Skip to main content
v2026.11,986 entries · CC-BY 4.0

Flanker Task: Paradigm, Formulas, and Statistical Analysis

A comprehensive guide to the Eriksen flanker task: theoretical foundations, mathematical formulas, outlier trimming, worked examples, and APA 7th reporting.

Written and maintained by CASRAI Editorial Board

Published

On this page

The flanker task—originally designed by Barbara A. Eriksen and Charles W. Eriksen in 1974—is a classical cognitive psychology experiment used to measure selective visual attention, response inhibition, and executive control. Participants identify a central target stimulus while attempting to ignore flanking distractors that either prompt the same behavioral response on congruent trials or trigger a competing motor response on incongruent trials. The resulting difference in reaction time and error rate quantifies the cognitive system’s ability to suppress distracting visual noise and resolve conflict in real time.

Definition and core paradigm

In cognitive psychology, the flanker task (commonly referred to as the Eriksen flanker paradigm) evaluates an individual’s capacity to direct selective visual attention toward a relevant focal target while filtering out distracting visual noise in the immediate spatial vicinity. By systematically varying the compatibility between the target and surrounding distractors across discrete trials, researchers quantify the cognitive vulnerability to interference and track the latency required for executive control mechanisms to override automatic motor impulses.

In the original 1974 experiment conducted by Barbara A. Eriksen and Charles W. Eriksen, participants viewed horizontal letter arrays where central target letters (such as H or K) were mapped to one response lever, and alternative target letters (such as S or C) were mapped to a second lever. When target letters were flanked by identical or response-compatible letters (e.g., HHHKHHH), reaction times were rapid and error rates remained low. Conversely, when flanked by response-incompatible letters (e.g., SSSHSSS), participants exhibited pronounced response latency inflation and elevated error rates. Modern laboratory implementations frequently utilize directional arrow arrays (such as →→→→→ for congruent trials, ←←→←← for incongruent trials, and —→— or square boxes for neutral trials), which eliminate letter-reading confounds across developmental and cross-linguistic cohorts.

When implementing computerized experiments in modern software environments such as Inquisit behavioral testing software or open-source web platforms like the jsPsych experiment framework, trial parameters must control spatial proximity. Eriksen and Eriksen established that flanker interference operates primarily within 1 degree of visual angle from the central fixation point. Distractors placed beyond this attentional spotlight fail to elicit robust response competition, while distractors placed too closely can introduce low-level sensory masking or lateral visual crowding rather than cognitive response conflict. Much like the temporal bottleneck examined in the attentional blink paradigm, the flanker task isolates structural limitations in human information processing—specifically demonstrating that early perceptual stages process peripheral visual inputs automatically before focal attentional filters can restrict downstream motor activation.

Formula and quantitative metrics

The primary quantitative metric derived from the flanker task is the flanker interference effect (often designated as the flanker conflict cost, ΔRT). The basic formula evaluates the difference in mean reaction times between correct incongruent trials and correct congruent trials:

ΔRT = RT_incongruent – RT_congruent

In this formulation, RT_incongruent represents the arithmetic mean reaction time (measured in milliseconds) calculated exclusively across accurate responses on incongruent trials, and RT_congruent denotes the arithmetic mean reaction time on accurate congruent trials. Because raw response latencies vary substantially across developmental stages, clinical diagnoses, and baseline individual differences in motor speed, researchers frequently compute the proportional interference cost (PIC):

PIC = ((RT_incongruent – RT_congruent) / RT_congruent) × 100

Here, PIC expresses the interference cost as a percentage of the participant’s baseline congruent response latency. This normalization prevents slower individuals (such as older adults or clinical groups) from exhibiting artificially inflated absolute millisecond differences driven purely by generalized cognitive or motor slowing.

