Neural dynamics and acoustic adaptations during the Lombard effect: evidence from EEG and dynamic causal modeling

Abstract The Lombard Effect (LE) is a vocal adaptation in which speakers involuntarily increase their vocal effort to preserve intelligibility in high-noise environments. This study explores acoustic and neurophysiological mechanisms that support LE in individuals with typical voices. Twenty-one participants produced 80 syllables under three conditions: Baseline (quiet), Lombard (noise at 80 dB SPL), and Recovery (quiet after five minutes of rest). Acoustic signals and electroencephalography (EEG) data were recorded synchronously, focusing on sound pressure level (SPL), H1–H2 (the difference in amplitude between the first and second harmonics), Cepstral Peak Prominence (CPP), Event-Related Potentials (ERPs) time-locked to the onset of self-produced vocalizations, and effective connectivity through Dynamic Causal Modeling (DCM). Results showed a significant increase in SPL during the Lombard condition compared to Baseline and Recovery. In this condition, H1–H2 values decreased and CPP increased, with no differences between Baseline and Recovery across the three acoustic measures. ERP analysis revealed a higher N1-P2 amplitude in the Lombard condition, associated with increased activations in frontal, limbic, and temporal brain regions. Bayesian model selection within the DCM framework indicated that the best-fitting model explaining the ERP data was a forward network from the primary auditory cortex (A1) to the temporal pole (TPO), inferior frontal gyrus (IFG), and parahippocampal gyrus (PHG), with modulatory connections highlighting feedback mechanisms. Within this network, the IFG and PHG seem to play a central role in error detection and feedback modulation, while the TPO supports auditory processing, together supporting the neural adjustments that sustain intelligible speech in noise. These results provide new insights into the cortical network underlying LE, emphasizing the adaptive mechanisms in speech production under noisy conditions.