Abstract
Over the last decade, the adoption of preregistration has gained more and more popularity in psychology (Simmons et al., 2021). Psychologists employed at universities in Germany, Austria, Luxembourg, and Switzerland, for instance, published only 5% of the empirical articles with a link to a preregistration in 2020. Afterward, this rate annually increased to 12% in 2023 (see Schulz-Hardt, 2025). Preregistrations consist of a time-stamped document specifying crucial parts of the methods and data analysis before the data collection (e.g., Parsons et al., 2022). Consequently, preregistration should reduce researchers’ degrees of freedom and biases in the research process, increase transparency, and eventually lead to higher replicability and a more trustworthy science (e.g., Hardwicke & Wagenmakers, 2023; Spitzer & Mueller, 2023; but see Rubin & Donkin, 2024). But do preregistrations live up to the standards required to decrease researchers’ degrees of freedom and increase transparency?
Because preregistration implementation only recently gained popularity, research on preregistration quality and efficacy is also just emerging (Claesen et al., 2022; Heirene et al., 2024; Van den Akker, Bakker, van Assen, et al., 2024; Van den Akker, Van Assen, Bakker, et al., 2024; Van den Akker et al., 2023). The results of the few existing studies indicate that as they are currently used, preregistrations do not reduce researchers’ degrees of freedom in all cases to the desired degree (e.g., because hypotheses from preregistrations are not all mentioned when the preregistered studies are published). Even though an incomplete preregistration is better than no preregistration at all (Ofosu & Posner, 2023) because it reduces researchers’ degrees of freedom at least to some extent, studies on preregistration quality have repeatedly concluded that there are substantial deficits in the level of detail or completeness of preregistrations (Bakker et al., 2020; Heirene et al., 2024).
However, this conclusion might be premature because there is a disagreement about adequate criteria for preregistration completeness (e.g., Simonsohn, 2024). Therefore, in this article, we start with a discussion of different sets of these criteria applicable to a broad set of methods and designs for confirmatory research 1 —differentiating between those reducing researchers’ degrees of freedom (referred to hereafter as “procedural specification”) and those solely increasing transparency but not affecting researchers’ degrees of freedom.
Furthermore, previous research has relied on samples of publications that put an emphasis on preregistrations, such as preregistration prize winners or early recipients of preregistration badges (Van den Akker, Van Assen, Bakker, et al., 2024). Alternatively, it has focused on specific subdisciplines (e.g., gambling studies; Heirene et al., 2024) or journals (Claesen et al., 2022). Thus, the existing studies provide information about selective research areas, which might not provide a representative picture of the overall completeness of preregistrations. Therefore, in the current study, we analyzed all preregistrations from one publication year in a region. Thus, the current sample includes journal articles from a broader range of psychological fields.
Taken together, the current research has two major contributions: (a) a reflection on how to assess the completeness of preregistrations in metascience research (i.e., differentiating procedural specification and transparency) and (b) an empirical test of the completeness of preregistration in a broad sample and its relation to aspects that might influence preregistration completeness (i.e., Journal Impact Factor [JIF], Transparency And Openness Promotion [TOP] Factor). In line with the dominant form of research in psychology and (potentially resulting from that) a common understanding of what preregistrations are most useful for—but by far not the only possible one (see e.g., Ledgerwood, 2018)—we focus on confirmatory research.
What Should Preregistrations Contain?
According to Wicherts et al. (2016), a preregistration should be specific (i.e., include all steps in the research process: hypothesizing, research design, data collection, data analyses, and results reporting), precise (i.e., unambiguous), and exhaustive (i.e., cover all decisions that can be made in advance). In other words, they should include everything researchers should know before starting to collect data. Other scholars have advocated that “succinctness is key” (Simonsohn, 2024) and suggested to include only what is necessary to reduce researchers’ degrees of freedom. To do justice to both positions, we suggest two sets of criteria: (a) procedural specifications reducing researchers’ degrees of freedom and (b) transparency elements that contextualize the research and the rationale for decisions but do not reduce researchers’ degrees of freedom. Note that procedural specifications also increase transparency. The crucial difference is that transparency elements only increase transparency, whereas procedural specifications also reduce researchers’ degrees of freedom.
