Quality of life, functions, bradykinesia, and dyskinesia in patients with schizophrenia.

An efficient recognition system should be able to prioritize the information that will constrain or inform perceptual representations and, eventually, be accumulated or retained. One way to facilitate this is to continuously generate predictions about upcoming inputs and to retain information—that is, revise beliefs—when these predictions are violated. Prediction is considered central in this account of the brain, by building a generative model of the world to minimize prediction error when sampling the sensorium. In predictive coding formulations, predictions are transmitted in a top-down manner, and, when there is a mismatch between the predicted and the observed input, a bottom-up prediction error is returned to update or revise the source of predictions at a higher hierarchical level. Crucially, prediction errors are weighted by the level of uncertainty (that is, their precision) associated with the given context, balancing top-down and bottom-up information streams to scale the influence of prior predictions and sensory evidence, respectively. This is sometimes framed in terms of precision-weighted prediction errors that instantiate the Kalman gain in Bayesian filtering formulations of predictive coding. The implicit encoding of uncertainty lends a dual aspect to predictive processing that encompasses both the predictions of a particular sensation and predictions of its predictability (that is, precision) that modulate the influence of the ensuing prediction error. Notably, these generative properties of predictive processing rely on ongoing integration of sensory inputs with internally generated, experience-dependent sequences, and are therefore thought to involve hippocampal–neocortical interactions. A hippocampal role in prediction is likely related to its function in extracting statistical regularities that can be applied to new situations. Therefore, the hippocampus should represent the expected information gain of an event before it occurs—as a function of predictability—and estimate the validity of the prediction upon observation of the event. Indeed, the hippocampus has long been postulated to hold a cognitive map that is used to form predictions about upcoming inputs. Such predictive information follow successor-like representation, for example, in place cell firing. The cortex may also have its own predictive role through communication across deep layers feeding-back predictions to the superficial layers of preceding regions in the processing hierarchy. Mechanistically, predictions and their violation have been associated with spiking activity and oscillatory dynamics in the cortex. Specifically, gamma-band activity has been associated with bottom-up prediction errors (reflecting surprise signals from primary sensory cortices) and alpha/beta oscillations with top-down predictions (from higher-level regions such as prefrontal cortex). However, the mechanism through which predictions of future sensory inputs are generated in the hippocampus and communicated to the cortex is still unclear. Irrespective of these mechanisms, they should manifest as a differential modulation of prediction–error responses in the visual cortex, depending on the predictability of the current context, which we hypothesize is itself recognized and broadcast through hippocampal processing. Specifically, when upcoming stimuli are unpredictable, they are inherently informative in the sense that they resolve uncertainty when observed (technically, they have a greater expected information gain). This leads to the hypothesis that the hippocampus has a role in precision weighting by modulating the electrophysiological correlates of prediction errors—that is, event-related gamma activity in the visual hierarchy—as a function of predictability (or entropy).