Preregistration is the practice of writing down your research questions, hypotheses, and analysis plan before you collect or examine data. By committing to a plan in advance, researchers create a clear record of what they set out to test, making it much harder to blur the line between a genuine prediction and a post-hoc explanation dressed up as one. The idea is simple, but its implications for the credibility of psychological research are significant.
Why Psychology Needed Preregistration
Starting around 2011, psychology entered what’s now called the replication crisis. Decades of loose research practices had quietly eroded the field’s foundations, and when a large-scale effort by the Open Science Collaboration tried to reproduce 100 published psychology findings in 2015, only about 36% replicated successfully. That number was a wake-up call. Many of those original findings had likely been shaped, whether intentionally or not, by flexible analysis choices that inflated results.
Two practices drew particular scrutiny. The first, known as p-hacking, involves running many different analyses on the same data and selectively reporting the ones that produce statistically significant results. The second, called HARKing (hypothesizing after results are known), is when researchers look at their data, find a pattern, and then write up the paper as though they predicted that pattern all along. Both practices make findings look stronger and more intentional than they really are. Preregistration was proposed as a direct countermeasure to both.
How Preregistration Works
A preregistration is a timestamped document, stored on a public registry, that locks in the key decisions of a study before the results are known. At minimum, it typically includes:
- Research questions and hypotheses: what specifically the study is trying to test
- Study design: the experimental setup, conditions, and measures
- Sample size justification: how many participants will be recruited and why that number is sufficient
- Exclusion criteria: what rules will be used to remove participants or data points, defined before anyone looks at the results
- Analysis plan: the exact statistical tests that will be run to evaluate each hypothesis
The document is uploaded to a registry that provides a permanent timestamp, so anyone can later verify that these decisions were made before data collection began. Researchers can still run additional analyses that weren’t in the plan, but they must label those as exploratory rather than presenting them as planned tests. This distinction is the heart of what preregistration does: it separates hypothesis testing (did my prediction hold up?) from hypothesis generating (I noticed something interesting in the data).
Where Researchers Preregister
The most widely used general-purpose platform is the Open Science Framework (OSF), run by the Center for Open Science. OSF provides a timestamped, persistent registry that meets the core requirements for a credible preregistration. Another popular option is AsPredicted, which offers a simpler, shorter template, though evaluations have noted it doesn’t fully meet all the criteria that make a preregistration maximally trustworthy, such as guaranteed long-term persistence and full public accessibility.
For clinical trials specifically, ClinicalTrials.gov serves a similar function, and PROSPERO handles systematic reviews. But for the typical psychology experiment, OSF is the standard. Researchers fill out a structured template, submit it, and receive a time-stamped URL they can later include in their published paper so reviewers and readers can check what was planned versus what was discovered along the way.
Preregistration vs. Registered Reports
A standard preregistration is something a researcher does on their own. They upload their plan to a registry, run the study, and submit the finished paper to a journal the usual way. The preregistration exists as a reference point, but it doesn’t change how the paper gets reviewed or whether it gets published.
Registered Reports take the concept further by building it into the publication process itself. In this format, researchers submit their introduction, hypotheses, and methods to a journal before collecting data. The journal peer-reviews this “Stage 1” submission, and if the question is important and the methods are sound, the paper receives “in-principle acceptance.” This means the results will be published regardless of whether they confirm the hypothesis, as long as the researcher follows the approved protocol. The full manuscript with results is then reviewed again (Stage 2) primarily to check that the plan was followed.
This is a meaningful difference. Standard preregistration still leaves researchers vulnerable to journals that prefer exciting, positive results. Registered Reports remove that pressure entirely. As of a 2021 review, 137 psychology journals had adopted the Registered Report format, though most of those journals (about 72%) had yet to actually publish one, suggesting adoption is still in early stages.
Does Preregistration Actually Work?
The evidence is more complicated than proponents hoped. Early analyses of preregistered studies found an increase in reports of null findings (results that didn’t support the hypothesis), which suggested the practice was reducing the bias toward only publishing positive results. That was encouraging.
However, a large comparison of preregistered and non-preregistered psychology studies found surprisingly little difference between the two groups. Preregistered studies reported positive results at essentially the same rate (69%) as non-preregistered studies (68%). Effect sizes were numerically smaller in preregistered work (averaging 0.29 versus 0.36), but the difference wasn’t statistically significant. The researchers concluded their study did not provide convincing evidence that preregistration prevents p-hacking and HARKing in the main results of published papers.
This doesn’t necessarily mean preregistration is useless. It may mean that the researchers who choose to preregister are already more careful, or that journals still exert selection pressure toward positive results regardless of preregistration status. Some institutions have experimented with rewarding preregistration through badges on published papers, but at least one major journal, Psychological Science, discontinued its preregistration badges after evidence suggested they weren’t working as intended.
Common Criticisms
One frequent concern is that preregistration could discourage exploratory analysis, the kind of open-ended data investigation that often leads to unexpected discoveries. In practice, preregistration doesn’t forbid exploration. It simply requires researchers to be transparent about which findings were predicted and which emerged from the data. The worry is more about culture than rules: if the field starts treating only preregistered, confirmatory findings as “real science,” genuinely valuable exploratory work could be devalued.
Another criticism is practical. Writing a detailed analysis plan requires anticipating every decision you’ll need to make during analysis, which is difficult even for experienced researchers. Studies that use preexisting datasets face a unique challenge, since the sample size and many design features are already fixed. Specialized preregistration templates exist for these situations, but they highlight how one-size-fits-all templates can miss important nuances.
There’s also the concern that preregistration can become performative. A researcher could write a vague preregistration that doesn’t actually constrain their choices, or they could preregister after peeking at the data. Without careful enforcement, the practice risks becoming a box-checking exercise that provides a false sense of rigor rather than genuine transparency.
What Preregistration Means for You as a Reader
If you’re reading about a psychology study and see that it was preregistered, it means the researchers publicly committed to their methods and hypotheses before seeing the results. That’s a meaningful signal of transparency. If the study is a Registered Report, it carries even more weight, because a journal agreed the question and methods were worthwhile before knowing the outcome.
But preregistration isn’t a stamp of truth. A preregistered study can still have flawed methods, an unrepresentative sample, or conclusions that overreach. What preregistration does is remove one specific source of doubt: you can be more confident the researchers aren’t just telling a convenient story that fits the data after the fact. For a field still rebuilding trust after the replication crisis, that’s a real, if incomplete, step forward.