ENS, Salle Langevin, 29 rue d’Ulm, 75005 Paris
Cuttlefish (Sepia officinalis) can alter the patterns on its skin and its postures for camouflage and communication. In this thesis, we aimed to decipher the symbolic patterning of the cuttlefish for communication. We also hypothesized that these patterns may analogically reflect the internal states of the cuttlefish. Initially, we studied a new mode of communication in cuttlefish through arm movements. In a first behavioral experiment, we showed videos of cuttlefish performing this signal and measured the subjects' response to the presentation of this stimulus in both a normal and a flipped version. In a second "playback" experiment, we transmitted the trace of the vibration produced by arm movements, recorded beforehand with a hydrophone, and presented this stimulus to cuttlefish with two controls: the reversed trace and the trace with a randomized phase. A third experimental setup aimed to record the skin patterning of the cuttlefish during a specific state, the eating phase, allowing us to achieve a sufficient degree of decoupling from external stimuli to quantitatively analyze the changes in skin patterning induced by the cuttlefish's internal state. Methodologically, we used artificial intelligence techniques (computer vision and machine learning) to analyze the data. Our results demonstrate that arm movements are perceived in a multimodal way through vision and mechanoreception via the lateral line. The analysis of skin pattern dynamics revealed a statistical signature corresponding to theta oscillations, associated with reward states in vertebrates. Overall, this thesis highlights a new communication signal in cuttlefish involving the mechanoreception modality, opening the discussion on the function of the lateral line in communication. Our artificial intelligence methodology allows for the study of the temporal dynamics of skin patterns and can be applied more broadly to their statistical analysis.