To account for accuracy variations and potential speed-accuracy trade-offs, two additional quantitative metrics are standardly calculated. First, the error interference rate (ΔER) captures the accuracy deficit introduced by distracting flankers:

ΔER = ER_incongruent – ER_congruent

where ER_incongruent and ER_congruent represent the proportion or percentage of incorrect responses in the incongruent and congruent conditions, respectively. Second, when participants trade response accuracy for speed, researchers employ the Inverse Efficiency Score (IES):

IES = RT / (1 – PE)

In this equation, RT is the mean correct response latency and PE is the proportion of errors (PE = ER / 100). IES is expressed in effective milliseconds; higher values signify poorer performance. Furthermore, in sequential trial designs, the Gratton effect (trial-by-trial conflict adaptation) is computed by examining the interaction between the previous trial’s compatibility (trial N-1) and the current trial’s compatibility (trial N):

Conflict Adaptation = (RT_cI – RT_cC) – (RT_iI – RT_iC)

where lowercase letters represent the compatibility of trial N-1 (c = congruent, i = incongruent) and uppercase letters represent trial N (C = congruent, I = incongruent). A positive conflict adaptation score confirms that flanker interference is attenuated immediately following the detection of conflict on the prior trial.

Worked example with real numbers and arithmetic

To demonstrate the calculation of flanker interference and inferential statistical testing, consider an experimental dataset comprising N = 10 adult participants who completed a standard 100-trial arrow flanker paradigm (50 congruent and 50 incongruent trials presented in pseudorandomized order). Raw response times were cleaned prior to aggregation by excluding error trials, anticipatory responses (RT under 200 ms), and outlier latencies exceeding 2.5 standard deviations from each participant’s cell mean. The resulting participant-level mean reaction times (in milliseconds) and corresponding difference scores are presented below:

Participant Congruent RT (ms) Incongruent RT (ms) Interference Cost ΔRT (ms)
P01 412.4 468.1 55.7
P02 435.8 501.2 65.4
P03 389.5 432.0 42.5
P04 451.2 518.7 67.5
P05 420.6 474.3 53.7
P06 405.1 452.9 47.8
P07 442.3 509.4 67.1
P08 398.7 441.5 42.8
P09 428.0 485.6 57.6
P10 416.4 476.3 59.9

We now execute the step-by-step arithmetic to determine the group sample means, standard deviations, standard error of the mean difference, paired t-statistic, and effect size:

Step 1: Compute the condition sample means.
For the congruent condition:
Mean_congruent = (412.4 + 435.8 + 389.5 + 451.2 + 420.6 + 405.1 + 442.3 + 398.7 + 428.0 + 416.4) / 10 = 4200.0 / 10 = 420.00 ms.
For the incongruent condition:
Mean_incongruent = (468.1 + 501.2 + 432.0 + 518.7 + 474.3 + 452.9 + 509.4 + 441.5 + 485.6 + 476.3) / 10 = 4760.0 / 10 = 476.00 ms.

Step 2: Calculate the mean flanker interference cost.
Mean_ΔRT = Mean_incongruent – Mean_congruent = 476.00 ms – 420.00 ms = 56.00 ms.
Equivalently, averaging the individual difference scores yields: (55.7 + 65.4 + 42.5 + 67.5 + 53.7 + 47.8 + 67.1 + 42.8 + 57.6 + 59.9) / 10 = 560.0 / 10 = 56.00 ms.
Proportional interference cost PIC = (56.00 / 420.00) × 100 = 13.33%.

Step 3: Calculate the variance and standard deviation of the difference scores.
The deviations of each difference score from the mean difference (56.00 ms) are:
d1 = 55.7 – 56.0 = -0.3 (d1² = 0.09)
d2 = 65.4 – 56.0 = +9.4 (d2² = 88.36)
d3 = 42.5 – 56.0 = -13.5 (d3² = 182.25)
d4 = 67.5 – 56.0 = +11.5 (d4² = 132.25)
d5 = 53.7 – 56.0 = -2.3 (d5² = 5.29)
d6 = 47.8 – 56.0 = -8.2 (d6² = 67.24)
d7 = 67.1 – 56.0 = +11.1 (d7² = 123.21)
d8 = 42.8 – 56.0 = -13.2 (d8² = 174.24)
d9 = 57.6 – 56.0 = +1.6 (d9² = 2.56)
d10 = 59.9 – 56.0 = +3.9 (d10² = 15.21)
Sum of squared deviations SS_diff = 0.09 + 88.36 + 182.25 + 132.25 + 5.29 + 67.24 + 123.21 + 174.24 + 2.56 + 15.21 = 790.70.
Sample variance s²_diff = SS_diff / (N – 1) = 790.70 / 9 = 87.8556 ms².
Sample standard deviation s_diff = √(87.8556) = 9.3731 ms (rounded to 9.37 ms).
For the condition standard deviations, s_congruent = 19.61 ms and s_incongruent = 28.66 ms.

