger extended) and a sweeping motion. This can be used for any number of referents greater than 1. |
| PNT_N | Numbered: The number of people (or objects) being referenced is incorporated into the handshape of the pronoun (for 1-5 referents). |
| PNT_1_2_S | 1st & 2nd pers pl, selective ('me and specific others of you') |
| PNT_1_2_M | 1st & 2nd pers pl, multiple/sweep ('me and all of you') |
| PNT_1_2_N | 1st & 2nd pers pl, numbered ('me and a certain number of you') |
| PNT_1_3_S | 1st & 3rd pers pl, selective ('me and specific others') |
| PNT_1_3_M | 1st & 3rd pers pl, multiple ('me and all others') |
| PNT_1_E | 1st pers pl, exclusive ('we', excluding addressees) |
| PNT_1_I | 1st pers pl, inclusive ('we', including addressees) |
| POSS_1 | 1st pers sg, possessive |
| POSS_2 | 2nd pers sg, possessive |
| POSS_3 | 3rd pers sg, possessive |
| POSS_1+ | 1st pers pl., possessive |
| POSS_2+ | 2nd pers pl., possessive |
| POSS_3+ | 3rd pers pl., possessive |
Examples of more complex pronouns:
| PNT_1 PNT_2 | 1st pers, 2nd pers, in succession ('me, you') |
| PNT_1_2*2 | 1st & 2nd pers (2) ('me and you two') |
| PNT_1_2*3 | 1st & 2nd pers (3) ('me and you three') |
| PNT_2_3*2 | 1st & 3rd pers (2) ('you and them two') |
In the fullest possible elaboration, a polycomponential construction includes:
Only the gloss and one classifier are obligatory. Locations, movement patterns, and paths of movement may be absent or may have several entries. There can only be one aspect entry. These component morpheme types are indicated in lower case, followed by an apostrophe and specification of the content component; e.g., -cl'TL indicates a two-legged animate being. The order of the components is: parenthetical gloss, classifier(s) (ground/figure), location/movement, modification, aspect (see examples).
Each of the eight possible components of polycomponential verb transcription is presented below, with examples at the end of this section.
The first symbol in the verb transcription is the approximate English gloss (e.g., jump, dismount, ride_seated, ride_mounted). The elements within a gloss are separated by underscores, in order to retrieve them as units.
| -pth'X | path of movement, when semantically meaningful |
| -src'X | movement from a place or from contact |
| -gol'X | movement to a place or to contact |
| -rel'X | movement relative to a fixed referent object |
The components "source" and "goal" can be combined with the component "locative relations" (see Section 3.4.1) to indicate which part of the figure and ground are in contact, e.g.: (jump)-cl'HP-cl'TBL-src'HP_TOP-pth'A ('two-legged figure jumps from the top of a horizontal plane in an arc path'). The components of path, source, and goal are indicated by uppercase letters from the following list. (The locative components of referent points in relative movement are the same as those for locative relations, and are listed in Section 3.4.1.)
| I | line |
| A | arc |
| C | circle |
| W | wandering |
| Z | zigzag |
| U | up |
| D | down |
| F | forward |
| B | backward |
| S | side |
| R | right |
| L | left |
| BF | back-and-forth |
| EO | two classifiers moving towards each other (e.g., -gol'EO) |
| OBJ(ref) | real-world object referent (e.g., -gol'OBJ(paper)) |
| X_cl'X | location/direction in relation to classifier (e.g., L_cl'CYL 'left of cylindrical object') |
| OUT | out |
| IN | in |
The notation -cl'X indicates a classifier of type X. The following list of classifier types is partial, and is open to refinement; eventually, handshape pictures will be provided with this section. Note that classifiers are given semantic (e.g., 'two-legged animate being') rather than phonological (e.g., 'V') definitions. Sometimes there is no single English word that summarizes the semantic content of a classifier; therefore the abbreviations should be treated as mnemonics for the category indicated by the handshape. When two classifiers are part of a single verb, the order of notation is ground followed by figure.
