- Atlas
- Social Sciences
- Attention
Attention
Attention is the mind's way of deciding what to process and what to ignore.
11 frameworks, from 1879 to today · Cognitive Science · Social Sciences
- Introspective Psychology of Attention 1879-1915
- Broadbent's filter model of attention 1958-1975
- Attenuation Theory 1960-1980
- Late Selection Theory 1963-1980
- Resource Theory 1973-1995
- Computational and Connectionist Models of Attention 1980-Present
- Spatial Orienting and Attentional Networks 1980-Present
- Selection for Action 1987-2010
- Biased competition theory 1995-Present
- Perceptual load theory 1995-Present
- Predictive Processing 2005-Present
Overview
Attention is the mind's way of deciding what to process and what to ignore1. At any moment the senses deliver far more information than the brain can fully analyse, and behaviour demands that some of it be prioritized: the voice you are listening to over the hum of the room, the brake light ahead over the billboards beside it. Cognitive science studies attention as a set of mechanisms—selection, filtering, orienting, sustaining, dividing, and switching—that allocate limited processing to some inputs, thoughts, or actions rather than others1. The field matters because almost every other mental capacity depends on it: perception, memory, learning, and skilled action all presuppose that something has been selected. Its history is a sequence of attempts to say precisely what is selected, at what stage, by what kind of mechanism, and under whose control.
Introspection and the First Experimental Psychology
Attention was a central topic of the introspective psychology that founded scientific psychology in the late nineteenth century. Wilhelm Wundt, whose laboratory at Leipzig opened in 1879, treated attention as the process by which contents of consciousness are apperceived—brought into clear focus—and distinguished passive from active attention. William James, in The Principles of Psychology (1890), gave the topic its most quoted formulation: everyone knows what attention is, he wrote, because it is "the taking possession by the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought." James distinguished sensory from intellectual attention and voluntary from involuntary attention, and he treated attention as a selection among competing streams of consciousness.
The method was introspection: trained observers reported the contents of their own experience under controlled conditions23. This produced rich descriptions but a problem the approach could not solve. Introspective reports were private, hard to compare across observers, and vulnerable to the very attentional processes being studied. When behaviourism swept Anglo-American psychology in the 1910s and 1920s, attention largely disappeared from serious research, since it was neither observable behaviour nor easily operationalized. It survived in Europe, notably in the work of the Dutch psychologist Franciscus Donders and later in Gestalt psychology, but the mainstream went quiet for several decades.
The Information-Processing Revolution and Filter Theory
Attention returned as a rigorous experimental topic with the information-processing revolution of the 1950s1. The immediate practical spur was wartime: radar operators and air traffic controllers had to monitor several channels at once, and engineers wanted to know the limits of human capacity. A key experimental tool was dichotic listening, in which different messages are played to each ear and the listener is asked to repeat one of them aloud—a task called shadowing1.
Colin Cherry, in the early 1950s, showed that listeners shadowing one message could report almost nothing about the unattended message beyond crude physical features such as whether it was speech or a tone, or whether the voice was male or female. In 1958 Donald Broadbent proposed the first major model. He argued that the senses feed a limited-capacity channel, and that a selective filter, operating on simple physical characteristics, admits only one input at a time to that channel145. The filter was conceived as an all-or-nothing gate. Broadbent's model explained why unattended speech was lost: it never reached the limited-capacity stage1. It was the first explicit statement of what came to be called filter theory, and it set the terms of debate for a generation.
Attenuation and Late Selection
Filter theory soon met evidence it could not accommodate. Neville Moray, in 1959, found that some listeners noticed their own name in the unattended ear—the "cocktail party" phenomenon. If the filter were an absolute gate, this should be impossible. Anne Treisman, working with Broadbent in the early 1960s, proposed attenuation theory: the filter does not block unattended inputs but turns their volume down. Attenuated messages can still be processed, especially if they are personally significant or if their content is primed, which is why a name can break through. Treisman later developed the idea that selection operates on features organized in a hierarchy, with some features requiring more processing than others.