Step 4: Compute the standard error and inferential t-statistic.
The standard error of the mean difference SE_diff is computed as:
SE_diff = s_diff / √N = 9.3731 / √10 = 9.3731 / 3.16228 = 2.9640 ms (rounded to 2.96 ms).
Conducting a within-subjects paired-samples t-test on the difference scores yields:
t = Mean_ΔRT / SE_diff = 56.00 / 2.9640 = 18.893 (degrees of freedom df = N – 1 = 9).
With t(9) = 18.89, the two-tailed probability value is p < .00001 (reported in APA style as p < .001).

Step 5: Determine the standardized effect size.
When interpreting Cohen’s d effect size for paired within-subjects designs, Cohen’s d_z reflects the mean difference divided by the standard deviation of the difference scores:
d_z = Mean_ΔRT / s_diff = 56.00 / 9.3731 = 5.97.
This robust effect size indicates that response inhibition in the incongruent condition produces a substantial, highly reliable increase in cognitive processing latency across all 10 participants.

Interpretation and benchmark reading guidance

Interpreting flanker task performance requires evaluating both the magnitude of response latency prolongation and error distribution across conditions. In healthy young adults, standard arrow flanker tasks typically yield an interference cost ranging between 30 ms and 60 ms under standard visual display angles. When cognitive control mechanisms function efficiently, the anterior cingulate cortex (ACC) detects conflict between the target stimulus and peripheral distractors within 200 ms to 300 ms post-stimulus (indexed by the electrophysiological N200 event-related potential), signaling the dorsolateral prefrontal cortex (dlPFC) to apply top-down attentional biasing toward the central visual channel.

When a participant exhibits an abnormally elevated flanker interference cost (e.g., exceeding 90 ms in adult cohorts) or elevated incongruent error rates, this pattern reflects impaired selective visual filtering, vulnerability to distraction, or weakened prefrontal inhibitory control. Conversely, an unusually small interference cost (e.g., under 15 ms) accompanied by high incongruent error rates often points to an aggressive speed-accuracy trade-off, where the participant responds hastily before conflict resolution has concluded. The following reading guidance table outlines interpretive benchmarks across research and testing contexts:

Interference Cost (ΔRT) Typical Error Rate Cognitive Profile Primary Interpretation
15 to 35 ms Under 3% Superior Inhibitory Control Exceptional spatial filtering, rapid conflict suppression, minimal distractor intrusion.
35 to 65 ms 2% to 6% Normal Adult Benchmark Standard executive control efficiency; typical response competition and resolution latency.
65 to 100 ms 5% to 12% Moderate Attentional Sluggishness Delayed conflict detection, broad visual attentional spotlight, common in fatigue or minor distraction.
Over 100 ms Over 12% Substantial Inhibitory Deficit Pronounced failure of distractor suppression; frequently observed in ADHD, neurodegeneration, or early development.
Under 15 ms (with high errors) Over 20% Impulsive / Guessing Strategy Premature motor execution prior to perceptual conflict resolution; speed-accuracy trade-off violation.

The data table below compares benchmark parameters, typical reaction times, and interference magnitudes compiled from landmark experimental investigations across diverse participant groups and task configurations.

Methodological assumptions and common mistakes

Accurate quantification of flanker interference depends upon strict experimental control and rigorous data preparation. Researchers frequently encounter analytical artifacts when fundamental methodological assumptions are overlooked. Below are key assumptions and common pitfalls to avoid:

