Interaction is a multimodal process in which visual speech conveys a substantial amount of information. Speech and gesture are biomechanically linked, and evidence suggests that both are controlled by overlapping motor systems. Consequently, vocal speech phenomena are intrinsically entangled with gestures. Previous work investigating whether the Lombard effect extends to manual gestures has found that speakers produce gestures with more submovements and longer hold times when speaking in noise. However, such effects could reflect deliberate communicative strategies rather than a genuine Lombard response.

In fact, Lombard speech is characterized not only as a communicative adaptation but also as a reflex, showing rapid, low-level modulation through sensorimotor feedback pathways. It occurs even in non-interactive contexts, where speakers rely solely on auditory and somatosensory feedback in the absence of listener feedback. Additionally, it scales continuously with stimulus intensity rather than appearing as a binary effect. Therefore, to determine whether gestures are truly part of the Lombard effect, we must determine whether these same key characteristics apply.

We aim to determine whether gestures in noise show Lombard-like characteristics in a behavioural study with around 30 participants. Participants will complete a story-retelling task in a sound-treated studio under three conditions: quiet, 75 dB, and 85 dB multitalker babble, presented through speakers. They will watch short silent cartoon clips and then retell each story freely, without an interaction partner or explicit mention of gesturing.

Speech will be recorded via a lapel-mounted microphone to verify the presence and magnitude of the vocal Lombard response and enable automatic transcription. Gestural kinematics will be captured with wrist-mounted inertial measurement units (IMUs) and complemented by AI-based video motion tracking. From these signals we will extract indices of gesture velocity, submovements, hold duration, and spatial amplitude. Gesture occurrences will also be annotated by an expert coder assisted by gesture-detection software. This design tests whether gestures scale with noise intensity in the absence of communicative feedback, indicating a reflexive, Lombard-like modulation rather than a deliberate communicative adjustment.

If speech and gesture share partially common motor control mechanisms, we expect a general motor gain in noise, emerging both as increased vocal effort and as enhanced gesturing effort, reflected in changes in velocity, vertical amplitude, hold duration, or submovements. This might be further modulated in an interaction situation through a feedback loop, accounting for idiosyncratic differences. Our findings would have implications for models of speech–gesture integration. Currently, there is no agreement on how these two modalities are integrated and we hope to shed light on this question. Our presentation will feature illustrations from recordings and deeper insights into the data collection.

Funding: This study has emanated from research conducted with the financial support of Taighde Éireann - Research Ireland under Grant number 22/FFP-A/11059.