The returning tide

Esta página en español. The atlas itself runs in both languages.

Why this channel exists

The atlas states its own gaps and until now could do nothing about them. Of the 313 connections it draws between 154 works, 163 are still marked "awaiting corroboration". The legend says so out loud and even offers a filter to isolate them. That is an uncommon kind of methodological honesty, but it was mute: anyone who knew where a given link came from had no way of saying so.

Two further imbalances lie beyond what one desk can fix. Of the 313 connections, 299 are resonances and only 14 are dissonances, a ratio of twenty to one that describes the reader rather than the field: continuities are easier to recognise than clashes. And the distribution by language reproduces the very academy the atlas criticises: 105 anglophone works against 28 Hispanic, 22 francophone, 6 Dutch, 2 Danish.

This is not participatory courtesy. It is the mechanism by which the atlas can grow precisely where its author has least reach, and by which a declaration of uncertainty stops being a gesture and starts being resolved.

The name comes from Brathwaite. Tidalectics has two movements, advance and withdrawal, and the atlas performed only one: the one running from corpus to reader. This is the water coming back. The public record of what has entered by another hand is called the tidemark, after the line water leaves on stone: memory and measurement at once.

Part one · For those who leave something in the current

What this is, and what it isn't

This is not a wiki. Nothing you send enters the atlas automatically. What you leave is a structured proposal that someone reads, weighs and decides on. Submissions are reviewed once a month, and that slowness is deliberate: a corpus that can be edited quickly is a corpus that can be degraded quickly.

Nor is it a suggestion box. The form asks of you exactly what the atlas demands of itself: a verifiable source and an argument. Without both, the submit button stays disabled. That isn't administrative friction; it is the project's rule of corroboration applied equally to the one who builds it and the one who reads it.

How to get in

The door isn't visible when the atlas loads. It appears once the visitor has actually travelled the map: six entries opened, or four minutes with the page active. Either one lights a line beneath the legend offering to "leave something in the current".

The reason is simple. A permanent "suggest a work" button turns the atlas into a letterbox and attracts submissions from people who have read nothing. A threshold of attention filters without demanding credentials, an institution or an account on any platform, which is the only barrier compatible with an open-access project.

Three shortcuts skip the threshold:

• The direct link con-el-agua-de-por-medio.html#reflujo, which opens the form at once. It exists so that someone can be invited without being made to travel first: paste it into an email, a seminar call or a syllabus.
• The A key in full-screen mode, for use in a class or a conference paper.
• The card titled "The returning tide" inside the atlas's conceptual framework, which leads to the same place.

Beyond that, every connection marked "awaiting corroboration" in a work's panel carries a small mark beside it that reads "know this one?". That is the short path: it opens the form with that connection already loaded, leaving only the source to be written.

The five kinds of contribution

Each corresponds to a distinct operation on the corpus, and choosing well saves everyone work.

• Missing work. Proposes a new entry. For a book, an article or an ethnographic manifestation the atlas ought to hold.
• Current. Draws a resonance between two entries already present that speak to each other without the map showing it.
• Friction. Draws a dissonance between two entries that clash where the map presents them as agreeing.
• Corroboration. Supplies the source for a connection already drawn but unverified.
• Correction. Fixes a fact: a year, an attribution, a place, a linguistic tradition.

Of the five, friction is the one most needed, given the twenty-to-one imbalance already noted; and corroboration gives the most return for the least effort, since 163 boxes stand open and closing one takes three minutes with the book at hand.

What counts as a verifiable source

A source is what lets another person check for themselves what you are claiming. Existing is not enough: it has to be locatable. A DOI works, as does an ISBN with a page, a full citation with a page, a stable URL from an institutional repository or journal, or an archive with its precise location. What does not work: "it is well known that", "I saw it in an article once", a link to an undated blog with no author, or a screenshot.

One nuance matters for corroboration: the source need not be either of the two works being linked. It may be a third text documenting the relation between them. If a commentator shows that Glissant read Brathwaite, that commentator is the source of the link, even if neither of the two says so in writing.

How to write the argument

The argument field takes a minimum of 200 characters and a maximum of 600. The minimum rules out "you should include X". The maximum forces a decision about what the argument actually is, rather than piling up everything one knows.

