The present study was preregistered before the data on open science and new statistics practices were gathered. The preregistration was submitted on 26th April 2024 on the OSF (see Düvel & Altemeier, 2024). The study was conducted as preregistered, with minor deviations noted in the methods description.

To examine recent shifts in publication practices in the Yearbook of Music Psychology, the most recent volumes were analysed. Since open science and new statistical practices are primarily documented in quantitative studies, only “research reports” and “research reports on thematic focus” were included. Other formats (e.g., spots, close-ups, reviews) were excluded. All eligible reports—including quantitative, qualitative, and non-empirical articles—were coded.

As preregistered, volumes up to Volume 31 were initially included, but during coding, five online-first papers from Volume 32 were added to capture recent developments. Data collection was extended back to Volume 22 (2012) to cover key changes. A stopping rule was defined: If none of the three earliest coded volumes (Vols. 22–24) contained open science practices (preregistration, open materials, or open data), coding would stop. If any such practice was found, earlier volumes would be added iteratively until three consecutive volumes showed no such practices. As no such practices were found in Volumes 22 to 24, data collection ended there. In total, research reports from Volumes 22 to 32 (including all papers through June 2024) were coded, resulting in N = 79 papers (see Figure 2 for distribution).

First, all available bibliometric data on the Yearbook was reviewed to retrieve general information (e.g., title, authors) and streamline manual coding. Metadata for all articles from Volume 22 onward was sought. According to the journal's website (https://jbdgm.psychopen.eu), the Yearbook is indexed in several databases, but most (DOAJ, Dimensions, BASE, Scilit, EBSCO, Semantic Scholar) cover only articles from Volume 28 onward, following the shift to online-first, open access publishing. DNB and OOIR list only the journal, not individual articles. While Google Scholar and PsychArchives index all articles, they do not provide a practical way to download metadata. Only PubPsych enables comprehensive metadata retrieval for all volumes using the search term "Jahrbuch Musikpsychologie."

All 257 records were downloaded from PubPsych as a RIS file. Initial inspection showed unclean data, including duplicates, encoding errors, unclear variable names, and missing entries. The data was cleaned in OpenRefine (https://openrefine.org): empty columns were deleted, authors split into separate variables, duplicates removed, and volume numbers retrieved. This resulted in 179 unique entries (from Vols. 1–30), though some older papers were missing. Key variables (volume, year, title, authors, address, language, doi, abstract, keywords, database) were generally complete. Keywords, though untidy and not analysed in this study, were retained for potential future use. Finally, the dataset was imported into R to generate a unique key for each paper—combining volume number, the first three letters of the first author’s surname, and the first three letters of the title. The final dataset was then exported as a CSV file (“03_PubPsych-Datenexport_N=179_fromR.csv”; see online supplementary material).

Main variables on open science practices and statistical methods were manually coded by the two authors (ND and FA), following coding prompts similar to the studies by Giofrè et al. (2017, 2023). To avoid errors due to coding in a spreadsheet, the coding form was implemented as an online questionnaire using SoSci Survey (https://www.soscisurvey.de; see questionnaire in the online supplementary material). Metadata from Phase 1 pre-filled the questionnaire fields: after entering each paper’s key, the corresponding case from the CSV file was retrieved and variables pre-filled, so coders only needed to check the information instead of re-entering it.

After recording general article characteristics, the research design was coded. For non-quantitative designs (theoretical, qualitative), coding ended; for quantitative designs (experimental, quasi-experimental, observational), items on new statistics and open science followed. Most variables included an “I’m not sure” option and a comment field was available on each page to avoid coding mistakes resulting from uncertainties. The two authors coded all papers from the relevant sections. The papers from each volume were split between the two coders in alternating order, with three papers double coded to assess inter-coder reliability. Completed data was exported from SoSci Survey as a CSV file with an R import script.

In the final phase, the authors discussed all remaining uncertainties and comments, documented decisions, and updated the dataset accordingly using an R script. Duplicates from double-coding were removed, and all “I’m not sure” ratings were eliminated. The final dataset was exported from R as a CSV file (“04_Jahrbuch_dataset.csv”; see online supplementary material).