| VP | Vertical Plane |
| HP | Horizontal Plane |
| HPG | generic ground (phonologically: horizontal plane, palm up) |
| HPS | specified ground (phonologically: horizontal plane, palm down) |
| SS | sitting surface (phonologically: two-finger horizontal plane) |
| NHP | Narrow Horizontal Plane |
| RHP | Rounded Horizontal Plane |
| PL | Plane (non-specific posture) |
| IP | Inclining Plane |
| DP | Declining Plane |
| TL | Two-Legged animate being |
| TBL | small animal, seated person, chair (Two Bent Legs) |
| FL | Four-Legged erect being (two-handed classifier) |
| FBL | four-legged animal (Four Bent Legs; two-handed classifier) |
| CN | Container |
| CYL | Cylindrical object |
| FD | Flat Disk |
| FT | Flat Thin object |
| TR | Tree |
| HOLD | Hold object |
| HOLD_SM | Hold small object (X-handshape) |
| HOLD_FL | Hold flat object (flat O-handshape) |
| HOLD_TN | Hold thin object (S-handshape) |
| HOLD_TK | Hold thick object (C_handshape) |
| LID | Lid |
| STK | Stick-like object |
| GUN | Gun-like object |
| VEH | Vehicle classifier |
| CIR | Circular object (F-handshape) |
| SPHERE | Spherical object (e.g. ball, balloon) |
| BOX | Box-shaped object |
| OBJ | Real object (specified in parentheses) |
Body part designation:
If the signer uses an actual body part, rather than a manual sign to designate a body part, use the following notation: cl'X'B, where X is the name of a body part, e.g.:
-cl'HEAD'B
If a body part of another person is used to indicate a body part in a signed utterance, indicate that person as B_2 (addressee) or B_3(other). Such body part designations can play various roles in polycomponential verbs. For example, a description of feeding a baby may include the baby's actual mouth as the goal: -gol'MOUTH'B(baby).
The notation -loc'X indicates a locative relation between figure and ground of type X. Locative components are used to indicate the location of the figure classifier with respect to the ground classifier, or of two classifiers to one another. The following is a partial list:
| INT | interior ('inside') |
| SUP | superior ('above') |
| INF | inferior ('below') |
| TOP | top ('upper surface') |
| BOT | bottom ('under surface') |
| EDG | edge |
| FRO | front |
| BAC | back |
| PAR | parallel |
| NXT | two classifiers articulated at the same time and are articulated next to each other, but do not indicate a figure/ground relationship. |
Note: Two locative components can be combined; e.g., FRO_EDG (front edge).
The component "position" (pos'X) indicates the position in signing space where a classifier or classifier configuration is articulated. The position components are the same as those for path, source, and goal (see Section 3.2, Paths of Movement).
The component "posture" (pst'X) indicates the posture of the figure for the subset of polycomponential verbs which indicate posture. Examples of such verbs are sit, stand, lie, mount, and ride. The following posture components have been defined:
| ERC | erect | STR | straddling |
| RCL | reclining | SIT | sitting |
| RCL_V | reclining ventrally | RCL_D | reclining dorsally |
The component "orientation" is used to indicate the orientation (relative to the signer) of either the figure or the ground. In a polycomponential construction, information about orientation and posture should follow the classifier to which it refers. The following orientation components have been defined:
| F | forward |
| B | back |
| L | left |
| R | right |
| U | up |
| D | down |
The notation -mvt'X indicates the movement pattern of a verb. Lexical movement (mvt'LEX) indicates the movement pattern that identifies the particular verb. The lexical movement pattern does not contrast with other movement patterns; its only function is to indicate that the configuration has the meaning of the particular verb.
