In order to form stable perceptual representations, populations of sensory neurons must pool their output to
overcome physiological noise; selective attention is then required to ensure that behaviorally relevant stimuli
dominate these ‘population codes’ to gain access to awareness. However, the role that attention plays in shaping
population response profiles has received little direct investigation, in part because most traditional
neurophysiological methods cannot simultaneously assess changes in activity across large populations of sensory
neurons. Based on single-unit recording studies, current theories hold that attending to a relevant feature sharpens
the population response profile and improves the signal-to-noise ratio of the resulting perceptual representation.
Here, we test this hypothesis using fMRI and an analysis approach that estimates the influence of feature-based
attentional modulations on population response profiles. We first derive orientation tuning functions for single
voxels in human primary visual cortex, and then use these tuning functions to sort voxels according to their
orientation preference. We then show that selective attention systematically biases population response profiles so
that behaviorally relevant stimuli are represented in the visual system at the expense of behaviorally irrelevant
stimuli. Collectively, the present results (1) provide a new approach for precisely characterizing feature-selective
responses in human sensory cortices and (2) reveal how behavioral goals can shape population response profiles
to support the formation of coherent perceptual representations.

link to paper here