“Researchers’ degrees of freedom” refers to the flexibility a researcher has in the research process because of multiple equally justifiable decisions (Parsons et al., 2022). To reduce researchers’ degrees of freedom, preregistrations should contain procedural specifications that limit this flexibility. Wicherts et al. (2016) created a comprehensive list of 34 researchers’ degrees of freedom, each of which can be reduced by preregistering the respective decision that a researcher can make. Accordingly, this list has been used to check for completeness in preregistration research (e.g., Bakker et al., 2020; Heirene et al., 2024).
It is, however, important to note that Wicherts et al.’s (2016) list was put together with a focus on experimental designs. Thus, some of the researchers’ degrees of freedom in their list make sense only if the design of the respective study is indeed experimental (e.g., how participants are assigned to conditions, how blinding is implemented, manipulation checks). This point demonstrates that doing justice to a heterogeneous set of studies in research on the completeness of preregistrations presupposes exclusively relying on a set of decisions that have to be made irrespective of the specific research design and methodology. Ofosu and Posner (2023) suggested such a set of “four key requirements for a complete, well-specified” (p. 182) preregistration for political science and economics: a clear hypothesis, the primary dependent variable(s), the independent variable(s), and the statistical model.
If the goal of preregistrations is to reduce researchers’ degrees of freedom, a preregistration of confirmatory research should prevent (a) hypothesizing after results are known (HARKing; Kerr, 1998), (b) selective reporting of variables (i.e., conditions and measures), (c) optional stopping or sampling until the desired results are found, (d) selective exclusion of participants, and (e) choice of analysis procedure contingent to results. These five purposes are best addressed if a preregistration contains the (a) hypothesized pattern of results, (b) assessed and manipulated variables, (c) sample size, (d) exclusion criteria, and (e) planned analyses to test the hypotheses. The final procedural specification of a preregistration required to reduce researchers’ degrees of freedom is a time stamp. A time-stamped preregistration document prevents researchers from changing the content after collecting or analyzing the data (e.g., Parsons et al., 2022). In sum, we consider the following six elements as procedural specifications that are independent of design and methodology and decrease researchers’ degrees of freedom: hypothesized pattern of results, (assessed and manipulated) variables, sample size, exclusion criteria, planned analyses to test the hypotheses, and time stamp. Hence, we argue that the six elements should be included in all preregistrations for confirmatory research.
Comparing this list with the one suggested by Wicherts et al. (2016), we find that the procedural specifications on our list address all areas (i.e., hypothesizing, design, collection, analyses, and reporting) and 25 of the 34 specific degrees of freedom they mentioned (for an overview, see S1 in the Supplemental Material available online). This coverage is made possible by describing the specifications at a more abstract level. Hence, this approach allows for a balance between broad applicability (i.e., across methods and designs) and reduction of key researchers’ degrees of freedom.
Other aspects of preregistrations that are often discussed as parts of preregistrations are dependent on research design and method. For example, a termination rule for data collection can be irrelevant when the data are collected via a panel provider, and missing-data handling is irrelevant when the configuration of an online survey requires that all questions have to be answered (but could still be specified as part of the analysis procedure). Likewise, a missing description of how a variable is computed leads to researchers’ degrees of freedom only if the variable is not captured with a single item.
Increasing transparency is a second aim of preregistrations. “Transparency” refers to a researcher’s openness and honesty about theoretical, methodological, and analytical decisions (Parsons et al., 2022). Even though the procedural specifications also increase transparency, they mainly aim at reducing researchers’ degrees of freedom. In contrast, transparency elements increase the comprehensibility of the researchers’ decision and provide additional background information already at the stage of preregistration. In other words, transparency elements provide additional information by justifying researchers’ decisions or giving additional details at the stage of preregistration rather than in a later publication.
