Dynamic Variability in Speech

A Forensic Phonetic Study of British English


About the DyViS Database

Introducing the DyViS Database: a large-scale forensically oriented speech corpus

- 100 male speakers of Standard Southern British English aged 18-25
- speaking in a variety of styles
- simulation of forensic conditions
- approximately 1.25 hours of speech per speaker
- 20 speakers recalled for repeat of reading tasks after 2 months

1. Police interview
The speaker is interrogated in a mock investigation concerning a drug trafficking incident. Target words include many SSBE vowels in contexts convenient for measurement. Speech is constructed spontaneously using visual stimuli, including prompts to lie.

2. Telephone conversation with accomplice
Recording at studio quality plus intercepted external BT landline. Subject speaks with his accomplice, Robert Freeman, who elicits the same target words during a debriefing of the interview.

3. News report (reading)
The subject reads a news report of the alleged crime which he has committed. The same target words are used for comparison between conversational and read styles.

4. Controlled sentences (reading)
A large number of sentences with many SSBE monophthongs and diphthongs in nuclear non-final position.
- also /ju:/ and /wi:/ sequences
- 6 repetitions

All speakers are anonymous and talk about an imagined scenario, and so there are no problems with data protection. The database is available for all research purposes.

The recordings are useful for instance
- for research into individual speaker characteristics and forensic speaker identification
- for the investigation of spoken discourse
- for early-stage student projects (which e.g. can exploit the 'ready-made' sentences)
- as the most extensive resource available of current-day 'RP'

Click here to see the press release.

Obtaining the DyViS Database

The database is now complete and ready to order.

The complete database is available, as .wav files, associated Praat .TextGrid files with an orthographic transcript of the target speakers' turns in the spontaneous tasks, and the materials used in making the recordings, from the Economic and Social Data Service (ESDS) at


Users have to register with ESDS. Commercial researchers will be directed to Cambridge Enterprise to arrange payment of a fee. This does not apply to academic users.