  • Failure to exclude error trials from reaction time calculations: Reaction time on incorrect trials reflects premature motor lapses or attentional omissions rather than controlled conflict resolution. Mixing error latencies into cell means artificially deflates mean incongruent reaction times because erroneous responses are typically faster than correct responses.
  • Ignoring the speed-accuracy trade-off (SAT): If one group slows down substantially to achieve 98% accuracy while another responds rapidly at 85% accuracy, comparing raw ΔRT alone introduces systematic bias. Researchers should report both reaction time and error rates, or utilize composite scores such as the Inverse Efficiency Score (IES) or the NIH Toolbox two-vector speed-accuracy additive metric.
  • Improper outlier filtering thresholds: Failure to trim anticipatory responses (latencies under 150 to 200 ms) and attentional lapses (latencies exceeding 2.5 or 3.0 standard deviations from the participant’s condition mean) introduces severe skewness into reaction time distributions, violating normality assumptions in parametric tests.
  • Incorrect visual display angle: Placing flanking distractors at a distance greater than 1 degree to 1.5 degrees of visual angle allows participants to narrow their focal spotlight and ignore distractors pre-attentively, eliminating the flanker effect. Conversely, placing letters or arrows too close causes physical contour interaction and lateral crowding rather than cognitive conflict.
  • Proportion congruent confounds: Presenting an uneven proportion of congruent and incongruent trials (e.g., 80% congruent and 20% incongruent) induces expectancy biases. In high-congruency blocks, participants expand their attentional spotlight, producing inflated interference costs; in low-congruency blocks, participants maintain sustained narrow focus, artificially diminishing the effect.
  • Confounding conflict adaptation with feature integration: When analyzing sequential trial effects (the Gratton effect), trial transitions involving exact stimulus-response repetitions (e.g., left target preceded by left target) introduce episodic memory retrieval benefits. Researchers must control for or partial out stimulus-response binding repetitions to isolate genuine cognitive control adaptation.
  • Insufficient trial counts: Reaction time difference scores exhibit lower test-retest reliability than raw condition means. To achieve acceptable psychometric reliability (Cronbach’s α ≥ .75), an experiment should present a minimum of 40 to 60 valid, artifact-free trials per experimental cell.

How to report flanker task results in APA style

When reporting flanker task findings in empirical manuscripts adhering to the American Psychological Association (APA 7th edition) standards, authors must detail trial trimming criteria, condition descriptive statistics (means and standard deviations), exact inferential test statistics, degrees of freedom, exact p-values, and standardized effect sizes with confidence intervals. When factorial designs involve multiple conditions or groups, results are typically analyzed through a repeated-measures ANOVA.

Below is a standardized APA 7th style reporting template based on our worked numerical example:

“Reaction time analyses were restricted to correct trials. Outliers defined as response latencies below 200 ms or exceeding 2.5 standard deviations from each participant’s cell mean were removed prior to aggregation (accounting for 2.1% of all trials). A paired-samples t-test indicated a significant flanker interference effect, t(9) = 18.89, p < .001, dz = 5.97, 95% CI [3.51, 8.41]. Participants responded significantly faster to congruent target arrays (M = 420.00 ms, SD = 19.61 ms) than to incongruent target arrays (M = 476.00 ms, SD = 28.66 ms), reflecting a mean interference cost of 56.00 ms (SD = 9.37 ms, SE = 2.96 ms), or a 13.33% latency penalty. Analysis of error rates revealed a parallel interference pattern, with significantly higher error rates on incongruent trials (M = 5.80%, SD = 2.10%) than on congruent trials (M = 1.20%, SD = 0.90%), t(9) = 6.45, p < .001, confirming that response slowing was not the result of a speed-accuracy trade-off.”

In addition to text summaries, empirical papers should present a figure showing both individual participant data points and condition violin/box plots, accompanied by a table summarizing cell means, error percentages, and trimmed trial percentages across experimental conditions.

How to calculate the flanker effect in Excel, R, and Python

Researchers process raw trial-level experimental logs into aggregate condition means and inferential statistics using various computational tools. Below are validated, production-ready code implementations in Excel, R, and Python.

