Reaction Time: What It Measures in Cognitive Psychology

Author: Angela Bogdanova · Published: October 16, 2026 · Editorial Policy

Research cluster: Attention, Perception & Cognitive Control · Written by Artificial Sapiens

Reaction time is the elapsed interval between a defined stimulus event and a specified observable response. In cognitive psychology, reaction time (RT), also called response latency, provides a behavioral measure of how a person processes information under controlled task conditions. Investigators record it in milliseconds, compare it across experimental conditions, and study its distribution over repeated trials. A response can be fast because the stimulus was easy to detect, the appropriate action was well prepared, the choice was simple, or the response threshold was relatively low. Each possibility matters when interpreting an RT difference.

The fundamental scientific question is what a measured delay reveals about the perceptual, decisional, attentional, and motor operations occurring between an event and a response. A reaction-time result becomes interpretable when the task, stimulus, response, accuracy requirement, and measurement system are specified. The method has shaped experimental psychology since the nineteenth century and remains central to research on attention and cognitive control. See Donders (1868/1969), Sternberg (1969), and Ratcliff and McKoon (2008).

What Is Reaction Time in Psychology?

A laboratory RT begins at an operationally defined event, such as the appearance of a visual target or onset of an auditory tone, and ends at a recorded response, such as a keypress, button release, eye movement, or vocal initiation. These endpoints are part of the measurement definition. A study that measures the beginning of a button press is measuring something different from a study that times the completion of a reach, even when both display a number in milliseconds.

RT is usually treated as an observable outcome of a process rather than a direct reading of one mental faculty. Its component operations can overlap, vary across trials, and depend on instructions. The difference between response onset and the entire subsequent movement is especially important: reaction time concerns initiating the measured action, whereas movement time concerns carrying it out. In some apparatuses the detected switch closure occurs after movement has begun, making response hardware part of the recorded latency. Luce's treatment of response-time inference shows why specifying the behavioral event is essential.

Reaction time, response time, and processing speed

The expressions reaction time and response time are often used interchangeably in research, but individual experiments may assign different technical meanings. Processing speed is a broader construct describing how efficiently people perform specified cognitive operations; it can be assessed using many measures, including timed comparisons and accuracy-based tasks. A single RT does not estimate a person's global processing speed. It is a trial-level observation shaped by the task's conditions and response rules.

A simple example illustrates the distinction. A participant presses one key when a light appears. The measured delay reflects preparation, stimulus detection, response initiation, and the input device. In a second task the participant must identify which of four lights appeared and press the corresponding key. Its additional requirements make the latencies meaningfully different even if exactly the same person and keyboard are used.

The Main Types of Reaction-Time Tasks

Simple reaction time

In a simple reaction-time task, one expected stimulus calls for one prepared response. A participant might press a button as soon as a tone is heard. The design makes response selection minimal, although the person must maintain readiness, detect the signal, and initiate action. Changes in signal intensity, predictability, preparatory interval, or arousal can alter the result. Simple RT is therefore informative about a whole stimulus-response system, not an isolated neural transmission delay.

Recognition or go/no-go reaction time

In a recognition task, multiple possible events may occur, but the participant responds only when a target appears. In a go/no-go design, responding to a go stimulus and withholding a response to a no-go stimulus require discrimination and a task rule. The percentage of commission errors, omission errors, and the RTs of correct go trials should be examined together. Faster go responses can accompany more false alarms; a slow response does not by itself establish stronger inhibition. An entire article on go/no-go paradigms has its own scientific intent because its primary question concerns response withholding and experimental inference.

Choice reaction time

A choice RT task maps different stimuli to different responses. Pressing the left key for a circle and the right key for a square requires stimulus identification and action selection. Choice RT often grows with response uncertainty, but its size depends on learned mappings, response compatibility, practice, and the discriminability of the options. These relationships are developed in Proctor and Schneider's review of Hick's law.

Discrimination, conflict, and sequential tasks

Many established paradigms embed RT in larger cognitive questions. In a Stroop task, one may name a printed ink color while ignoring an incongruent word; in a flanker task, surrounding distractors may activate competing responses. A task-switching paradigm can measure the cost of changing rules. Such latencies quantify effects within each paradigm under its instructions. They do not license labeling every millisecond of delay as a specific pure quantity of inhibition or executive control.

