Specifications
What it means to implement the paradigm correctly
A paradigm specification records the construct a paradigm operationalises, its defining contrast — the manipulation and prediction that make it that paradigm and not another — the design parameters that materially move results, and an executable test.
Human review
Has a domain expert judged the measurement claims? All 11 specifications sit at
sourced. That is the honest default and already useful. It is not an endorsement.
Machine checks
Does an implementation instantiate the contrast and scoring? A responder with a known policy is driven through the real task; the contrast must be logged as a first-class factor, scored, and recovered. A pass is drawn in plain ink here, never in green, because it is not review.
The human-review ladder
Four states, one axis. A specification climbs it only when a named person signs a claim.
sourcedLiterature-linked, drafted with LLM assistance or by a contributor, and unreviewed. The honest default.corroboratedIndependently cross-checked against the cited sources, but not signed by a named expert.expert-annotatedAll three claims attested by named experts. Derived, never self-declared.contestedA named expert disputes at least one claim. A real outcome, kept visible.
The specifications
| Paradigm | Family | Defining contrast | Human review | Executable test |
|---|---|---|---|---|
| Color Delayed Estimation (Visual Working Memory) | continuous-report | guess-rate (mixture-decomposition) | sourced — unreviewed | verify:task |
| Eriksen Flanker Task | congruency | flanker-effect-rt (condition-difference) | sourced — unreviewed | verify:task |
| Go/No-Go | go-no-go | commission-error-rate (selective-suppression) | sourced — unreviewed | verify:task |
| Lexical Decision Task | speeded-classification | lexicality-effect-rt (condition-difference) | sourced — unreviewed | verify:task |
| Motion Coherence Discrimination (Random-Dot Kinematogram) | psychophysics | coherence-threshold (psychometric) | sourced — unreviewed | verify:task |
| N-back | n-back | d-prime (load-response) | sourced — unreviewed | verify:task |
| Sequential Congruency (Gratton effect) | sequential-congruency | congruency-sequence-effect (interaction) | sourced — unreviewed | verify:task |
| Simon Task | congruency | simon-effect-rt (condition-difference) | sourced — unreviewed | verify:task |
| Stop-Signal Task | stop-signal | ssrt (race-model) | sourced — unreviewed | verify:task |
| Stroop Task | congruency | stroop-effect-rt (condition-difference) | sourced — unreviewed | verify:task |
| Task Switching | task-switching | switch-cost-rt (condition-difference) | sourced — unreviewed | verify:task |
Audited reference instances 9
Real third-party implementations driven end-to-end through the executable spec. The verdict says an implementation does or does not carry the paradigm's defining contrast and scoring. It says nothing about the spec's review state, and nothing about construct validity.
| Implementation | Platform | Specification | Verdict | Note |
|---|---|---|---|---|
| demos/eriksen_flanker | Pavlovia / PsychoJS | Eriksen Flanker Task | conforms | congruency condition logged, not inferred — the namesake paradigm validating against its own contract |
| arrow-direction-task | NCCR / jsPsych | Eriksen Flanker Task | non-conforming · contrast-not-logged | an arrow choice-RT task with no congruency manipulation — looks like a flanker, isn't |
| onlineworkshopstroop (dejan.draschkow) | Pavlovia / PsychoJS | Stroop Task | conforms | a classic 3-colour Stroop; same validator/analysis as the Flanker — the contract is framework-agnostic |
| demos/butterfly_simon | Pavlovia / PsychoJS | Simon Task | non-conforming · contrast-not-logged | a real Simon, but congruency is never logged — the not-logged/semantic floor; conforms only if the author emits the derived congruency |
| cpt-cities-mountains | NCCR / jsPsych | Go/No-Go | conforms | a gradual-onset CPT (gradCPT) — RT is on a different clock; a curator should tag the variant |
| demos/go_nogo | Pavlovia / PsychoJS | Go/No-Go | conforms | a newer PsychoJS (2023.2.3); the integration is not congruency-specific |
| TaskBeacon H000012-sst | psyflow-web / jsPsych | Stop-Signal Task | conforms | SSRT is estimable end-to-end, but a synthetic oracle isn't a race process — the value is a wiring check, not a validity claim |
| demos/n_back | Pavlovia / PsychoJS | N-back | conforms | mouse-driven; oracle policy recovered, d' ≈ 3.08 — an n-back the survey's static probe could only call undeterminable, shown conformant when driven |
| lexical-decision-task | NCCR / jsPsych | Lexical Decision Task | conforms | the clean centre — lexicality effect computable and recovered, nothing for a human to flag about the wiring |
None of this is expert-reviewed yet
The specs are the part of the catalog a machine cannot finish. Reviewing one is a bounded ask: three separable claims, an afternoon, a named and revocable sign-off.