Dynamic predictions: Oscillations and synchrony in top–down processing
Andreas K. Engel,Pascal Fries,Wolf Singer
DOI: https://doi.org/10.1038/35094565
IF: 38.755
2001-10-01
Nature Reviews Neuroscience
Abstract:Key Points Classical theories of sensory processing view the brain as a passive, stimulus-driven device. Newer theories, by contrast, view perception as an active, selective process controlled by top–down influences that include expectations and predictions, derived from experience, attention and working memory. The temporal binding model assumes that synchrony between distributed neurons is required for object representation, response selection, attention and sensorimotor integration. In a 'dynamicist' view of top–down influences on perception, synchrony generated intrinsically by interactions between higher and lower cortical areas could strongly influence perception, enhancing some representations and suppressing others. The firing rate and temporal response properties of neurons can be altered by attentional processes, working memory and behavioural context. For example, attention can enhance synchrony and/or gamma-band oscillations in neurons representing the attended stimulus. Spontaneous fluctuations in ongoing activity could represent not noise, but rather 'bias signals' that prime certain stimulus-evoked responses to allow rapid selection among inputs. Self-generated activity fluctuations during the preparatory period can also predict, for example, the direction, latency or speed of a movement. Contextual modulation can arise from networks of frontal, parietal and limbic areas as well as from sensorimotor areas. These areas can represent information related to goal definition, action planning, working memory and selective attention. Assemblies of neuronal populations implementing these aspects of top–down modulation could then entrain assemblies involved in the representation of new stimuli. A new concept of 'top–down' is proposed in which large-scale dynamics, expressing contextual influences and stored knowledge, can influence local processing. Rather than an anatomical hierarchy, this model proposes that any area could, in principle, modulate activity in any other area to which it is connected.
neurosciences