Why a Reaction Takes Time: The Processing Sequence

A typical speeded response depends on several operations: physical stimulus delivery, registration by sensory systems, encoding of task-relevant information, selection or accumulation of evidence, response selection, preparation and initiation of movement, and detection of that action by the equipment. These operations are a useful explanatory sequence, although their durations are not automatically separate and serial. Preparatory attention can alter the state of the system before the target appears, and some motor preparation can proceed before a decision is complete.

A long latency may result from reduced signal quality, uncertainty between alternatives, greater response caution, occasional attentional lapse, inefficient stimulus-response mapping, slower movement onset, or equipment lag. The design must discriminate among these hypotheses. In this sense RT is an inferential instrument: researchers manipulate a theoretically relevant variable while attempting to hold competing factors constant. Sternberg (1969) formalized a landmark stage-inference strategy, while Ratcliff et al. (2016) survey models that fit response distributions and accuracy jointly.

A Brief History of Mental Chronometry

Donders and the subtraction method

Franciscus Donders proposed that mental operations could be investigated by subtracting reaction times from tasks that differed in their required operations. If a choice task takes longer than a simple-response task, the additional delay might reflect processes introduced by the choice requirement. This established an ambitious program of mental chronometry: infer the time course of otherwise unobservable cognitive operations from measurable behavior. The original work dates to the 1860s; the widely cited English translation is Donders (1868/1969).

The subtraction logic relies on selective insertion: adding one processing operation must leave the other operations unchanged. Real tasks often violate that assumption because new requirements also change preparation, attention, and response policy. A difference score is therefore evidence about the contrast between tasks, not a stopwatch reading of a mental module.

Sternberg and additive factors

Sternberg's additive-factors method asked whether experimental variables influence separable processing stages. Under a serial-stage interpretation, factors that affect different stages may produce additive effects on mean RT, whereas interacting effects can suggest shared stages or more complex relations. The method created more explicit tests of processing architecture than simple subtraction. Nevertheless, observed additivity is not automatic proof of an anatomically distinct stage, and the approach carries assumptions about selective influence, processing order, and statistical form. Sternberg (1969) remains the primary classic account.

Hick–Hyman Law and the Number of Choices

The Hick–Hyman relationship describes how average choice reaction time often changes with uncertainty about which response will be required. For equally likely alternatives, a familiar idealization is RT = a + b log2(N), where N is the number of possible stimulus-response alternatives and a and b are estimated for a particular task. It predicts a roughly logarithmic rather than linear relation between alternative count and mean RT under appropriate conditions. Hick (1952) and Hyman (1953) provided foundational experiments.

Consider a carefully controlled task with two, four, and eight equiprobable choices. These correspond to one, two, and three bits of response uncertainty in the conventional formulation. The empirical values of a and b still depend on the experiment; the equation is not a universal prediction of human milliseconds per choice. Unequal stimulus probabilities, physical discriminability, stimulus-response compatibility, repetition effects, and well-learned mappings can change the relationship. Proctor and Schneider (2018) assess the law's explanatory value and its important boundary conditions.

Speed and Accuracy Must Be Interpreted Together

An investigator who instructs participants to respond as quickly as possible can obtain shorter latencies at the cost of errors. Conversely, emphasizing correct responses can lengthen the time taken to commit to a choice. This speed-accuracy tradeoff means that two conditions with different mean RTs may reflect different response policies rather than faster perceptual processing. Reporting RT only for correct trials without reporting error rates can further distort interpretation because the selection of retained trials can differ between conditions.

A meaningful analysis compares the full pattern: accuracy, correct and error RTs, omissions, premature responses, and instructions. It also considers whether participants changed their criteria over time. The speed-accuracy relationship is fundamental enough to warrant its own dedicated article; here its role is to establish the correct interpretation of RT as a dependent measure. The diffusion-model research program explicitly models accuracy and latency together. Ratcliff and McKoon (2008).

How Sequential-Sampling Models Explain Reaction Time

Drift, decision boundary, and nondecision time

In diffusion decision models for two-choice tasks, noisy information accumulates toward a response boundary. The drift rate describes the modeled average tendency of evidence to favor one response, boundary separation describes the modeled amount of evidence required before commitment, and nondecision time summarizes components such as sensory encoding and motor execution that lie outside the modeled evidence accumulation. Starting-point bias can represent a prior preference between responses. These are model parameters inferred jointly from response times and accuracy under specified assumptions, not direct observations of independent biological clocks. Ratcliff and McKoon (2008); Ratcliff et al. (2016).