The dataset includes variables for coder and key (based on volume number, first author, and title). General article characteristics from PubPsych are: title, authors, year, language, DOI, volume, abstract, and institution of the corresponding or first author. Additional manually coded variables include: country of origin, PDF filename, replication status, study design, and comments. For new statistics, the presence of p-values, confidence intervals, effect sizes, sample size determination, and data exclusion was recorded. For open science, preregistration, open materials, open data, funding, and conflict of interest statements were coded. Criteria for coding new statistics and open science items were set liberally—following Giofrè et al. (2017, 2023)—to allow for direct comparison and to detect early changes in publication practices.

The Yearbook is open to submissions in German and English language. The proportions of research reports in German and English are displayed in Figure 3. The Yearbook started with only German publications in Volume 22 to 24.

From Volume 25 on, an increase of the proportion of English publications is visible. In the current Volume 32, most papers (4 of 5) are published in English. Although no specific hypothesis was formulated, we quantified this increase by calculating a linear regression: F(1, 9) = 14.7, p = .004, with R² = 62.0% of the variance explained by the model. The slope of the regression line of 5.6% indicates that, on average, every volume included 5.6% more English articles than the previous one.

As operationalised by Giofrè et al. (2017, 2023), the country of origin of a paper equals the country of the corresponding author’s institution. For the articles in Volumes 27 and older, no corresponding author was indicated. In these instances, the institution of the first author was used, wherever it was available (n = 5 missing cases). The countries of origin are displayed in Figure 4. The majority of papers stem from researchers at German institutions (89.2%). Sporadic publications stem from the German-speaking countries Austria (4.1%) and Switzerland (1.4%), or from the UK (5.4%, published by German researchers that are employed in the UK).

Disregarding the actual research design, it was coded whether the paper reports a replication study or not. The criterion was set very liberally: Not only strict replications were included but also, for example, partial replications. In the dataset, three studies were coded as replication studies (3.8%).

The research design was coded based on the information in the abstract, methods, and results sections of a paper. The different designs distribute across the volumes as displayed in Figure 5. Visual inspection reveals that the proportion of experimental studies (black-striped in Figure 5) increased over time. Analogously, the proportion of non-empirical articles (white-dotted; e.g., theoretical contributions or reviews of a topic) decreased. Only a few papers feature quasi-experimental, meta-analytical (here in the broader sense of a paper jointly analysing several quantitative datasets), or qualitative designs. The remaining proportion of observational studies (grey-chequered in Figure 5) remains similar over time at about 50% of the publications.

Summed across all volumes, 48.1% of research reports presented observational studies, 24.1% non-empirical studies, 19.0% experimental studies, 3.8% meta-analyses, 3.8% quasi-experimental studies and 1.2% qualitative studies.

Research Question 4 specifically aimed at the development of authorship over time. The number of authors per paper was averaged across each volume and is displayed in Figure 6.

The number of authors increased over time: For Volume 22, the median was 1 author, since Volume 24, it was always 2 authors. For the most recent papers from Volume 32, the upper quartile reached 4 authors per paper. A significant linear regression supported the visual impression: F(1, 9) = 6.68, p = .029, R² = 42.6%. On average, in each volume the articles had 6.4% more authors than the articles in the previous volume. All in all, Hypothesis 4.1 could be confirmed. Summed across all volumes, the papers had on average (median) of 2 authors (25% quartile = 1, 75% quartile = 3).

The preregistration included the research question regarding the identification of the most prominent authors. To avoid highlighting individual authors in the manner of a Hall of Fame, we decided against listing author names here (see “08_most prevalent authors.pdf” in the online supplementary material for a non-anonymised list). In summary, one author has published seven papers, one has published six, one has published five, two have published four each, and eight have published three each. All other authors published two or just one paper in the selected volumes. All in all, 123 distinct authors are represented in the dataset.

For all quantitative papers it was coded whether the authors reported that their investigation had been pre-registered. No paper reported a preregistration—this practice has not yet entered the German music psychology publication practices. Therefore, H1.1a has to be rejected.

