The representations of information concerning the number, size, density and surface of sets of objects in a visual image are separable along the occipito-parietal cortex and independently modulated by attention.

Implementing neural changes associated with attention in a deep neural network causes performance changes that mimic those observed in humans and macaques.

Faced with multiple sources of sound, humans can better perceive all of a target sound's features when one of those features changes in time with a visual stimulus.

The sampling of visual space is not only warped by attention towards locations but also depends on attended features.

Low-frequency correlations among neurons in monkey visual area V4 impair the animal's ability to perform an attention-demanding task, suggesting a causal role of these fluctuations in perception.

A combined behavioural and electroencephalographic approach investigating the covert allocation of attention shows evidence for distributed periodic sampling away from a conscious visual image.

Spatial attention and saccadic processing co-ordinate to ensure that attention is available at a task-relevant location soon after the beginning of each eye fixation.

Prior expectations, induced by statistical learning, affect sensory processing only for attended stimuli, thereby constraining neurocomputational theories that cast perception as an inferential process integrating prior knowledge and sensory evidence.

To see the world stable across saccades, the brain compensates retinal shifts induced by the movements, pre-saccadic maps of sensitivity reveal that this process takes time and follows attentional dynamics.

Multivoxel pattern of fMRI data reveals how the brain distinguishes between relevant and irrelevant representations as representations adapt to the order in which they are required in multiple task sequences.