In a survey among chemists (April-May 2023), we found that 88% of them have already spotted an error (honest or not) in a publication, and 50% in their own papers. Errors are therefore common in scientific literature, and when they are detected, they must be corrected to mitigate the consequences of their propagation. The "cleaning of the published body of evidence" (Bouter, 2023) can take several forms: a formal correction of the scientific record with correction/retraction notices issued by publishers, or a variety of self-correcting processes that affect the scientific record in a less formal way (Dougherty, 2019). Such processes reflect researchers' behaviours when faced with errors. They need to be studied in order to assess which ones can be fostered and promoted. We surveyed researchers to identify their attitudes towards errors, and to gather their views on the effectiveness of different correction processes, including through free-text answers provided by 40% of the respondents. We analysed 977 responses from respondents having completed at least 67% of the survey. Among the respondents who have already spotted an error, 79% state they had taken action. Of these, very few undertook a replication project (4%), published a formal refutation (4%) or a comment on PubPeer (2%), or submitted a Letter/Comment/Note to the journal (13%). But those who did are the ones who most often agree that those actions do correct science (85-96%). By contrast, the most frequently cited action being taken is to mention the issue in private discussions; but only 55% agree that this corrects science. For 55% of respondents, the main motivation for taking any kind of action is "As a matter of principle, if there is an error, it should be corrected". This shows that researchers are quite willing to act to correct errors (their own as well) and have internalised this as a research integrity practice. Nevertheless, the verbatims show unambiguously that good intentions are often hindered by reluctant publishers/editors rather than by shyness or fear of hurting peers, even if social relationships, and in particular the presence of a "hierarchical" link with a peer, are sometimes presented as an obstacle to taking action to correct an error. Together with the other results provided by this survey, we present correction as drawing on many virtues and as a good practice to be stimulated, and promoted in training courses, codes of conduct and the media. Special efforts should be directed at researchers with editorial responsibilities in order to improve the peer-review process and post-publication correction processes. References Bouter, L. (2023). Research misconduct and questionable research practices form a continuum. Accountability in Research, 1–5. 10.1080/08989621.2023.2185141 Dougherty, M. V. (2019). Correcting the scholarly record for research integrity. Monash Bioethics Review, 37(1–2), 76–78. 10.1007/s40592-019-00093-x Note The survey was anonymous and has been assessed by Université Grenoble-Alpes and Maastricht University, registered at Université Grenoble-Alpes; Qualtrics survey tool has been provided by Maastricht University; questions have been proofread by NanoBubbles' ethics rapporteur.