Each quantitative research report was scanned and searched for statements regarding the availability of research materials, e.g., questionnaires, stimuli (audio examples), or analysis scripts. If only one material (e.g., only the questionnaire but not the stimuli) was available, this variable was already coded positively. The results are displayed in Figure 7. In most cases (30.5% overall), the materials could be downloaded directly, only a few cases (3.4%) stated that they are available upon request. Overall, 66.1% of the quantitative papers included no statement about the availability of research materials. In the most recent volumes, the majority of papers report availability of at least some research materials. A significant linear regression supported the visual impression of an increase in available materials (be it on request or downloadable), F(1, 9) = 68, p < .001, R² = 88.3%. On average, each volume had 9.4% more articles with some available materials than the previous volume. Overall, H1.1b was accepted.

As detailed in the introduction, data was understood here as the table with the characteristics of the subjects regarding variables. The analysis of the statements regarding the availability of the data is displayed in Figure 8.

Since Volume 28, almost all quantitative papers included statements regarding the data availability. Roughly half have made their dataset available for download (dark grey in Figure 8). In some cases, the data is available upon request (light grey). Only a few papers state that the data cannot be published due to data protection issues (white). Therefore, in recent years, most authors published their dataset alongside the paper or at least stated that it is available upon request.

Summed across all volumes, 55.9% of quantitative research reports included no statement regarding data availability, in 25.4% the data was available, in 15.2% it was stated that the data is available on request, and 3.3% stated that the data is not shared. The proportion of papers with data publications did not increase stepwise over time but rather suddenly between Volume 27 and 28. A significant linear regression supported the hypothesis of an increase in available data (be it on request or downloadable): F(1, 9) = 33.5, p < .001, R² = 78.8%. On average, each volume had 12.3% more articles with available research data than the previous volume. Therefore, H1.1c is accepted.

For reasons of transparency as well as acknowledgement, funders should be reported in the publications of the associated research projects. Until Volume 27, funders were rarely reported, and the statement was optional (see Figure 9). From Volume 28 on, a statement regarding funding was obligatory, although still most of the publications did not report any funding. A significant linear regression supported the hypothesis of an increase in funding statements: F(1, 9) = 24.9, p < .001, R² = 73.5%. On average, each volume had 12.9% more articles with funding statements than the previous volume. Therefore, funding statements did become more prevalent over time and H1.2a was accepted.

Summed across all volumes, 49.2% of quantitative research reports included no funding statement, 35.6% stated that there was no funding, and 15.3% named funders.

Similarly to the funding statement, a statement regarding potential conflicts of interest enhances transparency. This practice was not employed until Volume 27 and became obligatory from Volume 28 onwards. Only one quantitative research report (1.7%) stated potential conflicts of interest (see Figure 10). 52.5% included no statement regarding conflicts of interest, 45.8% stated that the authors reported no conflict of interest. A significant linear regression supported the hypothesis of an increase in conflicts of interest statements: F(1, 9) = 27, p < .001, R² = 75.0%. On average, each volume had 13.6% more articles with funding statements than the previous volume. All in all, conflicts of interest statements did become more prevalent over time and H1.2b was also accepted.

All in all, the analysis revealed changes over time regarding the observed open science practices and yielded answers to the leading RQ1. The publication of datasets (open data) and writing statements regarding funding and conflicts of interest have improved over time and employed thoroughly since Volume 28—H1.1c, H1.2a and H1.2b were accepted. Also, the proportion of published research materials (open materials) has increased over time (H1.1b was accepted). Only, the practice of preregistration has not yet been adopted for the German or German speaking music psychology publications (H1.1a was rejected).

The quantitative research reports were searched for p-values to gain insight into the prevalence of NHST. Most papers reported p-values and most of them included at least one exact p-value (79.7% of reports included at least one exact p-value while 15.2% included only relative p-values; 5.1% included no p-values; see Figure 11). Linear regression is non-significant, F(1, 9) = 2.04, p = .19, R² = 18.5, slope of the regression line is −1.6%, therefore no increase or decrease of papers including p-values was found. As assumed in H2.1, the use of NHST was quite popular among quantitative music psychology studies reported in the Yearbook and the hypothesis was accepted.

