Learning to reach in the sensorimotor loop, and the required neural dynamics, can be potentially explained by simple principles.

Structural representations of sentences generated by deep language models correlate with human listeners' behaviours and neural activity, providing a quantifiable framework to uncover the neural dynamics underpinning incremental speech comprehension.

For brain imaging to be useful despite its limitations in measuring neural activity, the neural code must be smooth both in a traditional sense and functionally.

During behavior, many neurons do not have classic trial-averaged responses to behaviorally relevant stimuli, but can still have activity and population dynamics related to stimulus and behavioral choice on single trials.

Six- to twelve-month old infants, who have little linguistic or object experience, classify objects by relying on a invariant representation of global shape known as the shape skeleton.

A recurrent network using a simple, biologically plausible learning rule can learn the successor representation, suggesting that long-horizon predictions are computations that are easily accessible in neural circuits.

A close approximation to the successor representation is learnt by a simple spike-time-dependent learning rule between cells undergoing theta phase precession.

Prediction of future input explains diverse neural tuning properties in sensory cortex.

The provided framework based on biological models and clinical literature can aid in the optimization of visual cortical prostheses through computational or behavioral simulation experiments, serving as a flexible tool for computational, clinical, and behavioral neuroscientists working on visual neuroprosthetics.

Computational modeling offers an explanation for why animals learn more quickly or slowly when their environment becomes more variable or stable.