| -mvt'LEX(verb) | movement which defines a lexical item but gives no further meaning; e.g., the ASL verb 'ride' (on an animal) consists of classifiers indicating the configuration of ground (vertical plane), figure (two-legged creature), and posture (straddling), plus a non-directional component of movement |
| -mvt'WIG | wiggling movement |
| -mvt'PIV | pivot (rotation of wrist) |
| -mvt'BOUNCE | bouncing movement |
| -mvt'CLOSE | hand closing movement |
| -mvt'JAB | short, jabbing movement |
| -mvt'LONG | showing long object, e.g. shelf of bed |
| -mvt'BEND | bending movement |
| -mvt'TWIST | twisting movement |
| -mvt'CP | change in posture |
| -mvt'CO | change in orientation |
| -mvt'WANDER | wandering movement |
| -mvt'TRACE | tracing movement (for SASSes and descriptive classifiers) |
See Section 4.2 for an explanation of non-manual components and a list of non-manual components that may be included as part of a polycomponential construction.
The notation -asp'X indicates an aspect of type X. Various aspects can be superimposed on a verb. A full list is not yet ready. Example codes are:
| CES | cessive |
| ITR | iterative |
| ITR_CUM | iterative cumulative (e.g., stacking of blocks, one on top of the other) |
For example, a single verb can indicate that riding was rapid and came to an end, resulting in a six-component verb:
(ride_mounted)-cl'VP-cl'TL-pst'STR-mvt'LEX(ride)-mod'RAP-asp'CES
A configuration of classifiers can act as a unit with respect to another component, including real-world objects. Curly brackets are used to indicate simultaneity. With regard to verbs, a configuration can move to a new location; for example, a doll on top of a board is moved to be located on a table:
(put){-cl'HP-cl'TL-loc'HP_TOP}-gol'OBJ(table)
In a series of utterances, a configuration can be held or continued from a previous utterance. For example, if the doll-on-board had already been set up in a previous utterance, the tilde (~) is used for each component which is continued, to indicate that this component was not created anew in the following utterance:
(put){-~cl'HP-~cl'TL-~loc'TOP}-gol'OBJ(table)
In addition, the component may serve as a different element in the second utterance, e.g., src'INT can become -~loc'INT in a subsequent utterance. This means that the element with the tilde is continued from the end product of the previous utterance.
The percent sign (%) is used when the handshape changes to form a new sign which adds meaning, yet the configuration is held over from the previous utterance. In the following example, in the first construction (look-around) cl'HEAD is formed with one handshape, but in the second construction (head-move) the index and middle fingers are extended forwards to form cl'%LOOK:
(head_move)-cl'CN-cl'HEAD-src'INT-pth'U-gol'OUT-cl'HEAD'B-mvt'ULR'B-^AFF'B .
(look_around)-~cl'CN-cl'%LOOK-~loc'INT_SUP-mvt'LR-cl'HEAD'B-mvt'LR'B ~^AFF'B
The following are examples of polycomponential verbs, with possible translations in parentheses. Note that this analysis reveals derivational relationships between verbs of location and verbs of movement.
(sit_on)-cl'VP-cl'TL-loc'VP_TOP-pst'STR = 'sit on a horse'
(mount)-cl'VP-cl'TL-gol'VP_TOP-pst'STR = 'get on a horse'
(ride_mounted)-cl'VP-cl'TL-loc'VP_TOP-pst'STR-mvt'LEX(ride) = 'ride a horse'
(dismount)-cl'VP-cl'TL-loc'VP_TOP-pst'STR-src'VP = 'get off of a horse'
(mount_seated)-cl'CN-cl'TBL-gol'CN-pst'SIT = 'get into a car'
(ride_seated)-cl'CN-cl'TBL-loc'CN_TOP-pst'SIT-mvt'LEX(ride) = 'ride in a car'
(jump)-cl'HP-cl'TL-pst'ERC-mvt'LEX(jump) = 'jump up and down'
(jump)-cl'HP-cl'TL-pst'ERC-src'HP = 'jump off of a horizontal plane'
(jump)-cl'HP-cl'TL-loc'HP_TOP-pst'ERC-src'HP-gol'HP = jump from one point to another on a horizontal plane'
(get_on)-cl'HP-cl'TBL-gol'HP_TOP-pos'USL = 'cat getting on a high, side, left table'
(give)-cl'LEX(give)-src'3-gol'1 = 'give from her to me'
(give)-cl'CYL-src'3-gol'1 = 'give cylindrical obj from her to me'
Verb agreement is indicated by the same conventions as used for transcribing directionality in verbs of motion (i.e., by use of src/goal and numeric indications, as in the examples of 'give', above). For example, 'You show me': (show)-cl'LEX(show)-src'2-gol'1.
