Information Processing Theory

At the very heart of cognitive psychology is the idea of information processing.

Information processing theory is a framework that views the human mind as a complex system through which information flows, much like a digital computer.

Computer-Mind Analogy

The development of the computer in the 1950s and 1960s had a profound influence on psychology.

It gave cognitive psychologists a compelling metaphor for the mind. That metaphor helped establish the cognitive approach as psychology’s dominant paradigm.

The analogy works simply.

A computer receives input, encodes it into a usable format, stores and manipulates the information, and produces an output. Cognitive psychologists proposed that the human mind works the same way.

In this framework, the brain corresponds to the computer’s physical hardware. The mind and its internal processes correspond to the software.

This split lets psychologists study the mind’s functional organisation somewhat independently of the neural hardware. Functional organisation means which operations occur, and in what order.

A program can be described without describing the transistors that run it.

The strongest version goes further: thinking is symbol manipulation. Newell and Simon (1976) defined a physical symbol system as a set of symbols combined into structures and transformed by processes.

Such a system, they argued, has “the necessary and sufficient means for general intelligent action.” On this reading, the brain is one kind of physical symbol system and a computer is another, so studying one illuminates the other.

Cognitive psychologists in this tradition treat people as active processors. They take in symbolic input, recode it, make decisions, store the results, and turn these internal states into behaviour.

A driver braking at a red light shows the stages at work.

• Input: light from the traffic signal strikes the retina.
• Encoding: the visual system extracts colour and shape and recognises a red traffic light.
• Retrieval: the cue calls up the learned rule “red means stop” from long-term memory.
• Decision: the driver decides to brake.
• Output: a motor command moves the foot to the pedal.

The behaviour is explained not by the stimulus alone but by the mental operations that intervene between stimulus and response.

Origins of the Computer Analogy

The analogy did not grow out of psychology alone. It was assembled in the 1950s at the crossroads of psychology, engineering, and the new science of computing. Three outside developments gave psychologists both a metaphor for the mind and a way to make mental models precise.

• Information theory: Shannon (1948) treated any message as information, measured in bits and sent through a channel of finite capacity. Coding and channel capacity entered psychology from here.
• The stored-program computer: Turing (1950) argued that a suitably programmed machine could, in principle, perform any well-defined symbolic operation. Von Neumann (1958) compared the computer’s logical organisation with the nervous system’s.
• Symbolic artificial intelligence: programs such as the Logic Theorist and the General Problem Solver proved theorems and solved puzzles by manipulating symbols (Newell & Simon, 1972).

Shannon’s ideas explain why Miller (1956) treated immediate memory as a limited channel. Like any transmission system, the human processor can hold only so much at once.

Within psychology, Broadbent (1958) made the analogy a research programme. His Perception and Communication proposed the first explicit flowchart model of the mind, described below. Neisser’s (1967) textbook Cognitive Psychology then consolidated the field, giving it its name and its information-processing charter.

Core Principles of Information Processing

• Coding: Sensory input becomes a usable mental format, much as a telecommunications system turns a voice into an electromagnetic code. The eye, for instance, converts light into electrical neural activity.
• Channel capacity: The system has finite limits: only so much information can be held and processed at once, a constraint that bears directly on attention and short-term memory.
• Serial and parallel processing: Information can be processed sequentially, one operation at a time, or simultaneously across multiple cognitive processes running in parallel.
• Storage and retrieval: Like a computer’s memory, the mind stores information and retrieves it when needed. Mental processes such as memory, perception, and attention carry out this work.

This finite channel capacity has a famous figure attached to it. Miller (1956) found that immediate memory holds only about seven items, plus or minus two, before it overflows.

The limit is a lasting one. Cognitive psychologists still cite this “magic number seven” as a benchmark for the mind’s limited channel.

The influence of this approach has extended well beyond its origins.

Cognitive psychology has influenced, and integrated with, many other approaches. Examples include social learning theory, cognitive neuropsychology, and artificial intelligence (AI).

Information Processing Models and Theories

Mental processes are unobservable, but flowcharts and box-and-arrow diagrams can model them precisely, showing how information passes between different systems. These models generate specific, testable predictions, evaluated through controlled experiments, computational simulations, and neuroimaging.

