Feature-based attention: effects and control

Abstract

Feature-based attention prioritizes the processing of non-spatial features across the visual field. Classical studies revealed a feature-similarity gain modulation of sensory neuron's activity. While early studies that quantified behavioral performance have provided support for this model, recent studies have revealed a non-monotonic, surround suppression effect in near feature space. The attentional suppression effects may give rise to a highly limited capacity when selecting multiple features, as documented by studies manipulating the number of attended features. These effects of feature-based attention are likely due to attentional control mechanisms exerting top-down modulations, which have been linked to neural signals in the dorsal frontoparietal network. The neural representation of attentional priority at multiple levels of the visual hierarchy thus shape visual perception and behavioral performance.

Figure 1.

Experimental paradigms to study feature-based attention, with example stimulus displays. A). A stimulus display in a visual search experiment. The task is to find the letter “T” among rotated T’s and L’s. B). A stimulus display in a feature-based attention experiment, showing two spatially superimposed sets of colored dots. Participants can be instructed to attend to a specific color. C). Another stimulus display used in feature-based attention experiment. Participants are instructed to attend to either the upward and downward motion in the right stimulus and ignore the left stimulus, while perceptual and neural effects of the left stimulus are measured. This protocol is often used to demonstrate the global spread of feature-based attention.

Figure 2.

A). Feature cueing effect as a function of the similarity between cued and target feature, measured as angular offset on a color wheel (adapted from [16]). The monotonic decline is consistent with feature-similarity gain model. B). Performance as a function of the similarity between two attended colors (adapted from [18]), showing a non-monotonic surround suppression effect. C). Detection threshold in an experiment manipulating the number of feature precues (zero, one, two). Note lower threshold indicates better performance (adapted from [31]).

Figure 3.

A). Key brain areas in the visual cortex and frontoparietal network, shown on an inflated right hemisphere. Visual areas include V1, ExS (extrastriate cortex), MT+. Frontoparietal areas include IFJ (inferior frontal junction), FEF (frontal eye field), IPS (intraparietal sulcus). IPS1–4 are topographic areas defined in independent mapping protocols. aIPS is the anterior portion of IPS that is often activated in attention tasks. B). Decoding of attended color in visual and frontoparietal areas. All areas showed significant above-chance decoding accuracy (adapted from [42]). C). Effect of TMS to posterior IPS on a task requiring feature selection (attention) and a control task (baseline, adapted from [45]).