Sampling-based representation of uncertainty during hippocampal theta sequences

Efficient planning in complex environments requires that uncertainty associated with current inferences and possible consequences of forthcoming actions is represented. Representation of uncertainty has been established in sensory systems during simple perceptual decision making tasks but it remains unclear if complex cognitive computations such as planning and navigation are also supported by probabilistic neural representations. The statistical structure of the sequential neural activity during theta oscillation in the hippocampus is consistent with repeatedly performing inference of current and prediction of future positions using a dynamical generative model. Since uncertainty increases gradually with the time span of the prediction in dynamical models, investigating neuronal activity during hippocampal theta sequences offers a unique possibility to identify neuronal correlates of uncertainty. We have established a framework to directly contrast competing hypotheses about the way probability distributions can be represented in neural populations. Importantly, new measures were developed to dissociate alternative coding schemes and to identify their hallmarks in the population activity. Reliability and robustness of these measures was validated on multiple synthetic data sets that were designed to match the statistics of neuronal and behavioural activity of recorded animals. Next we tested our framework in rats performing an open-field navigation task. In contrast with prominent theories, we found no evidence of encoding parameters of probability distributions in the momentary population activity. Instead, uncertainty was encoded sequentially by sampling motion trajectories randomly in subsequent theta cycles from the distribution of potential trajectories. Finally we confirm previous results in simpler mazes by showing that the trajectories sampled in subsequent theta cycles tend to be anti-correlated, a signature of efficient sampling algorithms. These results demonstrate that the brain employs probabilistic computations not only in sensory areas during perceptual decision making but also in associative cortices during naturalistic, high-level cognitive processes.