The information-processing model of memory is a series of stages, or boxes. Each box represents a stage of processing. Arrows indicate the flow of information from one stage to the next.

• Input processes are concerned with the analysis of the stimuli.
• Storage processes cover everything that happens to stimuli internally in the brain and can include coding and manipulation of the stimuli.
• Output processes are responsible for preparing an appropriate response to a stimulus.

A quintessential example of an information-processing model is the multi-store model of memory proposed by Atkinson and Shiffrin in 1968.

This model suggests information flows sequentially from sensory stores, into a limited-capacity short-term store via attention, and eventually into a practically unlimited long-term store via rehearsal.

Broadbent offers another example.

Donald Broadbent’s filter model of selective attention uses the concept of an information buffer, a temporary holding area for incoming information. It explains how the brain tunes out competing stimuli to prevent overload, representing the mind as a limited-capacity channel.

Broadbent’s Filter Theory (1958), inspired by the demands placed on air traffic controllers, proposed that attention operates as a bottleneck. Incoming sensory information is briefly held in a buffer. A selective filter then passes only one channel through to the central processor, chosen by physical characteristics such as pitch or location, while all other information is discarded entirely.

Later theorists kept Broadbent’s bottleneck but moved it. Treisman (1964) argued that unattended channels are weakened, or attenuated, rather than fully blocked. Deutsch and Deutsch (1963) went further, placing the point of selection later, near the response stage, after the brain has already analysed meaning.

The framework also corrected itself from within. Atkinson and Shiffrin’s single, serial short-term store proved too simple. That prompted Baddeley and Hitch’s (1974) working memory model.

This model replaced the single box with an active system of several components. A central executive coordinates a phonological loop and a visuospatial sketchpad. The design accommodates evidence that people can run several short-term operations in parallel. The serial box became a partly parallel workspace.

A related refinement, levels of processing, shifted the emphasis from where information is stored to how deeply it is encoded.

A model counts as more than an analogy when it is specified in enough detail to predict behaviour. Bruce and Young’s (1986) face-recognition model is a standard example.

Its flowchart separates stages from structural encoding to name generation. It correctly predicts dissociable failures, where one stage breaks down while the others keep working.

Sternberg’s Memory-Scanning Study

The information-processing approach is defined as much by its method as by its computer metaphor: it infers the mind’s hidden processing stages from precisely measured reaction times. Sternberg (1966) supplies the classic demonstration.

This makes Sternberg’s task a template for the whole information-processing approach: an unobservable mental operation is inferred from a measurable behavioural signature. Later researchers questioned whether certain parallel models could produce the same linear pattern. Reaction-time data alone, it seems, cannot always pin down the underlying architecture.

Serial & Parallel Processing

• Serial processing is a cognitive operation in which each process must complete before the next can begin.
• Parallel processing describes a mechanism in which two or more cognitive processes occur simultaneously.

The distinction between these two modes is foundational to cognitive psychology, with implications for how we learn, how attention operates, and how the brain is physically organised.

Computer Analogy and Brain Architecture

The traditional information-processing approach initially assumed that human cognition was primarily serial. Early computers shaped this assumption: they could perform only one operation at a time, in a strict linear sequence.

Modern cognitive science has revised this picture considerably.

The human brain is composed of billions of interconnected neurons and operates as a massively parallel structure, distributing operations across multiple systems simultaneously.

Functional neuroimaging data now indicates that parallel processing in the brain is the rule rather than the exception.

Controlled and Automatic Processing

How much a task relies on serial or parallel processing depends largely on how practised the individual is at performing it.

• Serial processing is associated with controlled processes: slow, deliberate, capacity-limited operations that proceed one step at a time and demand conscious attention.
• Parallel processing is associated with automatic processes: fast, simultaneous operations that place little or no demand on attentional resources and run largely outside conscious awareness.

A learner driver illustrates this clearly.

Steering, gear changes, and monitoring surrounding traffic each demand deliberate attention. A learner cannot easily do them at the same time, so the approach is serial and step by step.

An experienced driver, by contrast, performs all of these tasks in parallel, managing them seamlessly without conscious effort.

System 1 and System 2 Thinking

This distinction maps onto dual-process models of reasoning and judgement.