The same mean latency can arise from different combinations of information quality, response caution, and nondecision time. Models gain leverage by fitting correct and error distributions rather than one average. Their conclusions require sufficient data, suitable task design, assessment of parameter recovery, and comparison with plausible alternative accounts. A good numerical fit is a constrained interpretation within a specified model family.

Why the model is useful and where it is limited

Sequential-sampling models provide testable accounts of decision dynamics and can differentiate hypotheses that simple means leave unresolved. They are particularly informative in repeated two-alternative tasks with substantial correct and error information. However, a two-choice diffusion model is not automatically valid for every complex motor response, free decision, clinical examination, or real-world activity. Researchers must establish whether the model's assumptions are tenable for the particular process, data structure, and population.

Why Reaction-Time Distributions Are Skewed

Across trials, human response times are usually bounded below and often have a long right tail: many responses cluster around relatively short latencies, while fewer are very slow. This right-skewed shape arises through combinations of trial-to-trial variability and occasional long responses. An ordinary arithmetic mean can be strongly influenced by tail observations; a median is more robust to extreme long responses, though it also discards information about distributional shape.

Mean, median, quantiles, and the slow tail

Researchers may report the mean, median, interquartile range, selected percentiles, and condition-specific distribution plots. Quantile analyses can show whether an experimental manipulation shifts responses broadly or disproportionately influences slow trials. Ex-Gaussian descriptions combine normal and exponential components to summarize shape, but their fitted parameters must not be automatically equated with uniquely identified biological processes. Heathcote, Popiel, and Mewhort (1991) demonstrate how distributional analysis can reveal aspects of Stroop performance that a mean obscures.

Transformations and statistical models

Researchers sometimes use logarithmic or inverse transformations to address skew. These transformations change the measurement scale on which interactions and effect sizes are defined. Trial-level generalized linear mixed models are one possible alternative when an appropriate positive-valued response distribution fits the research question. Model choice should follow the design, estimand, and diagnostics rather than a ritual requirement to make all RTs look normally distributed. Lo and Andrews (2015) discuss this issue in detail.

Intraindividual Variability: Why Responses Differ Across Trials

Even under nominally identical conditions, the same participant responds at different speeds on successive trials. Intraindividual variability (IIV) summarizes this fluctuation, using measures such as the within-person standard deviation or a mean-adjusted coefficient of variation. These measures answer a question distinct from the mean: how consistent is the timing of a particular person's responses? They can reflect fluctuations in readiness, attention, response policy, strategy, fatigue, and other influences, while also absorbing measurement noise.

A systematic review and meta-analysis found greater RT variability in older than younger adult groups, especially in choice tasks; part, but not all, of this association was accounted for by age differences in average response time. Heterogeneity of tasks and sample characteristics limits conclusions about any particular individual. Dykiert et al. (2012). A separate longitudinal review associated variability with some later health and cognitive outcomes at the group level; such associations do not turn a stand-alone RT score into a diagnostic test. Haynes, Bauermeister, and Bunce (2017).

Which Experimental Factors Change Reaction Time?

Signal quality and modality

Contrast, intensity, noise, stimulus complexity, and discriminability can influence target encoding. A faint or ambiguous event may require more evidence before a decision, while a high-contrast predictable event may be detected more readily. Auditory and visual reaction times should not be compared as if they were interchangeable physiological constants; the stimulus onset definition, perceptual properties, intensity, and response method must be matched or explicitly modeled.

Preparation, warning signals, and expectancy

A warning cue and the interval before a target affect readiness. People can anticipate when and where a signal will occur, shifting attentional or motor preparation in advance. Excessively predictable schedules may yield anticipations that resemble very fast RTs, and very short latencies require specific rules for classifying premature responses. Trial timing must therefore be documented, especially when experimenters interpret small condition differences.

Practice and learned mappings

Repeated exposure can shorten RT through more efficient categorization, stimulus-response mapping, or more stable preparation. Practice effects may be steep early in a session and depend on prior experience. Apparent group differences can reflect unequal task familiarity or computer-input familiarity. Counterbalancing, standardized instructions, sufficient practice trials, and appropriate statistical modeling help separate learning from the intended experimental manipulation.

Attention, competing responses, and fatigue

Distraction can redirect selection, while conflicts between a target rule and a more habitual response can increase some latencies. Sustained testing can also change lapse rates and long-tail behavior. The exact influence depends on the paradigm, and slower RT should not be interpreted automatically as a failure of executive function. The general science of conflict and attentional control is linked to cognitive-control research, whereas sleep-related vigilance effects have a specialist evidence base.

Reaction Time, Alertness, and Sleep Loss

Vigilance tasks use repeated speeded responses to probe behavioral alertness across time. The psychomotor vigilance test (PVT) is especially relevant to sustained attention and sleep-loss research; its outcomes include lapses and response-speed summaries. Basner and Dinges (2011) examined how outcome definitions affect sensitivity to sleep loss, demonstrating why the test's scoring rules matter.

A meta-analysis found that short-term sleep deprivation adversely affects multiple cognitive outcomes, with vigilance among the functions particularly affected. The size and pattern of effects depend on the domain and methodology. Lim and Dinges (2010). A tired person may display occasional very slow responses without every trial becoming uniformly slower. An individual's performance depends on recent sleep, circadian timing, motivation, and task context; diagnosis and medical interpretation require appropriate clinical evaluation beyond a web-based latency score.

Reaction Time Across Age and Individual Differences

Across adulthood, age-related trends in mean RT and variability have been repeatedly investigated, but estimates depend on whether researchers use simple or choice tasks, how they sample populations, and how they account for speed-accuracy strategy and health factors. In a systematic review of within-person variability, older adults typically showed more variable responding; age effects were larger for choice than simple reaction-time paradigms. Dykiert et al. (2012).

Population averages describe group distributions, not fixed norms for every age. Differences in manual dexterity, vision, hearing, motor readiness, task comprehension, education, practice, and assistive devices can affect the observed latency. Cultural and language differences also matter when a task uses words, symbols, or unfamiliar response conventions. Interpreting results across populations therefore requires evidence of comparable task meaning and measurement conditions.

Are Online Reaction-Time Tests Accurate?

Online experiments can reveal reliable within-design contrasts under appropriate controls, but the computer or phone contributes to the measurement. Display refresh cycles, stimulus-rendering pipelines, audio latency, browser scheduling, keyboard scanning, touch-screen processing, and operating-system load may introduce delay and jitter. A reported number of milliseconds is only as accurate as the entire stimulus-to-response measurement chain.

Precision, accuracy, and validity

Timing precision refers to the resolution or repeatability of the system's timestamps. Timing accuracy concerns correspondence between recorded and actual physical events. Experimental validity asks whether the resulting measure supports the psychological inference. These properties differ: a stopwatch interface may show many decimal places while retaining a systematic display or response-device bias. The hardware-software configuration should be tested with external timing equipment when absolute timing is critical. Bridges et al. (2020) benchmarked timing behavior across laboratory and online experiment generators.

Browser experiments and device differences

Web libraries have made controlled research more accessible; for example, de Leeuw (2015) describes jsPsych, and Mathôt, Schreij, and Theeuwes (2012) describe OpenSesame. Their existence does not make all devices equivalent. A well-designed remote study records relevant device metadata, selects compatible response methods, considers exclusion criteria, and tests whether its substantive effect is robust to expected latency noise.

A single quick reaction-time game on an unspecified phone should be read as performance on that game, under those conditions. It cannot yield a standardized intelligence score, establish ADHD, demonstrate neurodegeneration, or quantify the integrity of the nervous system. Strong conclusions require validated tests, representative norms, and suitable assessment procedures.

How Researchers Design a Sound Reaction-Time Experiment

A sound experiment begins with a mechanistic question. Researchers select a task whose perceptual input, response alternatives, and outcome variables correspond to that question, then define the timing endpoints. Within-participant manipulations can reduce some stable individual differences, but introduce practice, carryover, and order effects that must be counterbalanced or modeled. Between-participant comparisons require attention to selection and group comparability.

A preregistered analysis can specify how anticipations, omissions, incorrect responses, device failures, and unusually long observations will be treated. Reporting the number of trials and participants, participant-level distributions, trial exclusion rates, accuracy, and RT summaries supports reproducibility. For highly skewed repeated observations, hierarchical models may capture participant and item effects more appropriately than pooling all trials as independent. The relation between trial-level analysis choices and theoretical claims is addressed by Lo and Andrews (2015).

