Retraction, correction and expression of concern
A correction — sometimes an erratum or corrigendum — fixes an error that does not undermine the paper's conclusions. The article stands; a notice records the change.
An expression of concern is a flag. The editor has been alerted to a potential problem that is not yet resolved, and is telling readers to treat the findings with caution while an investigation proceeds. It may be lifted, or may lead to retraction.
A retraction says the findings should not be relied upon. The article remains available — it is not deleted, because the record of what was published must stay intact — but it is marked prominently as retracted, and the notice states who requested it and why.
A retracted paper is never deleted. It stays available, clearly marked, so the record of what was published remains complete.
Not all retractions imply misconduct
A substantial share of retractions arise from honest error: a coding mistake, a mislabelled sample, a reagent later shown to be contaminated. Authors frequently request these themselves on discovering the problem.
Self-retraction of this kind is a mark of integrity, not a stain, and is generally understood as such within a field. It is also the correct action — leaving a result you know to be wrong in the literature causes real damage to people who build on it.
Retractions arising from misconduct — fabrication, falsification, plagiarism, image manipulation — are a different matter and carry consequences from the institution as well as the journal. The retraction notice normally makes clear which category applies, which is why the wording of that notice is worth attending to.
How the process runs, and what authors should do
Concerns typically reach the editor from readers, reviewers, or automated screening. The editor contacts the authors, and in serious cases the institution runs its own investigation. Publishers generally follow COPE guidance, which sets out the steps and the expectation that authors are given the chance to respond.
If you are contacted, respond promptly and factually, and involve your institution's research integrity office early — that is what it exists for. Do not attempt to resolve a serious allegation informally with the editor alone.
If you discover an error in your own published work, raise it yourself. Contact the editor, explain the problem, and say what you believe the appropriate correction is. Authors who self-report are treated very differently from those who are found out.