• System 1: fast, intuitive, and automatic, operating in parallel without deliberate effort.
• System 2: slower, analytical, and rule-governed, proceeding serially and demanding conscious engagement.

Most everyday cognition involves an interaction between the two, with System 1 handling routine processing and System 2 recruited when tasks exceed automatic capacity.

Bottlenecks and Limitations

Despite the brain’s capacity for parallel processing, multitasking has clear limits.

Research on the psychological refractory period demonstrates that early sensory and perceptual processing of two simultaneous tasks can overlap. The term describes the measurable delay in responding to a second task presented soon after a first.

A bottleneck tends to emerge at response selection, though. When the brain must decide how to respond to incoming stimuli, processing becomes serial regardless of how much parallelism preceded it. Parallel processing, in other words, does not extend indefinitely through the cognitive system.

Parallel Processing in Visual Perception

The interplay between serial and parallel processing has been studied extensively across specific cognitive domains.

In visual perception, the system relies heavily on parallel processing. Depth is computed from two parallel streams of input, one from each eye, while separate neural channels simultaneously handle fine detail, contrast, and colour.

Connectionism and Rival Frameworks

Not every cognitive scientist accepts the classical picture of a serial processor. The strongest challenges come from connectionism, which rebuilds the analogy, and from frameworks that reject it.

The Connectionist (PDP) Alternative

Connectionist models, also called parallel distributed processing (PDP) models, replace the serial processor shuffling discrete symbols with a brain-style network (Rumelhart & McClelland, 1986).

Knowledge is stored as patterns of activation across networks of simple, neuron-like units joined by weighted links. The units sit in layers: an input layer, one or more hidden layers, and an output layer.

Many units update at once, so processing is parallel. Learning works by adjusting the connection weights, classically through error back-propagation, which nudges the network’s output closer to the correct answer over many examples.

This shift matters in three ways.

• Learning from examples: networks can program themselves, whereas early symbolic models needed every rule written in by hand.
• Distributed knowledge: a concept is a pattern spread across the whole network, so damage degrades performance gradually (graceful degradation) rather than catastrophically, as brain injury often does.
• A brain-style computer: the analogy moves from the serial digital computer to the parallel neural network, matching a brain that computes across billions of interconnected neurons.

It also answers the seriality objection head-on. The assumption that cognition proceeds serially from one module to the next sits uneasily with a brain in which many regions are active together during a single task.

Connectionism did not overthrow the information-processing approach. It changed which computer the mind is compared to.

The two traditions still debate whether rule-governed thought, such as grammar and logic, needs genuine symbols. Some argue it can emerge from sub-symbolic patterns, and hybrid models use both.

Rival Frameworks

Other frameworks define themselves against the information-processing approach.

• Behaviourism: admitted only observable stimulus-response links. The computer analogy reopened the black box, showing that the processes in between could be modelled and tested.
• Embodied and enactive cognition: denies that cognition is disembodied symbol manipulation. Thinking is grounded in the body’s sensorimotor systems and its interaction with the environment. The mind is less like a computer than a skilled, situated activity.
• Predictive processing: models the brain as a prediction machine that generates top-down predictions and updates them against sensory evidence. It keeps computation but shifts the metaphor toward active, parallel, probabilistic processing.

The last of these is the most influential contemporary successor to the classical analogy, and its evidence is reviewed under Contemporary Research below.

Real-World Applications

The information-processing approach is not just theory. It shapes how psychologists build technology, teach students, and test the mind.

Artificial Intelligence and Human-Computer Interaction

Minds and machines trade ideas in both directions. Cognitive models inform artificial intelligence: the General Problem Solver, symbolic reasoning, and neural networks all began as attempts to mechanise human thought. AI returns the favour by supplying testable models of cognition.

Machines do not think like people.

The computer that beat the reigning world chess champion in 1997 evaluated up to 200 million positions every second, far more than any grandmaster considers. A shared outcome does not mean a shared process.

In human-computer interaction, designers apply what is known about attention, working memory, and response selection. They limit on-screen choices, chunk information, and respect the brain’s central bottleneck, so interfaces fit how people actually process information. Good interface design, in this sense, is applied cognitive psychology.

Education and Learning

Working memory has a strict channel capacity: the same limit Miller (1956) described as roughly seven items. Cognitive-load theory builds directly on this constraint.

