Behavior
Platform
The goal of the brain is to generate appropriate behavior according to the needs and goals of the animal. Carefully controlled behavioral tasks serve as the cornerstone for unraveling the complexities of brain function. However, significant challenges exist. For example, inherent variability of behavior is a major obstacle to establishing a causal link between the two. This is particularly challenging for behaviors such as learning and flexible decision making, where variability is a key feature. In addition, technology and tool development for each behavioral task require significant investment for benchmarking to ensure reliability and quality control. Furthermore, it is unlikely that a single task is well-suited for studying all aspects of brain function and behavior. The behavior team at the Allen Institute for Neural Dynamics is advancing technology and tools for a standardized, modular multi-task behavioral platform for head-fixed mice, making it possible to study brain function in across different behaviors at scale.
This modular design will enable the effortless adaptation of task logic and experimental instrumentation across diverse sensory modalities (visual, auditory, or olfactory), action modalities (directional licking, joystick, or locomotion), decision costs (water reward or effort), and environmental configurations (spatial distance or task geometry). Specific behavioral task workflows will draw from modular subsets of instrumentation. This system is designed to generate a wide range of complex behavioral tasks using a standardized framework. The accessibility of this system will facilitate new research questions on the neural basis of learning, decision making, and cognition.
We are interested in behaviors that give us access to neural dynamics across brain regions, with cell-type specificity over multiple timescales. This will require a careful balance of tasks that are well controlled that also respect the ethology by tapping into evolutionarily relevant behaviors across mammals, including mice. Foraging is a natural behavior where the efficient search and harvesting of resources through learning and adaptive decision making is essential for survival. Foraging engages the universal computation of balancing reward and cost, can be well-controlled in laboratory settings, and require rich neural implementation that engages brain-wide neural populations across multiple time. Furthermore, they are amenable to quantitative modeling, including rich theories from reinforcement learning and. The behavior team is developing a set of foraging behaviors to study cell-type specific dynamical interactions across brain regions as mice sample their environment and forage for resources.
Dynamic foraging
We have chosen a simple head-fixed, two-choice foraging task in mice, that have a single state (a go-cue), two actions (rightward or leftward licks), and a reward function (“dynamic foraging” task). The key feature of this task is that reward function evolves over time across the two available actions, and animals adaptively change their behavior to them. This task forms the basis of our first discovery projects.
Patch foraging
In stochastic natural environments where resources are distributed in patches, foraging animals sequentially sample their environment to choose between exploiting resources at their current location or leave to explore another, potentially superior one. Here, the animal’s goal is to understand the potentially hidden structure of the world based on repeated observations, and to adapt their behavioral strategy (policy) depending on sensory evidence, prior knowledge, and internal state. We leverage ethology in a virtual reality (VR) olfactory patch foraging task to study learning and decision making.
As next generation neurotechnology enables large-scale neural recordings during animal behavior, we also need new approaches to generate reproducible behavioral experiments at scale. We are developing engineering and procedural tools to standardize and automate training curriculum to study complex behaviors at scale.
Reducing a rich natural behavior into a controlled environment into the lab where we can take careful, quantified, and repeated measurements is challenging. It is unlikely that a single task is equally well-suited to study all aspects of learning and decision-making. Therefore, we are building infrastructure and modular tools to study a wide range of behaviors. We will make these resources available to the community as they mature and are benchmarked.
The Neuropixels platform uses pioneering technology for highly reproducible, targeted, brain-wide, cell-type-specific electrophysiology to record neural activity from defined neuron types across the brain.
The Molecular Anatomy platform combines innovative histology, imaging, and analysis techniques to map the morphology and molecular identity of neuron types across the whole brain.
The Fiber Photometry platform enables optical measurement of neural activity in live animals to study neural circuits' function and dynamics in behaving animals.
The Behavior platform uses advanced technology to implement a standardized, modular, multi-task virtual reality gymnasium for mice, with the goal to study brain function across different behaviors at scale.
