Written and maintained by CASRAI Editorial Board
Published
On this page
The Posner cueing task is a spatial-attention experiment that compares responses to targets at cued and uncued locations while participants maintain central fixation. Its reaction-time contrasts measure cueing effects; a neutral condition allows separate estimates of benefit and cost. The published example below yields a 19 ms benefit and a 28 ms cost, with calculations you can reproduce.
Definition of the Posner cueing task
A trial presents a cue before a target, and the participant detects the target or identifies a property of it. A valid cue indicates the target location; an invalid cue indicates another location. A neutral cue supplies no directional advantage. Testable’s peripheral-cue implementation, for example, flashes one box for directional cues and both boxes for neutral trials, then requires a response indicating the target’s side. This illustrates one implementation rather than a fixed recipe: the stimulus, required response and baseline must be specified for each experiment. Testable’s task documentation describes those conditions.
In his retrospective Orienting of Attention: Then and Now, Michael Posner emphasizes measuring attention shifts without eye movements. This is the meaning of covert orienting. Peripheral changes and central symbolic cues offer different ways of directing attention, but his account also cautions that arrows need not produce purely endogenous orienting. Describe the cue physically and state the instructions, rather than making the word “arrow” carry the entire theoretical interpretation.
Cueing-effect formula: benefits and costs
Define the subtraction before interpreting its sign. Using the positive-benefit convention documented in the HED Working Group’s Posner task catalog:
- E = I − V: total cueing effect.
- B = N − V: benefit relative to the neutral baseline.
- C = I − N: cost relative to the neutral baseline.
Here V is mean reaction time on valid trials, I is mean reaction time on invalid trials, and N is mean reaction time on neutral trials. E, B and C denote the resulting differences. All six quantities are expressed in milliseconds when the input reaction times use milliseconds. A positive B means the valid condition was faster than neutral; a positive C means invalid was slower than neutral.
The decomposition follows directly: B + C = (N − V) + (I − N) = I − V = E. The neutral terms cancel. Reversing a subtraction reverses its sign, so an analysis using valid minus neutral must label that convention explicitly. Without N, E can still be calculated, but B and C cannot be separated by these equations.
Worked example from published reaction times
Posner, Snyder and Davidson’s 1980 paper, Attention and the Detection of Signals, reports mean reaction times for location and form cues. In Table 1 on page 166, take the Neutral form row and the Expected, Neutral and Unexpected location columns. Their values are 252, 271 and 299 ms, respectively. Keeping the form row fixed avoids mixing location effects with different form-cue conditions.
- Assign V = 252, N = 271 and I = 299.
- Benefit: B = 271 − 252 = 19 ms.
- Cost: C = 299 − 271 = 28 ms.
- Total effect: E = 299 − 252 = 47 ms.
- Check the identity: 19 + 28 = 47 ms.
These are calculations from published condition means, not newly collected observations. In this comparison, the cost exceeds the benefit by 28 − 19 = 9 ms. Relative to the valid-condition mean, the invalid-condition mean is longer by 47 ms; relative to neutral, the two contributions are unequal. Reporting the total alone would conceal that asymmetry.
The same 47 ms total could be partitioned differently around a different neutral baseline. That is an algebraic reason to retain the baseline definition alongside the result. These three means alone do not supply participant-level variability, an uncertainty interval or a new significance test. The calculation reproduces a descriptive contrast, with no claim that 47 ms is a population norm.
Interpretation of cueing effects
A contrast describes the relationship between conditions under a specified design. Read it together with cue type, cue validity and the cue-to-target interval. HED lists these as task manipulations and includes inhibition of return at longer stimulus onset asynchronies, or SOAs. Consequently, the sign of a result belongs with its temporal and experimental context.
| Contrast pattern | Reading guidance under the stated equations |
|---|---|
| E is positive | The invalid-condition mean is longer than the valid-condition mean. |
| B and C are both positive | The neutral-condition mean lies between valid and invalid means. |
| E is negative | The valid-condition mean is longer than invalid; examine the task timing and conditions before interpreting why. |
The table’s directional statements are arithmetic consequences of the formulas. They are not diagnostic categories. A claim about a psychological mechanism needs evidence that the comparison isolates that mechanism; this is a construct-validity question.