Instruction should chunk material into fewer, larger units, avoid overloading the learner, and build long-term knowledge structures that lighten the load once a concept becomes familiar.

Findings on encoding and retrieval feed directly into study technique, and three methods stand out.

• Spaced practice: spreads learning out over time rather than cramming it into one session.
• Retrieval practice: self-testing strengthens memory more than simply re-reading notes.
• Elaboration: links new material to what a student already knows, giving the information more retrieval routes later.

Together, these techniques translate the theory’s own stages (encoding, storage, and retrieval) into concrete study habits.

Cognitive Assessment

Treating cognition as separable components licenses targeted testing.

Reaction-time tasks, working-memory span measures, processing-speed indices, and tasks like the Stroop task or dichotic listening all isolate one processing component at a time.

The same chronometric logic behind Sternberg’s memory-scanning task underlies much of this testing: a precisely measured reaction time can reveal something about an unobservable mental process. The principle scales.

These measures are used in intelligence testing, in assessing brain injury, and in tracking cognitive change across development and old age.

Because each component can be measured on its own, small deficits become visible long before they would affect everyday behaviour. A clinician can pinpoint which specific stage, input, storage, or output, has been disrupted, rather than relying on a single global score.

Human Factors and Ergonomics

Aviation, driving, and control-room design all depend on the same ideas.

Engineers use information-processing accounts of attention, workload, and dual-task interference to design displays, alarms, and workflows that respect human capacity limits and minimise error.

Air-traffic control is the very setting that inspired Broadbent’s model in the first place: a controller who must select one radio message from several competing channels. It remains a proving ground for these principles today.

A poorly designed cockpit display can overload a pilot’s limited attention, exactly as Broadbent’s model predicts. A serial bottleneck forms at the worst possible moment.

This connects directly to the central bottleneck already described above: a control-room operator, like anyone else, can only select one response at a time. Well-designed systems build in this limit rather than fighting it, however much parallel monitoring came before it.

Critical Evaluation

The approach has real strengths. It made the mind a legitimate object of experimental science again after behaviourism, without a return to subjective introspection.

Its models generate specific, falsifiable predictions. Sternberg’s (1966) memory-scanning task and Bruce and Young’s (1986) face-recognition model show this. It has also proved highly productive, organising research on attention, memory, perception, language, reasoning, and development, and supplying a shared vocabulary for intangible processes.

Before exploring each critique in detail, here are the main concerns raised about the information-processing approach:

1. Serial vs Parallel Processing: the models assume information is processed one step at a time, but dual-task research shows much processing happens in parallel.
2. Limits of the Computer Analogy: computers and human brains diverge in key ways, from emotion and motivation to biological plasticity.
3. Low Ecological Validity: the research rests on controlled lab experiments that may not reflect real-world cognition.
4. Oversimplified Processing: the earliest models assumed a rigid, step-by-step flow that ignores top-down influences on perception.

1. Serial vs Parallel Processing

• Serial processing effectively means one process has to be completed before the next starts.
• Parallel processing assumes some or all processes involved in a cognitive task(s) occur at the same time.

There is evidence from dual-task experiments that parallel processing is possible. Whether a task is processed serially or in parallel is hard to pin down. It likely depends on two things: the processes the task requires, and how much practice the person has had.

Practice matters. Parallel processing becomes more common as skill increases: a skilled typist thinks several letters ahead, while a novice focuses on just one letter at a time.

2. Limits of the Computer Analogy

Computers can be regarded as information processing systems insofar as they:

1. Combining information: they combine information presented with stored information to provide solutions to a variety of problems.
2. Limited capacity: most computers have a central processor of limited capacity, and it is usually assumed that capacity limitations affect the human attentional system.

