Why a hard search can make a bright distractor harmless
A useful way to picture this is a spotlight versus a floodlight. When the thing you're looking for is obvious, your attention spreads widely and processes the whole scene at once, in parallel. In that mode, a bright, distinctive object off to the side competes on equal footing and pulls your attention toward it. One account argues that this is exactly what happens when the target is salient: search proceeds in parallel, and the salient singleton "cannot be ignored nor suppressed but instead will capture attention" [1].
Flip the situation. If the target is dull and hard to tell apart from its neighbors, you need fine discrimination, so your attention narrows into a small window and you scan items one at a time, or nearly so. Salient objects outside that narrow window don't capture attention, according to this view, not because they were actively suppressed but because they were never attended to in the first place [1]. This reframes a lot of evidence that looked like suppression: in studies where search was likely serial or partly serial, the distractor may simply have been ignored rather than inhibited [1].
Suppression needs a heads-up: known distractors can be blocked, unpredictable ones usually can't
Active inhibitory control, the ability to deliberately damp down a known source of distraction, does appear to work, but only under limited conditions: the distractor's features must be known in advance. Distractors with unpredictable features cannot be ignored [2]. This is a crucial boundary. In everyday life, distractions are rarely predictable, which is why one line of research set out to test whether unpredictable distractors could ever be suppressed [2].
The answer from that work was largely no. Across three experiments, researchers made the search target less salient in different ways (highly similar shapes, complex array items, and a conjunction search where the target shared all its colors with non-targets). In every case, a color singleton distractor whose color unpredictably alternated was not suppressed but instead captured attention [2]. Notably, when the distractor's color was known in advance, it was suppressed, and reducing target saliency did not weaken that suppression [2]. The takeaway is that the machinery for suppressing a known distractor seems fairly independent of how hard the target is to find, but that machinery has little to work with when the distractor is a surprise [2].
Static versus moving distractors: some are easier to hold back than others
The type of distractor also matters. Most suppression research has used color singletons, so it was unclear whether other salient stimuli could be suppressed too. Using an eye-tracking version of an attentional capture task, one study compared a range of salient distractors and found that people could ignore a wide variety of them, but suppression was weaker and slower to develop for dynamic stimuli such as motion singletons than for static ones such as color singletons [3]. Adding a static salient feature to a moving distractor greatly improved suppression [3]. This suggests an inhibitory process is applied to all kinds of salient distractors, but it grips static features more readily than dynamic ones [3].
Getting distracted once may be how you learn not to be
There is also a learning angle. In three eye-tracking experiments with 120 college students, an initial instance of distraction occurred before salient distractors came to be suppressed, and when that initial capture was eliminated or weakened through experimental manipulations, the learned suppression of the distracting stimuli was greatly reduced [4]. The researchers propose that attentional capture can act as a learning signal that improves future attentional control, and that this learned suppression emerges rapidly [4]. In plain terms, the first grab of attention may be the experience that teaches the brain to block that distractor next time.
Inhibitory control also fluctuates with a person's state. In athletes, prior self-control exertion was associated with longer antisaccade latencies and greater inhibitory control costs, and during visual search the depletion group showed longer reaction times, delayed first fixation on the target, more fixations, and longer cumulative fixation time on distractors under an activation condition than a no-activation condition [5]. This is a latency-based cost in oculomotor response inhibition and greater difficulty suppressing salient distractors in that sample [5], a reminder that the same task can yield different capture depending on how taxed the person's control system is.
Sources used in this answer
The Attentional Capture Debate: When Can We Avoid Salient Distractors and When Not?
Capture by salient distractors can be avoided when the target is non-salient and search becomes serial, so distractors outside a narrow attentional window are ignored rather than suppressed; when the target is salient and search is parallel, the singleton captures attention.
Effects of Target Saliency on Inhibition of Salient Distractors
A color-unpredictable singleton distractor captured attention even when target saliency was reduced across three experiments, whereas a color-known distractor was suppressed and its suppression strength was unaffected by reduced target saliency.
Attentional suppression of dynamic versus static salient distractors
Participants could ignore a wide variety of salient distractors, but suppression was weaker and slower to develop for dynamic (motion) distractors than static (color) ones, and adding a static salient feature to a motion distractor improved suppression.
From Capture to Control: Initial Capture Increases Learned Suppression.
An initial instance of distraction preceded suppression of salient distractors, and eliminating or weakening that initial capture greatly reduced learned suppression, suggesting capture can serve as a learning signal for future attentional control.
Oculomotor response inhibition and distractor suppression following prior self-control exertion in athletes
In athletes, prior self-control exertion was associated with longer antisaccade latencies and greater inhibitory control costs, and with more difficulty suppressing salient distractors during visual search under an activation condition.