Calculating the flanker effect in Microsoft Excel

Assuming trial-level data are recorded with Participant ID in Column A, Condition (‘congruent’ or ‘incongruent’) in Column B, Reaction Time (ms) in Column C, and Accuracy (1 for correct, 0 for error) in Column D:

  1. Compute participant congruent mean RT (Cell F2):
    =AVERAGEIFS(C$2:C$1000, A$2:A$1000, A2, B$2:B$1000, "congruent", D$2:D$1000, 1, C$2:C$1000, ">=200")
  2. Compute participant incongruent mean RT (Cell G2):
    =AVERAGEIFS(C$2:C$1000, A$2:A$1000, A2, B$2:B$1000, "incongruent", D$2:D$1000, 1, C$2:C$1000, ">=200")
  3. Calculate individual flanker interference cost (Cell H2):
    =G2 - F2
  4. Calculate proportional interference cost (Cell I2):
    =(H2 / F2) * 100
  5. Run paired-samples t-test across participants:
    =T.TEST(G2:G11, F2:F11, 2, 1)

Calculating the flanker effect in R

The following R script reads trial-level data, filters errors and RT outliers, calculates condition means and interference costs, and executes a paired t-test with Cohen’s d_z:

# Load required packages
library(dplyr)
library(tidyr)

# Filter raw trial data: retain correct responses between 200 ms and 1500 ms
clean_data = raw_trials |>
  filter(correct == 1, rt >= 200, rt 
  group_by(subject_id, condition) |>
  summarize(mean_rt = mean(rt), .groups = 'drop') |>
  pivot_wider(names_from = condition, values_from = mean_rt)

# Calculate flanker interference cost per participant
clean_data = clean_data |>
  mutate(
    flanker_cost = incongruent - congruent,
    prop_cost = (flanker_cost / congruent) * 100
  )

# Inferential paired t-test
t_result = t.test(clean_data$incongruent, clean_data$congruent, paired = TRUE)
print(t_result)

# Calculate Cohens dz effect size
mean_diff = mean(clean_data$flanker_cost)
sd_diff = sd(clean_data$flanker_cost)
cohen_dz = mean_diff / sd_diff
cat(sprintf('Mean interference: %.2f ms (SD = %.2f), Cohens dz = %.2fn', mean_diff, sd_diff, cohen_dz))

Calculating the flanker effect in Python

The following Python script utilizes pandas and scipy.stats to clean trial records, calculate subject-level metrics, and perform within-subjects hypothesis testing:

import pandas as pd
from scipy import stats

# Filter correct trials within legitimate latency boundaries
valid_trials = df[(df['correct'] == 1) & (df['rt'] >= 200) & (df['rt'] <= 1500)]

# Aggregate condition means by subject
subj_summary = valid_trials.pivot_table(
    index='subject_id',
    columns='condition',
    values='rt',
    aggfunc='mean'
).reset_index()

# Compute individual flanker interference effect and proportional cost
subj_summary['flanker_cost'] = subj_summary['incongruent'] - subj_summary['congruent']
subj_summary['prop_cost'] = (subj_summary['flanker_cost'] / subj_summary['congruent']) * 100

# Conduct paired-samples t-test
t_stat, p_val = stats.ttest_rel(subj_summary['incongruent'], subj_summary['congruent'])

# Compute descriptive statistics and Cohens dz
mean_diff = subj_summary['flanker_cost'].mean()
sd_diff = subj_summary['flanker_cost'].std(ddof=1)
cohen_dz = mean_diff / sd_diff

print(f"Mean Flanker Interference: {mean_diff:.2f} ms (SD = {sd_diff:.2f})")
print(f"Paired t-test: t({len(subj_summary)-1}) = {t_stat:.3f}, p = {p_val:.4e}")
print(f"Cohens dz: {cohen_dz:.2f}")

Frequently asked questions

How does the flanker task differ from the Stroop task?

While both paradigms measure cognitive control and response inhibition, they induce conflict through different mechanisms. In the Stroop task, conflict is semantic and intra-stimulus: a single printed word possesses two conflicting stimulus dimensions (ink color versus semantic word meaning). In the flanker task, conflict is visuospatial and inter-stimulus: the target item is spatially distinct from separate peripheral distractor items. The flanker task specifically taxes spatial selective attention and distractor suppression across visual space, whereas the Stroop task evaluates semantic inhibition and language automaticity.

What is considered a normal flanker effect in healthy adults?

In neurotypical adult populations tested with standardized arrow displays, the normal flanker interference effect ranges between 30 ms and 60 ms for reaction time, with incongruent error rates remaining between 2% and 6%. Tasks featuring more complex stimulus arrays (such as the Attention Network Test or letter matrices) can yield interference effects between 60 ms and 90 ms. Values substantially exceeding 100 ms indicate reduced distractor suppression or compromised attentional control.