However:

1. Processing style: the human brain has the capacity for extensive parallel processing, while computers often rely on serial processing.
2. Emotion and motivation: humans are influenced in their cognitions by conflicting emotional and motivational factors.
3. No true understanding: computers manipulate formal symbols based on algorithms and syntax but have no inherent understanding of those symbols.
4. Biological plasticity: human brains are open, plastic biological systems that constantly adapt to their environment. They re-create memories and structures rather than pulling static files from a hard drive.
5. Systematic bias: the computer metaphor assumes rational information processing. Humans, however, make systematic, predictable mistakes due to cognitive biases and heuristics.
6. Flexible coding: human coding is flexible and cannot be reduced to fixed bits, so an absolute channel capacity cannot even be defined for people.
7. Machine reductionism: input-output stages risk losing the whole person. The “boxes” may be treated as literal mental objects rather than theoretical conveniences.
8. Homunculus problem: flowchart models often end in a “central executive” that does the intelligent work. That risks infinite regress. What explains the inner controller’s own intelligence?

3. Low Ecological Validity

Most laboratory studies are artificial and lack ecological validity: how well findings generalise to real-world settings.

In everyday life, cognitive processes serve a goal. You pay attention in class to pass the exam. Laboratory experiments strip away this real-world context, isolating cognition from other motivational factors.

The data are easy to interpret. But they may not apply outside the lab.

More recently, ecologically valid approaches to cognition have been proposed, such as Neisser’s (1976) Perceptual Cycle.

Attention has typically been studied in isolation from perception and memory, even though the three operate as one interdependent system. The more successfully we isolate one part of cognition, the less our findings may tell us about everyday life.

4. Oversimplification of processing dynamics

Early information-processing models were highly inflexible, assuming that human cognition was strictly a bottom-up, serial process where information flows step-by-step from one distinct stage to the next.

Modern cognitive science recognizes that this is a gross oversimplification.

Human cognition heavily relies on top-down processing, where an individual’s prior knowledge, expectations, and context continuously influence and alter perception and memory.

How did you read the text in the triangle above?

Expectation (top-down processing) often overrides information actually available in the stimulus (bottom-up), which we are, supposedly, attending to.

The human brain also contains billions of interconnected neurons that operate via massively parallel processing, carrying out multiple cognitive operations simultaneously. This contradicts the rigid serial assumptions of early computer models.

Contemporary Research

Recent research does not defend the original computer analogy so much as re-engineer it. Reviews now weigh the evidence unevenly: the strongest claims rest on meta-analyses and model-comparison studies, while more speculative ideas rely on single-laboratory demonstrations.

Modelling the Brain with Neural Networks

The most active theme replaces the classic serial computer with the artificial neural network and tests it directly against brain data, not by loose analogy.

Reviews argue that deep networks now work as genuine scientific models of biological information processing (Cichy & Kaiser, 2019; Kriegeskorte, 2015). Their internal representations can be compared quantitatively with neural activity.

Saxe et al. (2021) add a caution. A network that matches the brain’s behaviour does not necessarily match its mechanism, so researchers must ask what question each model actually answers.

Kietzmann et al. (2019) put the comparison to its most decisive test yet.

This is a quantitative model comparison against neural measurements, so it carries far more weight than analogy alone. Its scope is limited to early and mid-level vision, however, and whether the result generalises to higher cognition remains open.

The Predictive Brain

A second research front recasts cognition as prediction and error-correction, not passive input-output.

De Lange, Heilbron, and Kok (2018) reviewed the behavioural and neural evidence. They concluded that prior expectations systematically facilitate perception. This is a modern, Bayesian expression of the old idea that schemas shape what we see.

The evidence for the phenomenon itself is strong. Its neural mechanism is not. A critical review by Walsh, McGovern, Clark, and O’Connell (2020) found that predictive-processing models have historically been under-tested.

The evidence only partly supports the framework’s core claims. Priors shape perception, in short, but the mechanism that would replace the serial-computer analogy remains provisional.

Limits of Cognitive Training

The analogy’s more hopeful promise is that training one processing component, such as working memory, should transfer broadly to general ability. The strongest evidence says otherwise.

A meta-analysis by Melby-Lervåg, Redick, and Hulme (2016) found reliable short-term gains on trained and closely related tasks. It found no convincing far transfer to intelligence or school skills, once studies controlled for active comparison groups.

An extensive systematic review of the wider brain-training literature reached the same verdict (Simons et al., 2016). These reviews sit at the top of the evidence hierarchy. They show the mind’s processors are far less freely upgradeable than a strong hardware/software reading once implied.