Based on the OSF Registries and Psychological Research Preregistration–Quantitative (Bosnjak et al., 2022) preregistration templates, we identified six elements falling into the category of transparency that apply across designs and methodologies for confirmatory research: (a) description of the research question, (b) theoretical justification of the hypothesis, (c) justification of the sample size, (d) justification of the exclusion criteria, (e) data-collection mode (e.g., lab, online), and (f) planned sample composition (e.g., student-subject pool, online-sample provider). An inclusion of the research question, theoretical justification of the hypothesis, justification of the sample size, and exclusion criteria into a preregistration may slightly constrain what can be reported in the theory and method sections of an article. However, given that the conclusions drawn from these justifications are not strictly considered as required procedural specifications, the degrees of freedom gained by not including the justifications is very limited. The main benefit is transparency. Finally, data-collection mode and sample composition have likewise very little or no relevance to researchers’ degrees of freedom but increase transparency by giving more information about the context in which the study took place.
Research on the Completeness of Preregistrations
Because the implementation of preregistration is a recent development, there is limited research on the completeness of preregistration (beyond medical science). Focusing on the four key elements (i.e., clear hypothesis, primary dependent variable[s], independent variable[s], and statistical model), Ofosu and Posner (2023) found that only little more than half of their 195 sampled preregistrations in political science and economics included information on all of their key elements. This underwhelming result might be because these preregistrations had been written between 2011 and 2016—before the discussion about preregistrations and their content took off.
Bakker et al. (2020) investigated the completeness of 105 preregistrations from 2016 and compared unstructured (i.e., no guiding questions) and structured (i.e., guiding questions) preregistration formats. In their study, they used the list of researchers’ degrees of freedom from Wicherts et al. (2016). They found that although structured preregistrations outperformed unstructured preregistrations, completeness generally appeared to be low. However, this finding might, as mentioned above, result in part from applying criteria for experimental research to other designs. In a reanalysis of their data, we found that the procedural specification we identified above did not differ between structured and unstructured formats (see S2 in the Supplemental Material).
Heirene et al. (2024) investigated a sample of 53 preregistrations for gambling studies from 2017 to 2020 and compared them with the cross-disciplinary sample of structured preregistrations from Bakker et al. (2020). They found that preregistrations for gambling studies were overall more complete than the preregistrations from the cross-disciplinary sample. Our reanalysis of these data based on the procedural specification confirms their conclusion except for two criteria, that is, description of the dependent and independent variables. Note that their comparison of gambling studies and the cross-disciplinary sample might be flawed because it is confounded with the year of preregistration.
The Present Investigation
Based on the summarized finding, the results of research studying preregistrations seem to be contingent on the criteria used (effect of format with Wicherts et al. [2016] criteria but not with our procedural-specification indicator), and the completeness has been found to be low. Moreover, the sample sizes of some studies on preregistration completeness in psychology are low, and the samples have been drawn from more homogeneous populations (e.g., gambling studies in the case of Heirene et al., 2024).
Therefore, we aimed to study preregistration completeness in a larger sample drawn from a broader sample regarding research domain and journals using the separate indicators for procedural specification and transparency introduced above. In addition, we sought to study journal reputation as a correlate of preregistration completeness.
Many analyses of preregistrations focus on one or a few journals with high reputation. To find out whether the results of these studies are biased, it is important to know whether journal reputation and preregistration quality are related. We propose that journal reputation should relate to preregistration completeness for at least two reasons. First, the editorial process varies between journals. Rejection rates and effort of reviewers and editors differ substantially between journals. Second, the TOP Guidelines provide an index for journals’ efforts to increase transparency, openness, and replicability (Nosek et al., 2015). First evidence for the prediction that preregistration quality differs between journals came from Heirene et al. (2024), who found a higher rate of undisclosed deviations from preregistrations among articles in
One widespread indicator of journal reputation that is used for review, promotion, and tenure in research is the JIF (McKiernan et al., 2019). If journals with a higher JIF handle articles more rigorously, this might also mean that articles need more complete preregistrations. Therefore, we hypothesized that the higher the JIF is, the more complete preregistrations will be (Hypothesis 1). 2 Beyond our preregistered hypothesis, we also explore the relation between preregistration completeness and the TOP Factor. Whereas the JIF has previously been criticized as a measure of journal reputation (e.g., Schönbrodt et al., 2022), the TOP Factor is an alternative metric that indicates the alignment of journal policies with standards of transparency and openness (Nosek et al., 2015) and thus potentially captures journal quality better in relation to preregistrations. Hence, especially the TOP subscore for preregistration of study and the TOP subscore for preregistrations of analysis plan should correlate highly with preregistration completeness.
