BREF and ICY, large speech data bases, have been designed and recorded in the LIMSI (Laboratoire d'Informatique et de Mecanique pour les Sciences de Tlngenieur). The facilities for recording were built in the laboratory for a hundred speakers in artificial conditions : sound proofed room for the best recording quality, reading on a screen, instructions to be followed etc. In these situations, we have observed the influence of other parameters in addition to those taken in our initial hypothesis. The motivation is one of them and it is relevant in the context of the experimentations. In BREF, we observed the behaviour of the operator (guidance and control) and of the speakers: reading, pronouncing and tiring. The speakers read aloud the newspaper style texts on a screen. We took into account the chronological complexity of the texts to read and the familiarisation to the instructions together with the speaker being tired. In spite of indirect human-human communication, the results were excellent because the speakers were hired for this purpose. This consideration is examined.

Keywords: vocal data bases, ergonomics, behavioural sciences.