Single curly brackets surrounded by spaces enclose elements that co-occur in an utterance. For example, a child signs CANDY while pointing on a book with the non-dominant hand:
| *CHI: | { CANDY PNT(nh)(on_book) } . |
Note: Indicate which sign is on the non-dominant hand; the default is the dominant hand.
Curly brackets are surrounded by spaces. To indicate earlier onset of one sign in curly brackets, append (o) to the sign, e.g.:
| *CHI: | { CANDY(o) PNT(nh)(on_book) } . |
Overlaps are coded in the standard CHAT fashion:
| *CHI: | WANT < BOOK > [>] PNT_2 ? |
| *MOT: | < WANT > [<] . |
Note on utterance segmentation: If a signed utterance is grammatical, break the utterance by proposition or clause boundaries. If a signed utterance is ungrammatical, break the utterance by prosody (indicated by pauses, placing the hands down, etc.).
Non-manual elements are indicated by a carat (^). There are four types, as described below: operator (^opr), modification (^mod), affect (^aff), and discourse marker (^dis). Such an element can be added to a single sign; however, if the non-manual element has scope over several signs, this is indicated using the following conventions:
| SIGN-^opr'X | non-manual element associated with a single sign |
| ^opr'X SIGN SIGN ^ | non-manual element has scope over several signs |
If two different non-manuals are superimposed on a single sign or utterance, each has its own carat, using the following conventions:
| {^opr'X ^mod'X} SIGN | simultaneous onset of two non-manuals |
| SIGN ^ ^ | simultaneous offset of two non-manuals |
| ^opr'X ^mod'X SIGN | sequential onset of two non-manuals |
| ^aff'X ^opr'X SIGN opr^ SIGN aff^ | sequential offset of two non-manuals |
| SIGN-^opr'NEG | non-manual which adds a component to a sign |
| ^opr'X | grammatical operator which operates on a whole phrase or clause (e.g. negation, yes/no or wh- question, topical marker, relative clause marker, conditional marker) (partial list) |
| ^opr'NEG | negation |
| ^opr'YNQ | yes/no question |
| ^opr'WHQ | wh- question |
| ^opr'TOP | topical marker |
| ^opr'REL | relative clause marker |
| ^opr'COND | conditional marker |
| ^opr'AFR | affirmation (head nod) |
| ^opr'RHQ | rhetorical question |
| ^mod'X | modifies the referential meaning being expressed by adding a dimension (e.g., augmented/diminished size, rate, intensity) (partial list) |
| ^mod'RAP | rapid movement |
| ^mod'DUR | durative activity, situation |
| ^mod'AUG | augmented size, rate, or intensity |
| ^mod'EFF | with exaggerated effort |
| ^mod'CARE | with care or caution |
| ^aff'X | freely varying affective accompaniment to a lexical item or utterance to indicate the signer's attitudinal stance towards the situation being communicated (e.g., disgust, surprise) (partial list) |
| ^aff'DISGUST | disgust |
| ^aff'SURPRISE | surprise |
| ^aff'ANGER | |
| anger |
| ^dis'X | markers which regulate the flow of discourse (e.g., checking for agreement, comprehension, confirmation) (partial list) |
| ^dis'CONF | confirmation check |
| ^dis'AGR | agreement |
The signer can shift into the role of someone referred to, using various means of body and gaze shift into and out of the neutral position. Onset and offset of role shift are indicated by a reverse apostrophe (left single quote, grave accent) and indication of the person represented by the role shift is indicated in parentheses:
'RS(person) SIGN ... '
For example, a child role-shifts into his mother to indicate that she prohibited him from having a cookie:
| *CHI: | WANT COOKIE 'RS(mom) NO ' . |
Note: 'RS is in capital letters, since it is a meaningful element.