Posner’s retrospective also distinguishes orienting, alerting and executive attention in the Attention Network Test. Its surrounding flanker arrows introduce conflict. A flanker-task interference effect therefore answers a different experimental question from a spatial cue-validity contrast. Record which contrast the task actually supports before describing it as a general measure of attention.
Assumptions and common mistakes
Separate a neutral baseline from a no-target trial
A neutral trial still provides a target response from which N can be estimated. A catch trial with no target does not provide that same target-detection baseline. Millisecond’s Inquisit implementation makes no-target control trials optional; its documentation describes responding before a target and failing to respond as separate events. Treating a no-response timeout as an ordinary neutral RT would change the quantity being calculated.
The data table below compares three documented implementations by cue, required response and neutral condition. It shows why the task name alone does not establish that all three contrasts are available. A study using only valid and invalid conditions can estimate their difference, but cannot recover a missing neutral mean by relabeling an existing condition.
Interpret exported codes before calculating means
PsyToolkit’s documentation separates cue position, target position, cue validity, response time and response status. Its status codes are 1 for correct, 2 for wrong and 3 for timeout. Its validity variable also uses numbers, but with a different meaning: 1 is cued and 0 uncued. A filter on “1” is therefore meaningful only after identifying the column.
Testable describes an analysis restricted to correct responses. Applied to a trial export, that requires retaining accuracy information alongside condition labels, rather than averaging every numeric latency indiscriminately. Preserve the distinction between an incorrect response and a missing response; document how each enters the analysis. These are data-definition decisions, not an invitation to invent a universal acceptable-RT range.
Keep timing, practice and repeated trials visible
The Inquisit technical manual documents default SOAs of 100 and 500 ms and identifies practice versus test trials. It also distinguishes original trials from makeup trials, which repeat early-response and no-response trials by default. An exported row count can therefore include repeated attempts. State how those flags were handled before describing the number of observations analyzed.
For a temporal question, retain the SOA condition instead of collapsing it away during preprocessing. Likewise, keep the physical cue and response requirements in the analysis record. These recommendations apply the documented fields to the comparison being estimated; they do not establish one optimal timing schedule for every Posner experiment.
Report exclusions and accuracy alongside latency
APA’s JARS–Quant Table 1 calls for describing outlier processing, data distributions, missing-data handling and exclusion criteria. For this task, make that concrete: specify which response statuses were excluded, how reaction-time outliers were defined, and how many usable observations remained in each condition. Include accuracy summaries so readers can evaluate the latency contrast in its response context. Reporting those decisions makes the numerical result reproducible without presenting an arbitrary trimming rule as a property of the paradigm.
Posner cueing task and dot probe task
The dot probe task is related to spatial cueing but uses a different comparison. The HED dot-probe entry describes two stimuli followed by a probe at one of their locations. The participant detects or classifies the probe; reaction-time differences between congruent and incongruent trials index attentional bias associated with the stimulus pair. HED assigns it a separate task identifier from Posner spatial cueing.
For a methods record, specify whether the manipulation concerns a directional cue’s validity or a probe’s relationship to one member of a stimulus pair. Similar-looking subtraction does not make the conditions interchangeable. Preserve the stimulus categories and probe-location coding when documenting a dot-probe analysis. Do not rename its conditions “valid,” “invalid” and “neutral” unless those labels accurately describe the experiment that was run.
How to report it: an APA-style example
For an APA 7 manuscript, connect the task description to APA’s JARS–Quant reporting standard. Identify the outcomes, condition summaries and analytic decisions. JARS also requests effect-size estimates and confidence intervals where possible, exact p values when null-hypothesis tests are used, and a distinction between primary, secondary and exploratory analyses.
