Video and audio of authentic interaction & naturally occurring discourse
CABank (TalkBank)
It houses dozens of corpora of naturally occurring adult conversations. Many corpora include synchronized audio or video recordings, detailed conversation-analytic (CA) transcripts, speaker metadata, and downloadable files for analysis using CLAN software. The repository includes conversations in English as well as numerous other languages and interactional settings.
Key features
- Audio- and video-linked transcripts
- A large collection of naturally occurring conversations: Multiple interaction types: casual conversation, business meetings, telephone calls, academic discussions, medical encounters, interviews, multilingual interactions.
- Standardized CA transcription conventions, searchable transcripts, downloadable datasets, compatible with CLAN software for corpus searches.
Why CABank is valuable for L2 pragmatics
It enables researchers to examine how pragmatic phenomena occur in authentic interaction rather than in experimental tasks. It is particularly useful for studying: speech acts (requests, apologies, refusals, compliments, invitations), turn-taking organization, repair and self-repair, politeness strategies, discourse markers, conversational routines, overlap and interruption, hedging and mitigation, sequence organization, interactional competence, and non-verbal behaviors (when video is available).
Because the recordings are synchronized with transcripts, researchers can also analyze prosody, pauses, laughter, gazes, gestures, facial expressions, and timing, all of which contribute to pragmatic meaning. Spoken corpora with aligned recordings make it possible to investigate pragmatic functions that are difficult to capture from text alone.
CABank English CANDOR Corpus
The CANDOR Corpus is one of the newest and largest multimodal conversation corpora available through TalkBank. It contains approximately 1,450 participants, over 1,650 recorded Zoom conversations, more than 850 hours of video, and roughly 7 million words of transcribed conversation. The corpus also includes extensive participant questionnaires and conversational metadata. Access requires approval because of participant privacy.
Key features
- Large-scale video corpus, high-quality synchronized transcripts, two-person spontaneous conversations, rich demographic information, pre- and post-conversation surveys
- Measures of politeness, rapport, perceived liking, conversational enjoyment, and social background
- Downloadable transcript files, compatible with TalkBank tools
Why the CANDOR corpus is valuable for L2 pragmatics
- Although the participants are primarily native English speakers, the CANDOR corpus provides exceptionally rich models of authentic conversational pragmatics.
- Researchers can investigate greetings, small talk, topic initiation, topic shifts, humor, conversational closings, politeness strategies, indirectness, conversational alignment, turn-taking, backchannels1, repair sequences2, and interpersonal rapport.
- Because conversations are video-recorded, the corpus also supports analysis of facial expressions, eye gaze, gesture, prosody, timing, and multimodal communication.
- The CANDOR corpus is especially useful for developing L2 instructional materials that go beyond grammatical accuracy to include interactional competence.
- Potential applications in L2 pragmatic research: developing authentic teaching materials, comparing learner interactions with native-speaker models, building datasets for AI-based pragmatic feedback, studying conversational routines, training annotation schemes, investigating multimodal pragmatics, creating examples for speech-act instruction, and interpersonal communication.
1Brief verbal or nonverbal signals produced by a listener while another person is speaking to indicate attention, understanding, agreement, interest, or encouragement without taking over the speaking turn—e.g., uh-huh, mm-hmm, yeah, right, I see, okay, really?
2Refers to any process by which participants correct or clarify communication problems—e.g., pronunciation, word choice, grammar, factual errors, hearing problems, misunderstandings, ambiguous references.