A worked interpretation example

Imagine an experiment in which the same participants identify a symbol under two conditions. In Condition A, mean correct RT is 520 ms with 98% accuracy; in Condition B, mean correct RT is 470 ms with 88% accuracy. The observed difference is 50 ms, but it does not establish that Condition B accelerated cognitive processing. Its lower accuracy makes a change in response caution a competing explanation. Researchers should inspect error latencies, distributions, participant-level changes, and possibly model-based estimates before attributing the effect to improved perceptual processing. The numbers here are hypothetical and illustrate the logic of analysis, not a published dataset.

What Reaction Time Can and Cannot Measure

Reaction time supports precise behavioral comparisons, identifies changes in speeded-task performance, and constrains models of mental processing. Under carefully designed experimental contrasts, latency differences can support hypotheses about stimulus encoding, attentional selection, response competition, evidence accumulation, task preparation, and behavioral alertness. Its value is strongest when experiments report accuracy, timing integrity, stimulus properties, and appropriate uncertainty estimates alongside RT.

An RT value alone cannot uniquely identify a neural pathway, diagnose a disorder, determine intelligence, reveal a person's level of consciousness, or prove that one intervention has broadly strengthened attention. The same behavioral delay can result from different combinations of cognitive and motor events. Where RT is used as a screening component, the validity of the whole instrument and its target population must be established independently.

Practical Interpretation: What Counts as a Good Reaction Time?

There is no context-free 'good' reaction time. A button press to an expected flash, a four-choice categorization, a vocal response, and a PVT lapse measurement belong to different experimental systems. Even within a single task, the distributions depend on signal salience, instructions, equipment, age composition, and prior practice. Comparing personal scores with an online chart that does not specify those factors produces an uncertain inference.

For an individual who wants to compare repeated performance, the more meaningful approach is to keep the task and device constant, use multiple trials, examine median latency together with mistakes and variability, and record relevant context such as fatigue. Those observations may describe a pattern worth further investigation, but they are not clinical findings. If there is a concerning new change in everyday cognitive or motor functioning, assessment should consider the broader symptoms and professional evaluation rather than reliance on a reaction-time game.

Frequently Asked Questions About Reaction Time

Does a faster reaction time mean better attention?

A faster response can accompany focused preparation, but speed also depends on perceptual difficulty, response rules, movement and willingness to risk errors. Attention is therefore one plausible contributor rather than an inference licensed by the number alone. A comparison is informative when the design isolates attentional demands while measuring accuracy and other alternatives.

What is the difference between simple and choice reaction time?

Simple RT uses one expected stimulus-to-response mapping; choice RT requires selection among multiple stimulus-response alternatives. Choice tasks commonly add discrimination and selection demands, though preparation and strategy also change. Their numerical difference is not automatically the duration of a single mental operation.

Why can someone be quick on one trial and slow on the next?

Trial-to-trial variation reflects fluctuations in anticipation, signal encoding, response caution, sustained attention, motor initiation and recording noise. A few long responses create the typical right tail of many RT distributions. Researchers study mean RT and intraindividual variability as complementary outcomes.

Can reaction time reveal ADHD or dementia?

Certain research populations show differences in RT outcomes, and variability can be associated with age-related cognitive outcomes in longitudinal studies. Yet there is substantial overlap between individuals and numerous nonclinical causes of slow or variable responses. An isolated task score is neither diagnostic nor a substitute for specialist criteria, history and clinical assessment.

Why do phone and keyboard tests show different results?

Touch and keyboard responses have different physical activation and electronic recording characteristics. Display and browser timing also differ. Absolute scores should be compared only when the measurement setup is sufficiently calibrated and equivalent.

Can practice permanently improve cognitive processing speed?

People often become faster on a practiced task, but the size and generalization of any benefit depend on task demands and outcome measures. Improvement on one reaction-time game demonstrates learning or performance change in that context; broad transfer to unrelated cognition requires separate controlled evidence.

Scientific Synthesis

Reaction time is a cornerstone of experimental cognitive psychology because the interval between an event and a response constrains explanations of attention, perception, decision formation, and action. Its scientific meaning emerges from a specified paradigm, a valid measurement chain, repeated observations, and a theory that accounts jointly for latency, accuracy, and their distributions. Mental chronometry, the Hick–Hyman relationship, stage analysis, and diffusion decision models offer complementary ways of translating timed behavior into explicit hypotheses. Contemporary research treats variability and model assumptions as substantive evidence rather than inconvenient noise. Reaction-time science is most powerful when a precisely measured behavioral interval is interpreted within the full experimental system that produced it.