The question changes with the kind of contribution, and it is worth answering literally. For a missing work we do not ask what the book is about, but what it does that none of the 154 present entries does. For a friction we do not ask whether two authors differ, but what exactly they disagree on and in which text it shows. An example drawn from the corpus itself:

Not enough: "Mintz and Ortiz have different approaches to the Caribbean and it would be good to mark that."

Enough: "Anglophone political economy against Hispanic cultural transculturation. Mintz derives Caribbean culture from the socio-economic matrix of the plantation; Ortiz derives it from the counterpoint between cultural elements. This is not a difference of emphasis: it is what each takes as cause and what as effect."

Note what the second does and the first doesn't: it names the point of disagreement, locates it in each author's work, and explains why it is not a minor difference. That is what makes it possible to decide whether the connection enters.

Your name and your data

The form asks for a name, as you wish to be cited, and an email address. Affiliation and ORCID are optional. The address is used only to reply to you and is never published.

You may tick the anonymous box: the contribution is judged the same way and, if accepted, appears as "anonymous contribution". It exists because some contributions can expose those who make them, and because choosing not to appear is a legitimate decision that should not cost the contribution. Consent to data processing is required and states plainly what you are consenting to; you may request deletion by writing to the group.

What happens next

Submissions arrive at the group's address as a structured block. Once a month the accumulated contributions are read and decided on. If yours enters, it is ingested into the corpus and your name is recorded beside the exact item you contributed, not in a general acknowledgement, and the affected entry carries in its footer a mark of who proposed, corroborated or corrected it. Five accepted contributions bring you into the contributors of the next deposit of the dataset on Zenodo, under data curation: a citable line, not a thank-you note. If a contribution is declined, you are told why. A refusal with an argument is more use than silence.

Part two · The criteria for deciding

This part is published for the same reason the methodology is: whoever contributes has a right to know the measure being applied, and whoever audits the project has a right to check that the measure exists.

The five criteria

They are checked in this order, because the first ones rule out quickly and the last ones require reading.

• Source. Can it be verified in under ten minutes? If not, the precise reference is requested before anything else is weighed.
• Relevance. Is it the Greater Caribbean, or at least a work that thinks the basin from outside with consequence? If not, it is declined with an explanation.
• Non-duplication. Is the link or the work already there under another name? If so, the reply points to the existing entry.
• Argument. Does it name the exact gap, or merely praise the work? If it merely praises, it goes back with a request for precision; it is not declined outright.
• Balance. Does it correct a declared bias of the corpus, or deepen it? Not grounds for refusal, but grounds for priority.

The fifth criterion needs a warning. That a contribution corrects a bias does not by itself make it admissible: a poorly grounded Dutch-language work is still a poorly grounded contribution. Balance orders the queue; it does not lower the standard.

Difficult cases

Someone proposes their own work. Not grounds for refusal. Self-proposal is common in open corpora and often comes from those who know the field best. It is judged by the same five criteria, and in the record the verb is "proposed", which already declares the relation.

The contribution is good but the argument is badly written. The contribution is accepted and the entry rewritten. The record says "proposed", not "wrote". Distinguishing the two verbs is what keeps the credit honest.

Two people contribute the same thing. Credit goes to whoever arrived first by date of submission. If the second brings a better source, both are credited with different verbs: "proposed" and "corroborated".

A contribution questions a decision of the corpus, not a fact. That certain work should not be classed under post-plantation, say. That is not a correction but a methodological discussion, and it deserves an answer on the methodology page rather than a silent change in the data.

How credit works

Three levels, grounded in how academic credit actually functions rather than in a badge system.

• Notice. One accepted contribution: the name enters the tidemark beside the exact item and in the footer of the affected entry.
• Sustained contribution. Five accepted items, or a complete entry that enters as it arrived: the above, plus contributor in the next Zenodo deposit under data curation, and an entry in the repository's citation file.
• Co-authorship of the corpus. A body of entries that changes the shape of the atlas — taking the Dutch-language corpus from six to twenty-five, or doubling the recorded frictions — enters the citation of the DOI.

The verbs of the record are five and not interchangeable: proposed, wrote, corroborated, corrected, translated. A record that said "collaborated" in every case could not be audited, which is precisely what the tidemark must allow.

Where this stands

• The channel is active and submissions reach the group by email.
• The tidemark is published, with the open state of the corpus; the ledger starts empty and will grow with the first accepted contributions.
• The data-processing notice is awaiting review by the University's legal office.

As elsewhere in this project, what is missing is declared rather than hidden.