A more radical challenge came from late selection theory, advanced by J. Anthony Deutsch and Diana Deutsch in 1963 and developed experimentally by Donald Norman in the late 1960s. On this view, all inputs are processed to the level of meaning, and selection happens only at the stage of response—what to report or act on15. The evidence included findings that unattended words could influence behaviour, for instance in priming tasks. The debate between early selection (selection before meaning is extracted) and late selection (selection after) became one of the defining controversies of the 1960s and 1970s. Its resolution was not a victory for either side but a recognition that selection can occur at multiple stages, and that the question is empirical: which stage dominates depends on the task and the load.
Capacity, Resources, and Effort
A parallel tradition asked not where selection happens but how much processing is available. Resource theory, associated above all with Daniel Kahneman's Attention and Effort (1973), treated attention as a limited pool of capacity that can be allocated flexibly among tasks. The pool is not a single undifferentiated quantity; Kahneman and others distinguished arousal from the allocation policy that distributes effort. The theory explained why difficult tasks interfere with one another more than easy ones, and why performance degrades under high demand. It also connected attention to effort and to physiological arousal.
Resource models were powerful but vague about what a "resource" is, and later work—notably by Christopher Wickens and by Navon and Gopher in the 1980s—proposed multiple, partly independent resources (for example, separate pools for verbal and spatial processing). The approach's lasting contribution was to make capacity and interference central experimental variables. Its weakness was that a resource pool is hard to observe directly, and the theory risked explaining any result after the fact by positing the right kind of resource.
Orienting, Control, and the Neural Turn
From the 1970s onward, attention research became increasingly physiological. Michael Posner's work on spatial orienting distinguished three operations—disengaging attention from a location, moving it, and engaging it at a new location—and used reaction-time tasks with spatial cues to measure them. Posner and colleagues linked these operations to distinct brain networks, and the covert orienting paradigm (attending to a location without moving the eyes) became a workhorse of the field. This line of work established attention as a set of control processes with identifiable neural substrates, and it connected the psychology of attention to the study of eye movements and to clinical conditions such as hemispatial neglect, in which patients ignore one side of space.
The neural turn accelerated with single-cell recording in monkeys. Robert Wurtz, Michael Goldberg, and others showed in the 1970s and 1980s that neurons in parietal and frontal areas respond more strongly to stimuli at attended locations, and that attention can enhance responses in visual cortex itself. The question of how attention modulates sensory processing became central, and it set up the theoretical developments that followed.
Selection for Action and Biased Competition
In the 1980s and 1990s, two related ideas reframed attention. The first, selection for action, argued that attention is fundamentally about guiding behaviour rather than about consciousness. On this view, selection is the process by which the perceptual system picks out the objects relevant to the actions an organism is prepared to perform6. The idea, developed by Odmar Neumann and others, dissolved some of the early-versus-late selection debate by shifting the question from "what reaches awareness" to "what gets used."
The second, biased competition, was formulated by Robert Desimone and John Duncan in the 1990s. It holds that stimuli compete for representation in the brain, and that attention biases this competition in favour of behaviourally relevant inputs6. The competition can be resolved by bottom-up salience or by top-down signals from frontal and parietal areas that enhance the responses of neurons representing the attended object and suppress their rivals. Biased competition unified psychological findings about selection with neural data on receptive fields and response modulation, and it remains one of the field's central frameworks.
Load Theory and the Return of Early Selection
In the 1990s and 2000s, Nilli Lavie's load theory offered a reconciliation of the early-versus-late selection controversy. Lavie argued that the stage at which selection occurs depends on perceptual load1. When a task imposes high perceptual load, spare capacity is exhausted and irrelevant stimuli are filtered out early; when load is low, spare capacity spills over and irrelevant stimuli are processed, so selection appears late1. The theory made testable predictions about distractor interference as a function of task difficulty, and it revived the idea that capacity limits, not a fixed filter, determine what gets through. It is a good example of how the field's older debates were not so much settled as reframed in terms of interacting variables.
Computational and Connectionist Models
Alongside the experimental work, computational modelling gave attention a formal vocabulary. Connectionist models of the 1980s and 1990s represented attention as the modulation of activation in networks of simple units, and they explored how top-down signals could bias processing without a central controller. Models of visual search, such as those developed by Treisman and by Jeremy Wolfe, formalized the distinction between parallel processing of simple features and serial processing of conjunctions of features. Later computational work, including Bayesian and predictive-processing accounts, treated attention as precision-weighting: a way of increasing the influence of reliable or relevant signals in inference. These models made attention a computational problem—how to allocate limited representational resources—rather than only a psychological one.
Predictive Processing and the Present
The most recent major framework is predictive processing, developed from the 2000s onward by Karl Friston, Andy Clark, and others. On this view, the brain is constantly generating predictions about its inputs and updating them in light of prediction errors7. Attention is reconceived as the process that weights prediction errors—determining which errors are trusted and therefore drive learning and perception. This gives attention a role in learning itself, not just in selecting among already-processed inputs, and it connects attention to Bayesian inference and to the broader computational turn in cognitive science.
Today the field is plural. Psychophysical and behavioural paradigms—cueing, search, dual-task, flanker tasks—remain central. Cognitive neuroscience uses fMRI, EEG, and single-cell recording to map attentional networks and their interactions. Computational models, from connectionist networks to Bayesian and predictive accounts, provide formal hypotheses. Clinical and applied work studies attention in neglect, attention-deficit conditions, ageing, and human-machine interaction. The old questions persist in new form: what exactly is selected, at which stage, by what mechanism, and how do bottom-up salience and top-down goals interact? The field has not settled them, but it has built an increasingly precise vocabulary—filtering, capacity, orienting, competition, load, precision—for asking them.
References
- Attention (introduction, 1950–1999, Filter, Selection, Load), Wikipedia
- Functional psychology (History), Wikipedia
- Structuralism (psychology), Wikipedia
- Perceptual load theory (History), Wikipedia
- Selective auditory attention (History), Wikipedia
- Biased competition theory, Wikipedia
- Predictive coding, Wikipedia
The frameworks of Attention
Each framework is a way of working the field has used: when it held, and the question it answers.
- 1879-1915
Introspective Psychology of Attention
How does the mind bring contents into clear focus?
- 1958-1975
Broadbent's filter model of attention
Where does the mind filter out unwanted input?
- 1960-1980
- 1963-1980
Late Selection Theory
Is selection before or after meaning is extracted?
- 1973-1995
- 1980-Present
Computational and Connectionist Models of Attention
How can selection be implemented in a machine or network?
- 1980-Present
Spatial Orienting and Attentional Networks
Which brain systems move and hold attention?
- 1987-2010
- 1995-Present
Biased competition theory
How do stimuli compete for representation in the brain?
- 1995-Present
Perceptual load theory
Does perceptual load decide when selection happens?
- 2005-Present
Predictive Processing
How does attention weight prediction errors in inference?
How the ideas connect
Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.
Attenuation Theory reacted against Broadbent's filter model of attention
Treisman's attenuation theory arose directly as a revision of Broadbent's all-or-nothing filter, prompted by Moray's finding that a listener's own name could break through from the unattended ear. She kept the filter but made it attenuate rather than block, so unattended material could still be processed if primed or significant.
Carried by Anne Treisman.
Late Selection Theory reacted against Broadbent's filter model of attention
Late selection theory was formulated by the Deutsches as a direct challenge to Broadbent's early filter, arguing that all inputs reach the level of meaning and selection occurs only at response. The early-versus-late selection controversy was defined by this rejection of the filter's placement.
Carried by J. Anthony Deutsch; Diana Deutsch; Donald Norman.
Resource Theory influenced by Broadbent's filter model of attention
Kahneman's resource theory took over the limited-capacity channel central to Broadbent's model but recast capacity as a flexible pool allocated among tasks rather than a single filter admitting one input. The question of capacity limits passed from filter theory into the resource framework.