It is often essential to know where signers direct their gaze while signing. Gaze direction is indicated by an asterisk (*) and an indication, in lower case, of the object of gaze; e.g.:
| *mot | looks at mother |
| *book | looks at book |
Gaze direction is indicated only when the transcriber considers that it is relevant to analysis of the interaction. Special notations are used to indicate a recipient's view of particular signs, indicated by backslashes (\). Such information is especially important for assessing a child's comprehension.
| \- SIGN \ | recipient does not see SIGN |
| \q SIGN \ | unsure whether recipient sees SIGN |
Note: For any modification other than the extension of neutral signing space, insert a %com line to explain how the SIGN is modified.
| \@ SIGN \ | signer modifies location of SIGN outside its normal location |
| \@\- SIGN ... | simultaneous onset of two recipient markings (Gaze and modification markings may occur simultaneously) |
| ... SIGN \\ | simultaneous offset of two recipient markings |
| ... SIGN @\ -\ | sequential offset of two recipient markings |
If part of an utterance consists of non-signed but meaningful activity, notations of such activity is included as main line commentary in square brackets, as follows:
| [%ges: identification] | identifies the gesture and lexical interpretation for gestures occurring without the use of any object or prop, and/or outside of typical signing space (e.g., [%ges: write]) |
| [%act: identification] | identifies the activity that replaces some or all of an utterance, performed with or on some object (e.g., [%act: throws doll]) |
Various means are used to get the attention of the recipient. These devices are indicated by @ag. The @ag is part of the utterance line. The following attention getting-devices have been identified:
| t@ag | tap on person |
| w@ag | wave at person |
| g@ag | grab person |
| f@ag | touch face of person |
| p@ag | pound on surface |
An error is indicated by [*], in the standard CHAT fashion. If an entire utterance is un-grammatical, with no localizable error within the utterance, [*] is placed at the beginning of the line. If an error can be localized, the intended SIGN is given in square brackets with an equal sign, followed by [*]; e.g.:
| *CHI: | DAD CHAIR [= SIT] [*] HERE . |
Explanations of errors are given on a %err dependent tier, using codes including the following:
| $dir | directional error |
| $cl | classifier error (wrong classifier used) |
| $hs | handshape error |
| $lex | sign error (wrong sign used) |
| $loc | location error |
| $mvt | movement error |
| $po | palm orientation error |
| $syn | syntax error (ungrammatical utterance) |
If there is more than one error on a line, separate each explanation with a semicolon bounded by spaces. For example, if a child used the wrong handshape for DAD and signed CHAIR with a movement pattern that means SIT, the transcription and error coding would be as follows:
| *CHI: | DAD [*] CHAIR [= SIT] [*] HERE . |
| %err: | DAD $hs ; CHAIR $mvt = SIT ; |
In words with multiple components, use the [*] with a number to indicate which component has the error. However, if the child confuses src and gol, mark this as an error with the * symbol only on the first component, and then add [*] at the end of the classifier construction to mark the error, e.g:
| *CHI: | BOY (grab)-cl'HOLD-mvt'CLOSE-src'3*-gol'1 [*] . |
| %err: | src'3-gol'1 $agr = src'1-gol'3 ; |
| [*q] | possible error |
| [*u] | unspecified error somewhere in the utterance, but not tied to one particular SIGN |
Note: For phonological errors, the utterance line should represent what is semantically meant by the sign, i.e. what the addressee should get from the message, and not a phonological representation of what the signer signed. This issue is currently under discussion.