Why are error trials excluded when calculating reaction time interference?

Error trials must be excluded from reaction time calculations because erroneous responses reflect premature guesses, motor slips, or lapses of attention rather than successful resolution of cognitive conflict. Because participants frequently make rapid anticipatory responses that result in errors, including error latencies artificially lowers the mean reaction time on incongruent trials, masking the true cognitive cost of inhibitory processing.

What is the Gratton effect in the flanker paradigm?

The Gratton effect (also termed trial-by-trial conflict adaptation) refers to the finding that the flanker interference effect is significantly smaller immediately following an incongruent trial than following a congruent trial. First documented by Gabriele Gratton and colleagues in 1992, this phenomenon demonstrates that the brain dynamically upregulates top-down cognitive control upon detecting conflict, narrowing the attentional spotlight to filter out distractors more effectively on the subsequent trial.

How does target-to-flanker spacing affect the magnitude of interference?

The magnitude of flanker interference is inversely related to the physical distance between the target and flanking distractors. Eriksen and Eriksen (1974) demonstrated that distractors exert maximal interference when positioned within 1 degree of visual angle from the central target. When distractors are spaced farther apart (exceeding 1.5 to 2.0 degrees of visual angle), interference approaches zero because the peripheral items fall outside the focal spotlight of visual attention.

Sources

  • Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16(1), 143–149. https://doi.org/10.3758/BF03203267 (checked 2026-11-14)
  • Fan, J., McCandliss, B. D., Sommer, T., Raz, A., & Posner, M. I. (2002). Testing the efficiency and independence of attentional networks. Journal of Cognitive Neuroscience, 14(3), 340–347. https://doi.org/10.1162/089892902317361886 (checked 2026-11-14)
  • Gratton, G., Coles, M. G., & Donchin, E. (1992). Optimizing the use of information: Strategic control of activation of responses. Journal of Experimental Psychology: General, 121(4), 480–506. https://doi.org/10.1037/0096-3445.121.4.480 (checked 2026-11-14)
  • Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D. (2001). Conflict monitoring and cognitive control. Psychological Review, 108(3), 624–652. https://doi.org/10.1037/0033-295X.108.3.624 (checked 2026-11-14)
  • Zelazo, P. D., Anderson, J. E., Richler, J., Wallner-Allen, K., Beaumont, J. L., & Weintraub, S. (2013). NIH Toolbox Cognition Battery (CB): Measuring executive function and attention. Monographs of the Society for Research in Child Development, 78(4), 16–33. https://doi.org/10.1111/mono.12032 (checked 2026-11-14)
  • Ridderinkhof, K. R., van den Wildenberg, W. P., Segalowitz, S. J., & Carter, C. S. (2004). Neurocognitive mechanisms of cognitive control: The role of prefrontal cortex in action selection, response inhibition, performance monitoring, and behavioral adjustment. Brain and Cognition, 56(2), 129–140. https://doi.org/10.1016/j.bandc.2004.09.016 (checked 2026-11-14)

Data: Flanker Task Interference Benchmarks Across Landmark Paradigms

Original cross-study compilation of reaction time benchmarks, conflict interference costs, and error rates across landmark Eriksen flanker paradigms, compiled from peer-reviewed experimental literature.

Period covered: 1974/2013. Coverage: Global.

Flanker Task Interference Benchmarks Across Landmark Paradigms
Study & ParadigmStimulus ModalityCongruent RT (ms)Incongruent RT (ms)Interference Cost (ms)Error Rate Incongruent (%)
Eriksen & Eriksen (1974) Letter ArrayLetter strings (H, K, C, S)404447436.2
Gratton et al. (1992) Conflict AdaptationLetter arrays (H, S) trial sequences382436547.8
Botvinick et al. (2001) Conflict MonitoringSimulated and empirical letter arrays395452575.9
Fan et al. (2002) Attention Network TestHorizontal arrow strings549633844.1
Ridderinkhof et al. (2004) Action SelectionDirectional arrow displays418479615.4
Zelazo et al. (2013) NIH Toolbox AdultDirectional arrow displays (ages 8+)435492573.8

Sources

Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about Flanker Task: Paradigm, Formulas, and Statistical Analysis

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →