Feature-Based Attention Samples Stimuli Rhythmically

Attention supports the allocation of resources to relevant locations and objects in a scene. Under most conditions, several stimuli compete for neural representation. Attention biases neural representation toward the response associated with the attended object [1, 2]. Therefore, an attended stimulus enjoys a neural response that resembles the response to that stimulus in isolation. Factors that determine and generate attentional bias have been researched, ranging from endogenously controlled processes to exogenous capture of attention [1–4]. Recent studies investigate the temporal structure governing attention. When participants monitor a single location, visual-target detection depends on the phase of an ∼8-Hz brain rhythm [5, 6]. When two locations are monitored, performance fluctuates at 4 Hz for each location [7, 8]. The hypothesis is that 4-Hz sampling for two locations may reflect a common sampler that operates at 8 Hz globally, which is divided between relevant locations [5–7, 9]. The present study targets two properties of this phenomenon, called rhythmic-attentional sampling: first, sampling is typically described for selection over different locations. We examined whether rhythmic sampling is limited to selection over space or whether it extends to feature-based attention. Second, we examined whether sampling at 4 Hz results from the division of an 8-Hz rhythm over two objects. We found that two overlapping objects defined by features are sampled at ∼4 Hz per object. In addition, performance on a single object fluctuated at 8 Hz. Rhythmic sampling of features did not result from temporal structure in eye movements. •Feature-based attention unfolds rhythmically•Performance on two overlapping objects fluctuates rhythmically at 4 Hz•Performance on a single object fluctuates rhythmically at 8 Hz•Rhythmic sampling of one or two objects occurs in the absence of eye movements Re et al. demonstrate that feature-based attention operates rhythmically. A common sampler is distributed over the number of relevant items: performance on a single item fluctuates at 8 Hz and performance on two overlapping items fluctuates at 4 Hz per item. This phenomenon is independent of eye movements and is not specific to spatial selection.