Carried by Daniel Kahneman; Attention and Effort (1973).
Computational and Connectionist Models of Attention influenced by Broadbent's filter model of attention
Connectionist models of attention inherited Broadbent's filter as a selective gating mechanism, implementing selection as weighted connections or gating units in networks. The filter concept was formalized computationally rather than discarded.
Perceptual load theory influenced by Broadbent's filter model of attention
Load theory revived the early-selection intuition that a limited-capacity stage admits only some inputs, but made the amount of perceptual load determine whether selection is early or late. Lavie's framework thus carried forward Broadbent's capacity-limited filter as a limiting case under high load.
Carried by Nilli Lavie.
Late Selection Theory rival of Attenuation Theory
Attenuation theory and late selection theory competed through the 1960s and 1970s over where in processing selection occurs: Treisman placed a graded filter before meaning, while the Deutsches and Norman placed selection after semantic analysis. Both were responses to the same dichotic-listening and priming evidence.
Carried by Anne Treisman; J. Anthony Deutsch; Donald Norman.
Perceptual load theory influenced by Late Selection Theory
Load theory took the late-selection finding that distractors are processed to meaning and showed it holds only under low perceptual load, while high load produces early selection. Lavie's experiments were designed to adjudicate the late-selection claims of Deutsch and Norman.
Carried by Nilli Lavie.
Perceptual load theory grew out of Resource Theory
Perceptual load theory was framed as a resolution of the early-versus-late selection debate by making load the decisive variable, and it defined itself against the vague resource pool by specifying that perceptual load exhausts capacity at early stages. Lavie's experiments manipulated distractor interference as a function of load rather than positing a general resource.
Carried by Nilli Lavie.
Biased competition theory influenced by Computational and Connectionist Models of Attention
Biased competition drew on connectionist ideas of competition among representations, in which units inhibit one another and top-down signals bias the outcome. Desimone and Duncan's framework translated this competitive dynamics into neural terms.
Carried by Robert Desimone; John Duncan.
Biased competition theory influenced by Spatial Orienting and Attentional Networks
Biased competition incorporated Posner's finding that parietal and frontal areas modulate sensory responses at attended locations, treating those top-down signals as the bias in neural competition. The attentional network evidence supplied the mechanism by which competition is resolved.
Carried by Robert Desimone; John Duncan; Michael Posner.
Biased competition theory influenced by Selection for Action
Selection for action's claim that attention serves behaviourally relevant action shaped biased competition's emphasis on behaviourally relevant inputs winning the competition. Duncan and Desimone framed selection in terms of what the organism is prepared to do, not what reaches awareness.
Carried by Odmar Neumann; John Duncan.
Perceptual load theory influenced by Attenuation Theory
Load theory's claim that unattended stimuli are processed when capacity remains echoes Treisman's attenuation idea that unattended material is not blocked but reduced. Lavie treated the fate of distractors as depending on available capacity, a graded rather than all-or-nothing selection.
Carried by Nilli Lavie.
Computational and Connectionist Models of Attention influenced by Resource Theory
Computational and connectionist models took up the resource framework's question of how limited capacity is allocated, recasting it as explicit allocation algorithms and competition among processing units. The notion of a capacity limit passed into formal models as a constraint on processing.
Predictive Processing influenced by Biased competition theory
Predictive processing accounts of attention treat precision-weighting of prediction errors as the mechanism that biases competition among hypotheses, taking over biased competition's language of top-down modulation of sensory signals. The competition idea was recast as inference.
Beyond the field
Where Attention's ideas came from in other fields, and what they led to there: the strongest 12 of 13 links.
Introspective Psychology of Attention provoked Mental Testing and the IQ Tradition in Psychometrics
Binet's intelligence testing defined itself against the introspective attention experiments of Wundt's and Titchener's laboratories, replacing trained self-observation with standardized tasks scored by an examiner.
