The Lombard effect

How does it work? Although the adjustment of vocal intensity happens involuntarily when background noise levels change, the phenomenon is not truly a reflex. Much of what we do know about how the Lombard effect works at a neural level comes from comparative work on non-human primates and other mammals. From these studies we learn that the essential circuits responsible for the Lombard effect are located in the brainstem. Specifically, sets of audio-vocal neurons in the periolivary region and the pontine reticular formation are the most likely candidates for the integration of vocal production and auditory perception that is necessary for the Lombard effect. As mentioned above, however, the Lombard effect is not a true reflex, in that it is not controlled by a simple reflex arc. One clue that higher cortical areas are involved is that the effect, although involuntary, can be modulated by social context and can be inhibited with training involving feedback from a different sensory modality. Although the Lombard effect is robust and simply instructing a speaker to keep their voice level constant does little to inhibit it, when speakers are provided with visual feedback displaying their vocal intensity in real time, it is possible to train a speaker to inhibit the rise in voice amplitude. While the term ‘Lombard effect’ generally describes only the change in vocal amplitude, the effect is very often accompanied by a suite of other vocal changes, including a rise in fundamental frequency, a flattening of spectral slope (or ‘tilt’), and an elongation of signal duration. This collection of related vocal adjustments in response to noise is collectively referred to as ‘Lombard speech’ in humans. The degree to which these other traits are coupled to the rise in amplitude can depend on the type and context of the vocalization. For example, in human speech, vowels are more likely to be elongated in Lombard speech, while consonants usually are not. In humans, males tend to exhibit a more dramatic Lombard effect than females do, and speakers in general have stronger Lombard effect-related vocal changes when they are involved in communicative interactions than The Lombard effect