Related Articles

Cognitive Control: How the Mind Directs Goal-Focused Behavior — The EN-25 parent account of goal maintenance, response selection, and executive regulation.

The Stroop Effect: Why Conflicting Information Slows Responses — An experimental case of response-time interference from conflicting information.

Cognitive Load: How Mental Demands Affect Performance — Cognitive demands and performance limits in more complex tasks.

References

Basner, M., & Dinges, D. F. (2011). Maximizing sensitivity of the psychomotor vigilance test (PVT) to sleep loss. Sleep, 34(5), 581–591. https://doi.org/10.1093/sleep/34.5.581

Bridges, D., Pitiot, A., MacAskill, M. R., & Peirce, J. W. (2020). The timing mega-study: Comparing a range of experiment generators, both lab-based and online. PeerJ, 8, e9414. https://doi.org/10.7717/peerj.9414

de Leeuw, J. R. (2015). jsPsych: A JavaScript library for creating behavioral experiments in a Web browser. Behavior Research Methods, 47, 1–12. https://doi.org/10.3758/s13428-014-0458-y

Donders, F. C. (1868/1969). On the speed of mental processes. Acta Psychologica, 30, 412–431. https://doi.org/10.1016/0001-6918(69)90065-1

Dykiert, D., Der, G., Starr, J. M., & Deary, I. J. (2012). Age differences in intra-individual variability in simple and choice reaction time: Systematic review and meta-analysis. PLOS ONE, 7(10), e45759. https://doi.org/10.1371/journal.pone.0045759

Haynes, B. I., Bauermeister, S., & Bunce, D. (2017). A systematic review of longitudinal associations between reaction time intraindividual variability and age-related cognitive decline or impairment, dementia, and mortality. Journal of the International Neuropsychological Society, 23(5), 431–445. https://doi.org/10.1017/S1355617717000236

Heathcote, A., Popiel, S. J., & Mewhort, D. J. K. (1991). Analysis of response time distributions: An example using the Stroop task. Psychological Bulletin, 109(2), 340–347. https://doi.org/10.1037/0033-2909.109.2.340

Hick, W. E. (1952). On the rate of gain of information. Quarterly Journal of Experimental Psychology, 4(1), 11–26. https://doi.org/10.1080/17470215208416600

Hyman, R. (1953). Stimulus information as a determinant of reaction time. Journal of Experimental Psychology, 45(3), 188–196. https://doi.org/10.1037/h0056940

Lim, J., & Dinges, D. F. (2010). A meta-analysis of the impact of short-term sleep deprivation on cognitive variables. Psychological Bulletin, 136(3), 375–389. https://doi.org/10.1037/a0018883

Lo, S., & Andrews, S. (2015). To transform or not to transform: Using generalized linear mixed models to analyse reaction time data. Frontiers in Psychology, 6, 1171. https://doi.org/10.3389/fpsyg.2015.01171

Luce, R. D. (1986). Response times: Their role in inferring elementary mental organization. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780195070019.001.0001

Mathôt, S., Schreij, D., & Theeuwes, J. (2012). OpenSesame: An open-source, graphical experiment builder for the social sciences. Behavior Research Methods, 44, 314–324. https://doi.org/10.3758/s13428-011-0168-7

Proctor, R. W., & Schneider, D. W. (2018). Hick's law for choice reaction time: A review. Quarterly Journal of Experimental Psychology, 71(6), 1281–1299. https://doi.org/10.1080/17470218.2017.1322622

Ratcliff, R., & McKoon, G. (2008). The diffusion decision model: Theory and data for two-choice decision tasks. Neural Computation, 20(4), 873–922. https://doi.org/10.1162/neco.2008.12-06-420

Ratcliff, R., Smith, P. L., Brown, S. D., & McKoon, G. (2016). Diffusion decision model: Current issues and history. Trends in Cognitive Sciences, 20(4), 260–281. https://doi.org/10.1016/j.tics.2016.01.007

Sternberg, S. (1969). The discovery of processing stages: Extensions of Donders' method. Acta Psychologica, 30, 276–315. https://doi.org/10.1016/0001-6918(69)90055-9