Carried by Alfred Binet.
Introspective Psychology of Attention influenced by Empiricist Philosophy of Mind in Philosophy Of Consciousness
Empiricist philosophy of mind, especially the associationist tradition running from Locke and Hume to Mill, supplied the introspective psychology of attention with its method of inner observation and its associationist account of how ideas combine. Wundt and Titchener built their experimental introspection on that inheritance.
Carried by John Locke; David Hume; John Stuart Mill.
Resource Theory influenced Attentional Styles Framework in Mental Skills Training
Nideffer's attentional styles drew on resource models of attention, treating attentional capacity as limited and allocation as a skill, when constructing his test of attentional and interpersonal style.
Carried by Robert Nideffer.
Computational and Connectionist Models of Attention influenced by Information processing and computational vision in Perception
Computational models of attention drew on the information-processing and computational vision tradition, especially Marr's levels of analysis and the modelling of early visual selection. The idea of attention as a computational stage operating on representations came from that programme.
Carried by David Marr.
Computational and Connectionist Models of Attention influenced by Artificial neural network modeling in Computational Neuroscience
Connectionist models of attention were built directly on artificial neural network modelling, using distributed units and learned weights to implement selection and competition. The neural network tradition supplied the architecture and learning rules for these attention models.
Carried by David Rumelhart; James McClelland.
Computational and Connectionist Models of Attention influenced by Computational Theory of Mind in Philosophy Of Consciousness
Computational models of attention took from the computational theory of mind the assumption that cognition is symbol manipulation or computation over representations, and cast attention as a computational process of selection and control. The framework of computation over representations carried directly into these models.
Carried by David Marr.
Selection for Action influenced by Ecological Psychology in Action And Motor Cognition
Allport's selection-for-action proposal drew on Gibsonian ecological psychology's insistence that perception is for action and that the environment affords behaviour. The action-oriented framing of selection came directly from that ecological tradition.
Carried by David Allport; James J. Gibson.
Selection for Action influenced by Ecological approach to perception in Cognitive Psychology
The ecological approach to perception, especially Gibson's direct-perception account, supplied selection for action with its premise that perceptual systems are geared to guiding behaviour rather than to passive representation. Allport built his argument on that action-centred view of perception.
Carried by David Allport; James J. Gibson.
Resource Theory influenced Attentional Styles Framework in Performance Psychology
The attentional styles framework borrowed from resource theories of attention the idea that attentional capacity is limited and can be allocated across tasks.
Carried by Daniel Kahneman; Attention and Effort (1973).
Resource Theory influenced by Arousal–Performance Theory in Mental Skills Training
Arousal-performance theory, from the Yerkes-Dodson law onward, supplied resource theory with the idea that arousal governs the availability and allocation of attentional capacity. Kahneman explicitly incorporated arousal effects into his capacity model of attention.
Carried by Daniel Kahneman.
Computational and Connectionist Models of Attention influenced Transformer in Deep Learning
Connectionist models of attention from the 1980s supplied the notion of learned attentional weighting that attention mechanisms in neural networks echo. The passage is conceptual rather than a direct technical derivation.
Computational and Connectionist Models of Attention influenced Attentional Focus and Practice Design in Motor Control
Computational and connectionist models of attention supplied the notion of limited-capacity selection that attentional-focus work on movement invokes when explaining why directing attention externally helps performance. The motor-learning framework draws on this attention literature conceptually rather than deriving from it.

Attention is a planet in Nousfarer
Its city has a tower for each of these 11 frameworks. Land, walk in, and climb one concept at a time.
Land on AttentionRelated fields
Fields its ideas are linked to
- Philosophy Of Consciousness 2 links
- Mental Skills Training 2 links
- Deep Learning 1 link
- Action And Motor Cognition 1 link
- Perception 1 link
- Cognitive Neuroscience 1 link
- Computational Neuroscience 1 link
- Cognitive Psychology 1 link
- Psychometrics 1 link
- Performance Psychology 1 link
- Motor Control 1 link