Toward a joint definition of neural-behavioral states

Understanding brain states requires more than the isolated study of either neural dynamics or behavior — it demands an integrative framework that acknowledges their complex, reciprocal interaction. For instance, this perspective led to developing key methods in the field (e.g., CEBRA, Schneider et al, Nature 2023) of shared latent states that underlie observable dynamics and behavioral outputs, offering a more comprehensive and biologically grounded definition of brain states. In this workshop, we bring together computational and experimental neuroscientists studying brain states with such an integrative perspective, and we hope that through this exchange, we find a clearer idea of a joint definition of brain-behavior state. We aim to stimulate discussions on the benefits and challenges of different frameworks and perspectives, and how they can facilitate a multi-faceted understanding of brain computation (consequently, behavior) in relation to neural dynamics. We believe that through this exchange of ideas, we can find new paths toward systemizing an approach to define brain state jointly based on neural dynamics and behavior (or at least the pros and cons of each approach). Moreover, the complementary perspectives of our speakers synergized with discussions in our workshop would promote new research directions and collaborations that will broaden the extent of brain research to gain a more holistic understanding of brain computation, dynamics, and behavior. Our speakers will discuss diverse perspectives to jointly characterizing states for information processing systems (both biological and artificial). These range from well-established (yet diverse) approaches such as task-optimized RNNs (Valerio Mante), neurophysiology (Dennis Nestvogel), and learning low dimensional embedding from joint neural and behavioral data (Xiaohang Yu’s CEBRA 2.0 from Mackenzie Mathis lab; Maryam Shanechi’s work, e.g., DPAD published in Nat Neuro 2024, and BRAID published in ICLR 2025), to less explored approaches that put more weight on behavior (recent work of Tatiana Engel and Meming Park). Our speakers also cover novel approaches that exploit recent developments in data-driven dynamical systems to introduce a common language to study states both in biological and artificial systems (Arman Behrad from Shervin Safavi’s lab). Moreover, our workshop benefits from recent developments in experimental neuroscience to have new perspectives on the joint definition of neural-behavioral state. First, through the presentation of Virginia Ruetten (from Misha Ahrens’s lab in Janelia campus), who intends to define the joint state, by recording neurons simultaneously with body ‘cells’ (which is a novel and unique approach) and to empirically find such a joint state in their data.

Second, we have invited Andrea Navas Olivé (from Cisvari lab in IST Austria), who studies Sharp Wave-Ripples (SWR), a well-known neural phenomena with rich connection to both sides of neural dynamics and behavior: SWR are characteristic transient, coordinated, neural activities observed in LFP (with signature is other scales) and have been shown to be involved in a wide range of cognitive processes supporting goal-driven behavior (from decision-making to memory consolidation and planning). Lastly, Mehran Ahmadlou (PI at Oxford Uni.) demonstrates how clear states in behavior (perseverative, exploratory, and disengaged states and switching between them), are controlled by thalamic neural dynamics (Ahmadlou et al. Nature 2025).