Publishing a study

The publication states

Which moves are legal

The graph is enforced server-side, so a status can never be set by accident from a UI that got out of date. The common path is the top line:

• under_review may also go to approved first, and approved then to published — the two-step route used with reviewer governance.
• needs_revision can go back to submitted, under_review, draft, or rejected.
• rejected is not a dead end: it can return to draft or submitted.
• published has exactly one outgoing move: retracted, which needs a reason and a public notice. Nothing can quietly revert to draft after people have started citing it — see moderation and integrity.
• retracted is not a dead end either: corrected work can go back to draft or submitted and be reviewed again. The retraction stays on the record.
• submitted and under_review can return to draft — that is you withdrawing your own submission from review.

Authors linked by email

When you publish from BCILattice, each author you list with an email address is invited to confirm authorship. These invitations never hold the study back: the study publishes, and an author who has not answered is shown as invited next to their name.

• If the email belongs to a verified BCINexus account, that researcher gets an in-app notification and an email. Once they accept, their name links to their profile and the study appears under Co-authored Studies there.
• If the email is not registered yet, they get an email invitation. When they sign up and verify that address, the authorship joins their new account.
• Only verified addresses are matched, so nobody can claim an authorship by registering someone else's email.
• Email addresses are never shown on the study page. Which account an author is stays private until they accept.

Publishing, step by step

1. 1

   Fill in the study metadata

   Title, signal modality, task type, and an abstract or description are the hard minimum. Licence, keywords, author byline with affiliations, and references are strongly recommended and feed the readiness score.

2. 2

   Attach the work itself

   At least one pipeline definition — NeuralFlow, ML Suite, or Workflow — must be present, either on the study or on one of its experiments. A study with metadata and no pipeline has nothing to publish.

3. 3

   Invite co-authors and wait for acceptance

   Add the byline with contributor roles. Required co-authors receive an invitation and must accept before the study can move past submission.

4. 4

   Run the readiness preflight

   The desktop publish dialog shows blocking issues, warnings, and a score. The server runs the same rule set again on the publish hop, so passing locally is not a formality that can be skipped by calling the API directly.

5. 5

   Submit

   Submitting creates a review submission on the default review pipeline and drops the study into the pool that eligible reviewers pick from. You do not have to find a reviewer yourself.

6. 6

   Respond to review, then publish

   Answer comments, resubmit if changes were requested, and on final approval the study is published and becomes publicly visible.

What happens the moment it publishes

• Visibility flips to public. This is the one place that rule lives, so every publish route behaves identically.
• The study is indexed for similarity and recommendation, so related work starts surfacing next to it.
• You and any linked co-authors gain the researcher capability, which is what puts the badge on your profile.
• A transition record and an audit entry are written, giving the study a visible, permanent history.
• The submission's completed reviewers are added to its contributor list — anonymised if the pipeline was blind.

How a published study is shown

A published study has two faces. Both are made from what you ticked in BCILattice's Publish page, and nothing else.

The Summary

• Key result. Each model's score as mean ± SD across participants, with the validation it was measured under. It is never a single best number, and never a training-set score.
• Key figures. You choose them: on the Publish page, open Images & tables and tick Key figure on up to four. BCILattice never picks them for you. Until you tick one, the Summary shows none.
• Method in brief. One row per step you actually did, in the order the work is done. Each row opens that step in the full study.
• Findings. Group Analysis results, one line per test, with their caveats printed in full beneath.
• Citation. Shown on the page with a copy button; other formats are one click away.

The Publish page's Summary tab previews this exactly, and prints its length. Aim for under 600 words: a long abstract or conclusion is cut at a sentence end.

The full study

• Each page opens with one line saying what the step did, then its figures (named as in the app), a few facts, and the first rows of its table.
• Everything else you published is still there, in folds that are closed until opened: All settings and Raw data.
• A step one experiment takes unchanged from another is marked same as Experiment 1 and shown once in the Summary.
• A run opens as the app's result tabs: each model, each participant, the graphs for one model and participant at a time, and the training curves.
• Model Analysis and Phase Analysis show only the pictures and tables you saved with Save for publishing: a figure, a table, a group of headline numbers, or the projection explorer's picture.
• Every picture and table is also a row in What to publish, under its page, and can be unticked there or on Images & tables. It is the same tick.

The study's models

However many models a study trains, they are one entry in the model registry: a model set named after the study. Its page lists each model, then its runs, participants and sessions. Publishing the study again updates the same set and the same models, at the same addresses.

Training knowledge goes out only with a trained model, picked under the run that trained it. A pipeline or workflow is published as its design, and its page names the models it holds.

Versions and updates

Published studies are versioned rather than edited in place. A new version is a new snapshot with its own history, so a citation keeps pointing at what the citer actually read. Review submissions carry their own independent version counter — resubmitting during review bumps that, not the public version.

Citation and identifiers

• Every published study gets a citation export you can paste into a reference manager.
• Author identity is strongest when co-authors have verified ORCID identifiers on their accounts.
• Affiliations let work roll up to an institution, which is why the readiness check nags about them.

Related