When and how is preregistration in experimental biology worth the effort?
This is an uncorrected proof.
Figures
Citation: Amaral OB, Carneiro CFD, Valério-Gomes B, Neves K, Abreu MB (2026) When and how is preregistration in experimental biology worth the effort? PLoS Biol 24(11): e3004129. https://doi.org/10.1371/journal.pbio.3004129
Published: November 20, 2026
Copyright: © 2026 Amaral et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The Brazilian Reproducibility Initiative, which has inspired the reflections in this article, has been supported by grants from the Serrapilheira Institute, FAPERJ (E-26/200.824/2021) and CNPq (310813/2021-2) to OBA, and by the Einstein Foundation Institutional Award for Improving Research Quality to the consortium. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Kleber Neves and Clarissa Carneiro are now employed by the Serrapilheira Institute (a philanthropic organization with no financial stake in the results). Their participation as authors in this article relates to their previous participation in the project and is independent of their current work at the Institute.
Research programs in biology advance through iteration of exploratory work that generates hypotheses and confirmatory experiments that test them [1]. The scientific literature, however, does not follow this distinction strictly. Exploratory analyses undertaken after data are collected can suggest that a hypothesis was confirmed by results, when they in fact helped generate it; a problem known as HARKing (hypothesizing after results are known) [2]. This allows researcher biases to define what analyses get carried out or reported, distorting the scientific record [3].
One way of addressing this problem is to register study protocols ahead of data collection. The proposal has a long history, but it started to gain traction in the 1990s, when the US FDA started mandating the registration of clinical trials. Later, the so-called ‘reproducibility crisis’ in psychology led to a bottom-up movement for preregistration of analysis plans. This also led to the creation of the Registered Reports format, where a study protocol is peer-reviewed before data collection, leading to in-principle acceptance at a journal regardless of the results [4].
Preregistration has many proposed benefits [5]. It generates transparency in data collection and analysis, allowing readers to distinguish between a priori and a posteriori decisions. It stimulates authors to consider analysis issues before data collection and, with Registered Reports, allows peer review to occur when it can impact research design. It can also mitigate publication bias, as it allows tracking of unpublished studies and, in the case of Registered Reports, reduces the impact of results on publication decisions [6].
Nevertheless, there has been pushback. Arguments against preregistration include limiting researcher flexibility, increasing workload, delaying the research process and facilitating scooping. While negative reactions to preregistration can occur in other fields [3,5], they seem to be the norm in biology: a survey of animal researchers in Switzerland presented mostly negative views, with only 10% of researchers having tried the practice [7].
Our experience shows that implementing preregistration in lab biology is not trivial. In the Brazilian Reproducibility Initiative [8], a multicenter effort to replicate biomedical experiments, we made an extensive attempt to preregister methods, requiring researchers to prespecify protocol decisions and criteria to assess the validity of experiments [9]. Nevertheless, they had a hard time defining these in advance and frequently ended up breaking their own protocols [10]. This raises the question of whether this is a cultural issue, or whether there are aspects of experimental biology that make it ill-suited for preregistration.
It may not be accidental that preregistration has had its greatest impact in fields such as psychology and clinical medicine, where data collection is usually straightforward (e.g., a survey or clinical assessment) but analysis options abound, as many variables are typically collected. Lab-based researchers, on the other hand, usually have a harder time collecting data than analyzing them. If data collection is complex, protocol adaptations can be inevitable, not as an attempt to nudge results but simply to get a method to work. Flexibility in experimental protocols is thus welcome as it allows researchers to adapt to conditions. This is especially true when non-standard methods are set up in an ad hoc manner for an experiment; something that frequently happens in lab science. The need for flexibility is also greater when experiments are performed by researchers who are still on the learning curve towards an established routine.
For data analysis, however, preregistration increases in value. On the cost side, the lack of external contingencies makes it easier to plan ahead (and to test whether plans are feasible with mock data). The benefits of limiting flexibility are also greater: analyses can be repeated cheaply in myriad ways, leading to the possibility of p-hacking (i.e., experimenting with analysis options in order to reach a significant or desired result) if analysis pipelines are not prespecified [11]. While completely specifying an analysis beforehand effectively prevents p-hacking, it does little to prevent the selective reporting of experiments, as one can still push for a particular result by redoing the same procedure and cherry-picking what to report.
It is thus likely that the concept of preregistration needs to be adapted to fit the needs of experimental biology. Some types of articles, such as replications or extensively planned confirmatory experiments, are better suited to the format and present fewer difficulties. However, even for these cases, it is useful to reflect on what is worth preregistering in detail and when to do it (Fig 1). Experimental protocols meant to be reviewed, as in the case of Registered Reports, should be detailed enough to allow one to judge whether methods are appropriate. In this case, pilot experiments are worth doing before registration to confirm the protocol’s feasibility and prevent deviations. Nevertheless, specific details such as running a PCR for 25 or 30 cycles or growing cells for 5 or 7 days are frequently data-driven decisions within an experiment, and it may be better to allow flexibility in such steps.
Preregistration is better suited for confirmatory research (left side), and Registered Reports should only be considered in those cases; the amount of experimental protocol detail needed depends on whether peer review will be sought, while analysis decisions and inclusion/exclusion criteria for experiments, animals or samples should always be prespecified. For exploratory research (right side), full preregistration is unlikely to work, but options exist to preserve parts of the concept.