Illustrative reporting example using published aggregate data: “For the neutral-form condition reported by Posner et al. (1980, Table 1), mean reaction times were 252 ms at the expected location, 271 ms with neutral location information, and 299 ms at the unexpected location. Defining benefit as neutral minus expected and cost as unexpected minus neutral yielded a 19 ms benefit and a 28 ms cost; unexpected minus expected yielded a 47 ms cueing effect.”
This example reports the descriptive calculation only. In a report of your own study, accompany it with sample and cell counts, appropriate dispersion, and the actual inferential analysis. Give the software and analysis specification. A confidence interval belongs to an estimate with an identified analysis; it cannot be reconstructed from three condition means alone. If trials are summarized within participants, state that aggregation explicitly so readers can distinguish participant summaries from the underlying observations.
How to calculate it in Excel, R and Python
The following expressions reproduce the published-mean example. They accept condition summaries rather than raw trial data. Keep V, N and I in the same units and in the same order across programs; otherwise an apparently successful subtraction can answer a different question.
Excel
Enter valid, neutral and invalid means in A2, B2 and C2 respectively: 252, 271 and 299. Calculate benefit with =B2-A2, cost with =C2-B2, and total effect with =C2-A2. The results are 19, 28 and 47. A separate check, =(B2-A2)+(C2-B2)-(C2-A2), returns zero.
R
V = 252; N = 271; I = 299
c(benefit = N - V, cost = I - N, effect = I - V)
# benefit 19; cost 28; effect 47
Python
V, N, I = 252, 271, 299
scores = {"benefit": N - V, "cost": I - N, "effect": I - V}
assert scores["benefit"] + scores["cost"] == scores["effect"]
print(scores) # {'benefit': 19, 'cost': 28, 'effect': 47}
The zero residual and matching outputs check the arithmetic and sign convention. They do not check trial selection, fixation, accuracy or the experimental design. Those checks belong upstream, before the condition means enter these expressions.
Sources
- Posner, Snyder and Davidson, Attention and the Detection of Signals (1980), especially Experiment 2 and Table 1, p. 166. Checked 2026-11-16.
- HED Working Group, Posner Spatial Cueing Task. Checked 2026-11-16.
- HED Working Group, Dot-Probe Task. Checked 2026-11-16.
- PsyToolkit, Spatial orientation / cueing / Posner Task. Checked 2026-11-16.
- Testable, Posner Cueing Task. Checked 2026-11-16.
- Michael I. Posner, Orienting of Attention: Then and Now. Checked 2026-11-16.
- American Psychological Association, JARS–Quant Table 1. Checked 2026-11-16.
- Millisecond Software, Inquisit Posner Cueing Task technical manual. Checked 2026-11-16.
Data: Posner cueing implementations: response and neutral baseline
Original comparison compiled from the original Posner, Snyder and Davidson Experiment 2 and the platform documentation for the PsyToolkit and Testable demos. It distinguishes cue presentation, target response and whether a neutral target condition is documented. These are implementation-specific designs, not universal requirements or a performance ranking.
| Implementation | Directional cue | Required target response | Neutral target condition | Source |
|---|---|---|---|---|
| Posner, Snyder and Davidson (1980), Experiment 2 | Central left or right arrow for location information | Simple reaction-time response to the appearance of a letter | Plus-sign location cue with either location equally probable | Original paper, Experiment 2 methods and Table 1 |
| PsyToolkit cueing demo | X in the left or right box | A for left GO target; L for right GO target | Not included in the documented valid/invalid demo design | PsyToolkit implementation and output documentation |
| Testable peripheral-cue demo | One peripheral box flashes | Arrow-key response indicating the side of the target | Both boxes flash before a target appears on one side | Testable task description and trial conditions |
Sources
- Posner, Snyder and Davidson (1980), Attention and the Detection of Signals, Experiment 2 (accessed 2026-11-16)
- PsyToolkit: Spatial orientation / cueing / Posner Task (accessed 2026-11-16)
- Testable: Posner Cueing Task (accessed 2026-11-16)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