Disclosures
Preregistration
The study was preregistered (https://aspredicted.org/qwc8-wh9x.pdf) after revising the coding scheme based on the experiences from coding 16 preregistrations. Substantial deviations from the preregistration are explicitly acknowledged in the article and justified in more detail in S5 in the Supplemental Material. The coding scheme contains a wide variety of indicators. We originally preregistered one broader completeness index including more fine-grained coding of the researchers’ degrees of freedom mentioned by Wicherts et al. (2016) and preregistration guidelines (for analysis with the original preregistered completeness index, see S6 in the Supplemental Material). The original completeness index turned out to be problematic given that it contained aspects that are not independent of design and method. For example, a method of correction for multiple tests is relevant only if multiple tests are planned. Based on this insight, we started the reflection above and formed the new indices for procedural specifications and the transparency elements. Thus, the considerations leading to these indices are partly informed by the coding process of the study reported below and a deeper reflection on what applies to all preregistrations for confirmatory research independent of design and methodology. Following the suggestion of a reviewer, we wrote a new preregistration based on what we learned conducting the study and that can be used for future studies (see S7 in the Supplemental Material).
Data and materials
The data drawn from PSYNDEX, a reference database for psychological literature; which articles were excluded from our sample, the final data, and the analysis script are openly available (data: https://doi.org/10.23668/psycharchives.16435; analysis script: https://doi.org/10.23668/psycharchives.16436).
Reporting
We report how we determined our sample size, all data exclusions, all manipulations, and all variables in the study. The coding procedure is reported following the guidelines by Conry-Murray et al. (2024).
Ethical approval
The study was approved by the local ethics committee of Institut für Wissensmedien (LEK 2022/021).
Preprint
The article is published as a preprint (https://doi.org/10.31219/osf.io/wc7qr), and the results were presented at the 53rd German Psychological Society (DGPs) Congress and the META-REP 2024 conference.
Method
Sample
To obtain a sample of preregistrations from diverse journals, we used the PSYNDEX (https://psyndex.de/en/) database, which encompasses all psychological publications from German-speaking countries (i.e., Germany, Austria, Switzerland, and Luxembourg; for more information on the PSYNDEX database, see S8 in the Supplemental Material; for information about PSYNDEX’s inclusion criteria, see Leibniz Institute for Psychology, 2024). The original search was performed in June 2022 and was updated in December 2024 (search term: (((DB=PSYNDEX)) AND PY=“2020”) AND DT=“Journal Article”)). Selecting only journal articles published in 2020 led to 9,146 hits. Besides other information, the metadata in PSYNDEX for 2020 include information on whether a journal article was accompanied by a preregistration. This initial screening led to 464 journal articles with preregistrations. Hence, only 5.09% of journal articles were preregistered (including clinical-trial registration). Deviating from our preregistration, we excluded specific types of research that were too rare to be analyzed separately. To be more precise, we excluded seven Registered Reports because at Stage 1, they undergo peer review, which should warrant a substantially higher completeness of the content compared with regular preregistrations. Likewise, we excluded two Many Labs studies because more than one research team reviewed the preregistration. In addition, we excluded 11 studies with preregistrations for nonconfirmatory research because they are irrelevant to the current research question (i.e., the completeness of preregistrations in confirmatory research). Finally, we had to exclude 13 articles because the preregistration link in PSYNDEX was incorrect and we were unable to find the correct link within searching for 5 min. After applying the preregistered and described deviating exclusion criteria, a sample of 146 journal articles remained (see Fig. 1; for a list of all included journal articles, see Appendix A). The 146 articles reported 223 preregistrations. For the final sample, we randomly drew one preregistration from articles with more than one preregistration to avoid dependencies in the data.

Flowchart of sample-selection procedure.
Measures
All authors developed and tested the coding scheme (see S9 in the Supplemental Material). We coded most criteria dichotomously (not included, included). Four undergraduate research assistants were trained in an online meeting by explaining the coding scheme and evaluating a preregistration together. All 223 preregistrations were coded at least once, and a subsample of 114 preregistrations was drawn and coded by two raters to assess interrater agreement. Variables below the preregistered rater agreement were excluded. Preregistrations were randomly drawn and distributed to each coder. Each coder worked on between 52 and 208 preregistrations. Disagreement between coders for variables above the preregistered agreement criteria was resolved by a third rater. All coders were independent of each other and blind to the research question and hypothesis.
Deviating from the preregistration, we did not use κ as a measure of coder agreement because the marginal totals were symmetrically unbalanced, and in these cases, κ is known to be overly conservative (Delgado & Tibau, 2019; Feinstein & Cicchetti, 1990). Instead, we used the percentage of agreement as a measure of interrater agreement and excluded variables below 70%. For an overview of the variables used for the analysis and intercoder agreement, see Table 1. For an overview of all coded variables, see S10 in the Supplemental Material.
Intercoder Agreement and Frequencies of the Completeness Variables
Note: All variables are included in the completeness index; the variables that comprise the procedural-specification elements are in bold.
Beyond the coding, we included the following additional variables from external sources: JIF from the Web of Science 2022 and the first version of the TOP Factors from January 6, 2020 (Mellor et al., 2025). The JIF ranged from 0.58 to 29.9 (
The completeness index was planned as an unweighted sum of 21 variables. In line with the preregistration, three variables were not included in the index because of low intercoder agreement. Deviating from our preregistration, we also excluded six variables that were contingent on specific methods and designs (e.g., manipulation checks make sense only when manipulations are implemented). The resulting index included the following 12 variables: hypothesized pattern of results, assessed variables mentioned, planned sample size, exclusion criteria mentioned, planned analyses to test the hypotheses, time stamp, research question, theoretical justification of hypothesis, justification for sample size, justification for exclusion criteria, data-collection mode, and sample composition. It could theoretically range from 0 to 12. However, in our data, it ranged from 5 to 12 (
Based on the discussion about the quality criteria for preregistrations above, we also computed an index of procedural-specification-related and transparency-related preregistration content, that is, the unweighted sum of the variables hypothesized pattern of results, assessed variables mentioned, planned sample size, exclusion criteria mentioned, planned analysis to test the hypotheses, and time stamp for procedural specification and the unweighted sum of the variables research question, theoretical justification of hypothesis, justification for sample size, justification for exclusion criteria, data-collection mode, and sample composition for transparency. The frequencies for the variables included in these indices are also listed in Table 1. The index of procedural specification ranged from 3 to 6 (
Results
Preliminary analysis
Eleven (7.53%) preregistrations reached a high score of 12 in our completeness index, and 78 (53.42%) preregistrations included all six procedural specifications. Fifteen (10.27%) preregistrations included all six transparency elements (see Table 2). Hence, it seems that most—although not all—preregistrations include procedural-specification elements but that transparency elements were less frequently included—although to a different varying extent across criteria (see Table 1).
Absolute, Relative, and Cumulative Frequency for the Completeness, Procedural-Specification, and Transparency Indices
Note: Differences between percentage and cumulative percentage are because of rounding.
Confirmatory analysis
We preregistered a sensitivity analysis because our sample size was determined by the number of published articles with preregistrations in PSYNDEX rather than fixed a priori. The sensitivity analysis using G*Power 3.1 (Faul et al., 2007) with 138 cases (because of missing values for preregistered studies published in journals without JIF), 1 – β = .80, and α = .05 for a two-tailed correlation revealed that we could find an expected population correlation of at least |ρ| = .23.
We predicted a positive relationship between JIF and the preregistration completeness. However, this correlation was not significant,
Exploratory analysis
As preregistered, we corrected the α level for exploratory analysis (.05 /
As indicated above, we explored the relationship between the TOP Factor and preregistration completeness. The correlation between the completeness index and neither the TOP Factor,
In a final step, we considered the procedural-specification-related and transparency-related elements of the preregistrations (see Table 3). The correlations for procedural specification and the TOP Factor, the preregistration-of-study TOP subscore, and the preregistration-of-analysis-plan TOP subscore were negative (range:
Correlations (
Note: The correlations with the TOP Factor and the preregistration TOP subscores include the three outliers (for analysis without outliers, see S17 in the Supplemental Material available online). TOP = Transparency and Openness Promotion Factor; JIF = Journal Impact Factor.

Scattergrams and distributions for the completeness indices and JIF, TOP Factor, preregistration-of-study TOP subscore, and preregistration-of-analysis-plan TOP subscore. Circle size indicates frequency; bigger circles indicate higher frequency. TOP = Transparency and Openness Promotion Factor; JIF = Journal Impact Factor.
In addition, we performed Bayesian correlation analysis (see Table 4) with the null hypothesis that the variables are not correlated (H0: ρ = 0) and the alternative hypothesis that the correlation is higher than .23 (H1: ρ > .23). We used an uninformative prior distribution over the (−1, 1) interval for ρ. Our data indicated moderate to strong evidence that correlations are zero rather than
Bayesian Analyses of Correlations for the Journal-Reputation and the Completeness Indices.
Note: Bayesian correlation with H0 ρ = 0, H1 ρ > .23, and uninform prior distribution from (−1, 1). BF10 is the strength of evidence in favor of H1 relative to H0, and BF01 is the strength of evidence in favor of H0 relative to H1. BF = Bayes’s factor; TOP = Transparency and Openness Promotion Factor; JIF = Journal Impact Factor.
Discussion
In this study, we investigated the completeness of preregistrations of articles published in a broad range of journals and the relation of different preregistration-completeness indices to the JIF and the TOP Factor. More than half of the preregistrations contained all six procedural specifications we identified that are independent of design and methodology and reduce researchers’ degrees of freedom in confirmatory research. However, this finding also means that similar to previous studies (Ofosu & Posner, 2023), almost half of the preregistrations do not include all six of these key criteria, which are required to reduce researchers’ degrees of freedom. Even fewer preregistrations contained all transparency information, indicating that information that increases the comprehensibility of research already at the stage of preregistration is often not included. These findings are surprising because coding the variables dichotomously (i.e., included vs. not included) sets a rather low threshold.
Furthermore, our results suggest that the completeness of preregistrations is independent of the standards set by the journal in which the studies are published. This result occurred when we used the often criticized JIF (Schönbrodt et al., 2022) and the TOP Factor dedicated to transparency and openness promotion (Nosek et al., 2015) and its specific subscores for preregistration as indicators for journal reputation. Even when we differentiate between procedural-specification-related and transparency-related elements of preregistrations, we do not find a positive relation between preregistration completeness and journal reputation. Thus, our hypothesis was not supported by the data. In sum, this finding suggests that preregistration completeness is unrelated to journal standards regarding open science and implies that a check on whether a preregistration contains sufficient detail seems to be missing.
Strength and limitations
The strengths of the current study are (a) the consideration of a sample of all publications with preregistrations of a whole publication year from one region; (b) the differentiated approach to preregistration research, for the first time differentiating between procedural-specification-related and transparency-related content; and (c) the consideration of several indicators for journal reputation.
Even though our sampling method led to a sample of preregistrations of studies published in journals with diverse reputations, sampling all preregistrations from German-speaking countries that are included in PSYNDEX leads to restricted regional diversity. Hence, the generalizability to other areas of the world should be tested in future research. Second, even though our sample size was larger than in other similar studies, we could detect only
Moreover, in the current article, we focused on preregistrations of confirmatory research—in line with the finding that fewer than 10% of the preregistrations in our sample were exploratory. Nonetheless, exploratory research should also be preregistered to prevent, among other things, HARKing (i.e., that the results are published as confirmatory). Follow-up research should assess preregistrations of exploratory research.
Finally, a limitation might be the selected variables for the completeness index and that we assessed them dichotomously (included vs. not included). The dichotomous coding does not allow us to evaluate the precision of the preregistration (Wicherts et al., 2016). That is, we often coded whether a variable was mentioned. Future research should apply a more fine-grained spectrum, from mentioning an aspect in preregistration to describing it in enough detail to actually reduce researchers’ degrees of freedom completely. However, we note, based on our experience in the current study, that such a coding would require coders with some research or even domain experience.
Lessons learned
First, beyond the quantitative analysis, the current research again indicated that researchers should make their preregistration unambiguously accessible (Claesen et al., 2022). In the current sample, 13 articles (about 6.44%) were excluded because the preregistration link did not work and the preregistration was not found within 5 min of searching. In addition, to avoid ambiguity as to which studies in multistudy articles refer to which preregistrations, the preregistration documents should be named unambiguously. Beyond these pragmatic questions, there is currently a discussion regarding other formal conditions for (pre)registrations, such as the type of document and repository required to properly (pre)register a study (e.g., Mayo-Wilson et al., 2024). These important questions are beyond the scope of the current article but clearly deserve attention.
Second, it is of utmost importance that there is an unambiguous time stamp because the document being time-stamped is a key part of the definition of “preregistration” (e.g., Parsons et al., 2022)—again pointing to the question where and how a preregistration should be stored. We encountered most problems regarding time stamps for preregistration documents published on OSF. These documents often had a time stamp for when they were uploaded to OSF and a time stamp that was supposedly from the date of creation or the date of the last modification. To avoid such ambiguities, preregistrations should be created, modified, and ultimately published on the same platform, and it should be confirmed that they are unambiguously time-stamped.
Third, there should be set of criteria required to consider a study as preregistered (Thibault et al., 2023)—the six elements we marked as procedural specifications may be a good starting point: hypothesized pattern of results, variables, sample size, exclusion criteria, planned analyses to test the hypothesis, and time stamp. Including more information (e.g., a preregistration including contingency plans for assumption violations) can further reduce researchers’ degrees of freedom and increase transparency. However, we propose that this minimum set is required for each and every confirmatory study to label it “preregistered.” We additionally note that a preregistration along these lines does not undoubtedly warrant higher replicability. There has been a debate about this point in recent years (Rubin & Donkin, 2024; Szollosi et al., 2020), but empirical evidence is scarce.
Finally, still close to half of the preregistrations completely left out at least one of the procedural specifications. This result indicates that although using preregistration has been successfully established as the norm at least in some subfields of psychology (Glöckner et al., 2024), the completeness of preregistrations does not seem to be checked sufficiently. In other words, many instances of the preregistrations we coded were not living up to minimum requirements. Authors and journals do not seem to spend enough attention on preregistrations given that neither high-impact journals nor journals with a high TOP Factor stood out in terms of preregistration completeness. Hence, the self-disclosure of the journals to transparency and openness, such as the TOP Factor, does not automatically lead to the implementation of these changes. However, the six procedural-specification elements could be easily checked in the review process. Hence, our results highlight a need to reconsider journals’ editorial procedures. As iterated previously (Lindsay, 2023), we also do not believe this check should be an additional task of peer reviewers. Checking for these criteria could be done by trained editorial assistants, or authors could be asked to fill in a checklist regarding the content of their preregistration (for an example of a transparency checklist, see Aczel et al., 2021). That journal procedures need to change and the implementation of preregistrations needs to be reinforced are also reflected by the fact that only about 5.07% of journal articles had preregistrations. Even though the proportion is steadily rising (6.54% in 2021, 8.69% in 2022, and 12.05% in 2023 in the database we used for the current study; Schulz-Hardt, 2025), the absolute number is still underwhelming.
Conclusions
In the current study, we advance research on preregistrations by investigating their completeness from articles published in journals with diverse reputations and differentiating procedural specifications that should be included in all preregistrations to reduce researchers’ degrees of freedom and provide information that increases transparency. Close to half of the preregistrations do not include all six procedural specifications, and the six transparency elements were included in even fewer preregistrations. In addition, we found that the completeness of preregistrations is independent of a journal’s reputation.
Based on our results, there is a need for more consistent use of preregistrations in line with emerging norms, and the completeness of preregistrations has to be improved and more thoroughly checked in the review process. Furthermore, future research needs to evaluate if the steps taken to reduce researchers’ degrees of freedom actually increase replicability.
Supplemental Material
sj-pdf-1-amp-10.1177_25152459251357568 – Supplemental material for A Cross-Sectional Study of the Completeness of Preregistrations by Psychological Authors From German-Speaking Institutions
Supplemental material, sj-pdf-1-amp-10.1177_25152459251357568 for A Cross-Sectional Study of the Completeness of Preregistrations by Psychological Authors From German-Speaking Institutions by Lena Hahn, Andreas Glöckner, Mario Gollwitzer, Jens Hellmann, Jens Lange, Simon Schindler and Kai Sassenberg in Advances in Methods and Practices in Psychological Science
Footnotes
Transparency
ORCID iDs
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