If preregistration is not meant for peer review, specifying experimental details becomes less important, and the focus should be on the analysis. Statistical analysis plans should be as detailed as possible and, in cases where decisions depend on observing the data (such as whether data transformations should be performed), clear decision rules should be provided. This also applies to rules concerning the inclusion or exclusion of experiments, animals or samples. This does not mean that unplanned analyses cannot be carried out, but it is important that they are labeled as such.
Ultimately, however, most traditional biomedical papers will not be good candidates for a full preregistration or Registered Report. A cell biology article where the authors start by screening for molecules that interfere with a pathway, select a promising candidate, and perform a series of custom-designed assays to establish a mechanism of action cannot specify all these steps in advance, as the initial experiments define what will be done later. Still, it is worth asking what aspects of preregistration can be preserved in these cases.
One option is to break down long projects into exploratory and confirmatory stages [12], which need not be published together. In the exploratory stage, researchers are free to perform pilot experiments, refine their hypotheses, and set the stage for a confirmatory experiment, which is then preregistered. Publication-wise, this may lead to two different articles, and a solid research story would ideally include both.
A distinct approach is to try to bring in pieces of the preregistration concept within exploratory projects. Even when experiments inform each other, each of them will still have a protocol, and these can be preregistered sequentially in time-stamped fashion as they unfold. They will not be suitable for a Registered Report, but will still provide an experimental log to understand researcher decisions, especially if shared along with a lab notebook [13]. This also allows a reader to evaluate whether experiments were developed as reported, mitigating authors’ tendencies to rework their narrative to tell a better story [14].
A third option is to preregister general standards for frequently used methods, rather than individual experiments. These can include analysis pipelines, validity criteria for samples and other decisions that may be biased by results. Labs would generally be expected to follow their preregistered standards, and deviations from those practices would be identifiable as exploratory. This, however, is only feasible for well-established methods, and the ‘minimum package’ to be registered should be established by the research community using them.
Finally, one should remember that preregistration will not solve selective reporting on its own. Even when methods are set, there will still be the temptation to discard experiments or data points that do not conform to expectations based on post-hoc criteria. Attempts to predefine rules for exclusion are desirable but will never cover all cases; thus, alternatives such as providing a complete experimental log that records discarded experiments, as well as sensitivity analyses that include them, can be used to make these decisions transparent.
Although there is resistance to preregistration in lab biology, some of which may be justified, minimizing bias is an ethical duty for researchers, and preregistration should be considered as a potential strategy. Limiting flexibility in analysis and preventing selective reporting should be priorities, while trying to anticipate every decision in data collection may prove unfeasible and hamper adoption.
References
- 1. Kimmelman J, Mogil JS, Dirnagl U. Distinguishing between exploratory and confirmatory preclinical research will improve translation. PLoS Biol. 2014;12(5):e1001863. pmid:24844265
- 2. Kerr NL. HARKing: hypothesizing after the results are known. Pers Soc Psychol Rev. 1998;2(3):196–217. pmid:15647155
- 3. Lakens D, Mesquida C, Rasti S, Ditroilo M. The benefits of preregistration and Registered Reports. Evid Based Toxicol. 2024;2(1).
- 4. Pariente N, Editors on behalf of the PBS. Premiering pre-registration at PLOS Biology. PLOS Biol. 2022;20: e3001611. doi:10.1371/journal.pbio.3001611
- 5. Nosek BA, Ebersole CR, DeHaven AC, Mellor DT. The preregistration revolution. Proc Natl Acad Sci U S A. 2018;115(11):2600–6. pmid:29531091
- 6. Pick JL, Takola E, Bairos-Novak KR, Ivimey-Cook ER, Morillo D, Nakagawa S, et al. Pre-registration and registered reports in ecology and evolutionary biology: an evidence-based appraisal by SORTEE. 2026 [cited 17 Sept 2026]. Available: https://ecoevorxiv.org/repository/view/12377/
- 7. Priboi C, Mayer B, Vergauwe E, Elger BS, Würbel H. A survey of researchers’ attitudes to preregistration in animal research reveals multiple perceived barriers to adoption. PLoS Biol. 2026;24(7):e3003511. pmid:42520002
- 8. Brazilian Reproducibility Initiative. Estimating the replicability of Brazilian biomedical science. eLife. 2026;15.
- 9. Neves K, Amaral OB. Addressing selective reporting of experiments through predefined exclusion criteria. Elife. 2020;9:e56626. pmid:32441650
- 10. Amaral OB, Valério B, Carneiro CFD, Mota GPS, Neves K, Abreu M. Challenges for building up confirmatory science in lab biology: Lessons learned from the Brazilian Reproducibility Initiative. MetaArXiv. 2026.
- 11. Fraser H, Parker T, Nakagawa S, Barnett A, Fidler F. Questionable research practices in ecology and evolution. PLoS One. 2018;13(7):e0200303. pmid:30011289
- 12. Mogil JS, Macleod MR. No publication without confirmation. Nature. 2017;542(7642):409–11. pmid:28230138
- 13. Dirnagl U. Preregistration of exploratory research: Learning from the golden age of discovery. PLoS Biol. 2020;18(3):e3000690. pmid:32214315
- 14. Sanes JR. Tell me a story. Elife. 2019;8:e50527. pmid:31383254