References

Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation (Vol. 2, pp. 89–195). Academic Press. https://doi.org/10.1016/S0079-7421(08)60422-3

Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The psychology of learning and motivation (Vol. 8, pp. 47–89). Academic Press. https://doi.org/10.1016/S0079-7421(08)60452-1

Broadbent, D. E. (1958). Perception and communication. Pergamon Press. https://doi.org/10.1037/10037-000

Bruce, V., & Young, A. (1986). Understanding face recognition. British Journal of Psychology, 77(3), 305–327. https://doi.org/10.1111/j.2044-8295.1986.tb02199.x

Cichy, R. M., & Kaiser, D. (2019). Deep neural networks as scientific models. Trends in Cognitive Sciences, 23(4), 305–317. https://doi.org/10.1016/j.tics.2019.01.009

de Lange, F. P., Heilbron, M., & Kok, P. (2018). How do expectations shape perception? Trends in Cognitive Sciences, 22(9), 764–779. https://doi.org/10.1016/j.tics.2018.06.002

Deutsch, J. A., & Deutsch, D. (1963). Attention: Some theoretical considerations. Psychological Review, 70(1), 80–90. https://doi.org/10.1037/h0039515

Kietzmann, T. C., Spoerer, C. J., Sörensen, L. K. A., Cichy, R. M., Hauk, O., & Kriegeskorte, N. (2019). Recurrence is required to capture the representational dynamics of the human visual system. Proceedings of the National Academy of Sciences, 116(43), 21854–21863. https://doi.org/10.1073/pnas.1905544116

Kriegeskorte, N. (2015). Deep neural networks: A new framework for modeling biological vision and brain information processing. Annual Review of Vision Science, 1, 417–446. https://doi.org/10.1146/annurev-vision-082114-035447

Melby-Lervåg, M., Redick, T. S., & Hulme, C. (2016). Working memory training does not improve performance on measures of intelligence or other measures of “far transfer”: Evidence from a meta-analytic review. Perspectives on Psychological Science, 11(4), 512–534. https://doi.org/10.1177/1745691616635612

Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. https://doi.org/10.1037/h0043158

Neisser, U. (1967). Cognitive psychology. Appleton-Century-Crofts.

Neisser, U. (1976). Cognition and reality: Principles and implications of cognitive psychology. W. H. Freeman.

Newell, A., & Simon, H. A. (1972). Human problem solving. Prentice-Hall.

Newell, A., & Simon, H. A. (1976). Computer science as empirical inquiry: Symbols and search. Communications of the ACM, 19(3), 113–126. https://doi.org/10.1145/360018.360022

Rumelhart, D. E., & McClelland, J. L. (1986). Parallel distributed processing: Explorations in the microstructure of cognition. Volume 1: Foundations. MIT Press.

Saxe, A., Nelli, S., & Summerfield, C. (2021). If deep learning is the answer, what is the question? Nature Reviews Neuroscience, 22(1), 55–67. https://doi.org/10.1038/s41583-020-00395-8

Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423. https://doi.org/10.1002/j.1538-7305.1948.tb01338.x

Simons, D. J., Boot, W. R., Charness, N., Gathercole, S. E., Chabris, C. F., Hambrick, D. Z., & Stine-Morrow, E. A. L. (2016). Do “brain-training” programs work? Psychological Science in the Public Interest, 17(3), 103–186. https://doi.org/10.1177/1529100616661983

Sternberg, S. (1966). High-speed scanning in human memory. Science, 153(3736), 652–654. https://doi.org/10.1126/science.153.3736.652

Treisman, A. (1964). Selective attention in man. British Medical Bulletin, 20(1), 12–16. https://doi.org/10.1093/oxfordjournals.bmb.a070274

Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59(236), 433–460. https://doi.org/10.1093/mind/LIX.236.433

von Neumann, J. (1958). The computer and the brain. Yale University Press.

Walsh, K. S., McGovern, D. P., Clark, A., & O’Connell, R. G. (2020). Evaluating the neurophysiological evidence for predictive processing as a model of perception. Annals of the New York Academy of Sciences, 1464(1), 242–268. https://doi.org/10.1111/nyas.14321

Further information

• Why Your Brain is Not a Computer
• BBC Radio: Donald Broadbent and the Cocktail Party
• Attention Journal Article
• Attention Essay