If a turn consists of a definite, but non-signed response, use the standard CHAT convention of beginning the utterance line with zero (e.g., *CHI: 0). The zero is used when the interlocutor uses only an attention getter (*CHI: 0 t@ag), action, and/or gesture.
If a sign is continued or held from the previous utterance, the tilde (~) is used to indicate that this sign was not created anew in the following utterance, e.g. ~SIGN.
Interruption and continuation after interruption are coded in the standard CHAT fashion. For example:
| *MOT: | WANT +/ . |
| *CHI: | PNT_3(on_book) . |
| *MOT: | +, READ BOOK . |
Standard CHAT conventions are used for retracing and retracing with correction, as in these examples:
| *CHI: | <MOTHER> [/] MOTHER LEAVE . |
| *CHI: | <BEAR> [*] [//] BEAR . |
| %err: | BEAR $mov |
The following dependent tiers are being developed for transcription of sign language:
| %act | modifies the preceding utterance line, describing actions of signer or recipient that are necessary for the understanding of the transcription |
| %att | describes participants' attention (e.g., CHI and MOT not attending to one another) |
| %gls | gloss (written-language paraphrase for particular complex utterance lines) |
| %ges | phonological description of gesture |
| %pho | phonology |
| %mor | morphology |
| %sem | semantics |
| %spa | speech act |
This is a segment of joint drawing activity between a mother and her daughter of 2;8 (data of Nini Hoiting):
| *MOT: | PNT(nh)(with_pen_on_slate) < FATHER > [>] . |
| *CHI: | < MAN > [<] FATHER PNT_3(on_slate) . |
| *MOT: | t@ag MAN . |
| *CHI: | MAN . |
| *MOT: | PNT_3(at_drawing) . |
| *CHI: | PNT_3(at_drawing) PNT_1 [%ges: long ears] . |
| *MOT: | PNT_2 PNT_3(on_slate) . |
| *CHI: | PNT_3(on_slate) GRANDPARENTS . |
| *MOT: | GRANDMOTHER . |
| *CHI: | PNT_3(on_slate)(*N) < A_LOT(*N) > [>] . |
| *MOT: | <^dis'CONF A_LOT FACES > [<] A_LOT FACES ^ . |
This is a segment of book reading between a mother and her daughter of 1;9 (data of Reyna Lindert):
| *MOT: | t@ag(*2) w@ag ^opr'WHQ SEE WHAT(1h) ^ ? |
| *CHI: | MOUSE(*N) . |
| *MOT: | t@ag g@ag(nh): \- ^opr'WHQ WHAT(1h) ^ \ ? |
| *CHI: | 0 [%act: lifts panel in book] *mot . |
| *MOT: | ^opr'WHQ WHAT(1h) ^ ? |
| *CHI: | < PNT_3[on book] *m > [>] . |
| *MOT: | < ^opr'WHQ WHAT(1h)(*2) ^ > [<] ? |
| *MOT: | ^opr'YNQ \- CAT \ t@ag(nh) CAT PNT_3(at_cat_in_book) ^ < CAT > [>] ? |
| *CHI: | < ^opr'NEG 0 ^ > [<] [%ges: don't know/not me] . |
| %ges: | open 5s, wrists rotate out |
| *MOT: | ^opr'WHQ WHAT(1h) ^ ? |
| *CHI: | <BEAR> [*] [//] BEAR . |
| %err: | BEAR $mvt |
1 See the CHILDES website for the full set of required conventions
(The North American site, organized by Brian MacWhinney, is at Carnegie-Mellon University: http://childes.psy.cmu.edu. There is a European mirroring site organized by Steven Gillis in Antwerp, Belgium: http://atila-www.uia.ac.be/childes/chat.pdf. There is a Japanese mirroring site organized by Hidetosi Sirai in Chukyo: http://jchat.sccs.chukyo-u.ac.jp/CHILDES/chat.pdf.)