Speech breathing and the Lombard effect.

Respiratory measurements were made using linearized magnetometers placed antero-posteriorly over the rib cages and abdomens of five healthy young women. Background noise was introduced over headphones simultaneously as "babble" presented binaurally at 55 dB ("moderate noise") and 70 dB ("high noise"). Speech during oral reading and spontaneous monologue was transduced with a microphone positioned near the lips, from which a speaking intensity signal (dBA) was derived. Subjects were instructed to speak during the noise conditions, but no instruction was given to alter speaking intensity. Compared with a "no noise" condition, the speaking intensities of all the subjects increased significantly for both speech tasks in the moderate and high noise conditions, thereby replicating the well-documented Lombard effect. No consistent trend of lung volume change was observed, in contrast to the linear increases in speech intensity as the noise level increased. For the higher speech intensities during the moderate and high noise conditions both initiation and termination lung volumes either increased or decreased. These preliminary findings suggest that when speech intensity is increased following the introduction of noise via headphones rather than by specific instruction to speak more loudly, speakers employ variable lung volume strategies for intensity control.