Attention

Attention is the set of processes by which the mind gives priority to some information over the rest. Sources distinguish three forms: selective attention, which picks out what to process among competing inputs; divided attention, which spreads processing across tasks done at the same time; and sustained attention, which keeps processing going over a long period (Smith, n.d.; Wickens, 2008; Mishler & Chen, 2024). The classic question about selection is how early unattended information is filtered out. Perceptual load theory, the most influential modern answer, holds that it depends on how demanding the main task is. That answer is now in doubt: its central effect has failed to replicate, and few individuals show it (Snowden & Gray, 2025; Marciano & Yeshurun, 2017). Its authors maintain that it is robust (Lavie et al., 2014). For sustained attention, researchers dispute whether performance declines because people run out of resources or because they disengage (Mishler & Chen, 2024).

History

Early and late selection

Three classic models place the selection of information at different points. Broadbent’s filter model places it very early: unselected information is not processed further. Treisman’s attenuation model holds that unattended information is weakened rather than blocked. Deutsch’s late-selection model holds that unattended input is processed all the way to meaning before selection happens (Smith, n.d.).

From one channel to multiple resources

Theories of divided attention moved through three stages. The first treated attention as a single channel, an all-or-none bottleneck that only one task can pass at a time. Kahneman (1973) replaced it with one undifferentiated pool of resources shared among tasks. Multiple-resource models then followed, because tasks interfere more when they compete for the same structures (Wickens, 2008).

Core concepts

Selective attention and perceptual load

Perceptual load theory holds that a perceptually demanding task uses up the capacity of perception, so that nothing is left over to process distractors; when the task is easy, spare capacity spills over to them (Lavie & Torralbo, 2010). The theory therefore predicts that distractors interfere less under high load (Snowden & Gray, 2025). Load has also been linked to awareness. Inattentional blindness is the failure to notice an unexpected object in plain view, and a 2022 meta-analysis of studies on it found that perceptual load significantly raises the chance of missing the object, with an odds ratio of about 2.5, although the overall estimate depended on which trials were excluded (Hutchinson et al., 2022).

Divided attention

Multiple-resource models predict well how much two tasks will interfere with each other. Correlations between predicted and observed interference are lower than perfect, however, which reveals additional resources the models do not yet include, and the models bear only partly on mental workload as people experience it (Wickens, 2008).

Sustained attention and the vigilance decrement

The vigilance decrement is the decline in detecting rare signals over a long, monotonous task, such as supervising a partially automated car. Two accounts compete. The overload account says the task depletes mental resources. The underload account says the task is too boring, so attention disengages. The two make opposite predictions for an added secondary task: it should make the decrement worse under overload and lessen it under underload (Mishler & Chen, 2024).

The attentional blink

When two targets appear in a rapid stream of images, people often miss the second if it follows the first closely; this gap is the attentional blink. Two direct replications of the original 1992 study, with 98 and 29 participants against the original 10, reproduced the effect. The odds of missing the second target at lag 2 were 8.77 and 10.41 times higher, against 6.75 in the original, though overall accuracy, the timing of the blink and the sparing of a target immediately after the first differed from the original (Grassi et al., 2021).

Questions, controversies, and new developments

Does perceptual load reduce distraction?

The challenges to load theory are of three kinds.

• Failed replications. Critics report that various studies failed to replicate the traditional load effects (Benoni & Tsal, 2013). In six experiments, Snowden and Gray found no reduction of distractor effects under high load (Snowden & Gray, 2025).
• Individual differences. A reanalysis at the level of individual participants found that only a minority behave as the theory predicts (Marciano & Yeshurun, 2017).
• Rival explanations. Load is vaguely defined, which makes the theory hard to test (Benoni & Tsal, 2013). High-load displays usually contain more neutral items, and reduced interference may come from these items diluting the distractor rather than from load; interference can even increase under high load (Benoni, 2025). Some studies attribute the efficacy of selection to how salient the distractor is rather than to load (Benoni & Tsal, 2013), and other accounts invoke attentional control settings or the relative activation of targets and distractors (Benoni, 2025).

Load theory’s proponents answer that perceptual-load modulation is robust across paradigms. They acknowledge boundary conditions: some effects are confined to visual motion, and load that is not perceptual can increase interference (Lavie et al., 2014). Against the dilution critique, Lavie and Torralbo report an experiment that they say supports load theory’s spillover account, while conceding that it is a single experiment; they call the dilution account underspecified (Lavie & Torralbo, 2010).

The question is contested. The one meta-analysis supports load effects on awareness (Hutchinson et al., 2022), while the effect on distraction, the theory’s original prediction, has the failed replications.

Overload or underload?

In a study of drivers in partially automated cars, an added secondary task produced evidence for both accounts: some results pointed to resource depletion and others to disengagement (Mishler & Chen, 2024). The same authors read their results as evidence against overload theories, in dispute with Greenlee and colleagues (2019), who favoured overload in a similar driving setting (Mishler & Chen, 2024).

Replication record

Broader connections

Attention and memory depend on each other: attention selects what gets encoded, and memory guides what is attended (Chun & Turk-Browne, 2007), in a cycle where memory sets what is relevant and attention decides what becomes memory (Turner & Sloutsky, 2024). Attention is also intertwined with working memory, in ways that depend on the specific variety of each (Awh et al., 2006). Whether attention can change what is seen, rather than only what is processed further, is part of the dispute over top-down effects in Perception, where one account explains action effects on perceived size by shifts in spatial attention (Kirsch et al., 2021).

Further reading

• Wickens (2008): the history of resource models of divided attention, by the author of the multiple-resource model.
• Lavie et al. (2014): load theory and its boundary conditions, by its proponents.
• Benoni and Tsal (2013): the main conceptual and methodological critique of load theory.