The papers were scanned for CIs that were reported in the text (mostly, the results section), in a figure (as error bars to central tendencies—the error bars had to be labelled as CIs), or in a table. If one was reported somewhere, the item was rated positively. In almost all considered volumes, roughly half of the quantitative studies reported at least one CI (see Figure 12). There is no obvious tendency over time: linear regression was non-significant, F(1, 9) = 4.2, p = .072, R² = 31.6%, slope of the regression line is 3.2%. Therefore, H2.2a was rejected.

Summed across all volumes, 55.9% of quantitative research reports included no CI while the other 44.1% included at least one.

The reporting of (standardised) effect sizes in the paper was observed. Aside from common effect sizes like the correlation coefficient r or Cohen’s d, in many cases, an enquiry was conducted to clarify if a measure is an effect size or not. In these cases, statistics books (Eid et al., 2015; Sedlmeier & Renkewitz, 2013) and the internet were consulted. The criterion was, again, set quite liberally: If the measure was described as fulfilling the function of an effect size or enabling interpretation similar to an effect size, the variable was coded positively. The measures were decided upon as noted in Table 1.

The prevalence of effect sizes over time is displayed in Figure 13. Across all volumes, 84.7% of the quantitative papers contained at least one effect size while 15.3% contained none. The linear regression was significant, F(1, 9) = 5.6, p = .042, R² = 38.4, the regression line having a slope of 2.7%. Therefore, the reporting of effect sizes increased over time—albeit with just 2.7% per volume—and H2.2b was accepted.

It was noted if a reason for the employed sample size was given and if so, on what rationale it was based. If the size of the sample was determined by a statistical procedure (e.g., power analysis) or based on previous studies, the item was coded positively. If other reasons were given (pragmatic reasons like the size of a natural group under investigation) or no reason was reported, the item was coded negatively.

All in all, a statement regarding the sample size determination could be rarely observed (13.6% while the other 86.4% included no such statement; see Figure 14). The first one was included in Volume 28. During the last two volumes, the proportion exceeded 50%. Taking into account that power analyses have been developed for centuries, German music psychology seems to be lagging behind the standard. Although the Yearbook itself published a paper on the topic by Platz et al. (2012), it did not seem to have had a major influence on researchers’ practices. The linear regression revealed a significant positive trend: F(1, 9) = 12.6, p = .006, R² = 58.2%, the regression line featuring a slope of 5.5%. Therefore, the prevalence of sample size determination has increased over the last years and H2.3a can be accepted.

This item was taken over from Giofrè et al. (2017, 2023) and was rated positively if at least one of the two criteria were fulfilled: (1) if the researchers adopted a stopping rule for data collection and reported what the stopping rule was or, (2) if they reported criteria to exclude data or participants.

Stopping rules could not be observed in the present dataset and do not seem to be used in music psychology very often. Therefore, Figure 15 displays whether the authors reported the exclusion of data or participants from the analysis. 44.1% of the quantitative publication included such a statement. The other 55.9% included no such statement. Linear regression revealed a significant positive tendency, F(1, 9) = 15.4, p = .004, R² = 63.0%, the regression line having a slope of 7.7%. Therefore, H2.3b was accepted.

The analyses revealed that NHST was popular in all regarded volumes—p-values were reported in most quantitative papers (H2.1 was confirmed). On the other hand, effect sizes were also wide-spread and CIs were at least reported in about half of the publications. On the one hand, the reporting of effect size was wide-spread and showed a slight significant increase, H2.2b was accepted. On the other hand, CIs were reported in 44% of the research reports and their use did not increase over time, therefore, H2.2a was rejected. A strict use of new statistics (in the sense that it relied on effect sizes and CIs and omitted p-values) was observed in one recent paper by Ruth et al. (2024). Statements regarding the determination of the sample size were still quite rare, reporting the data exclusion was somewhat more prevalent. Both showed an increase over time, H2.3a and b could be confirmed.

Anglada-Tort and Sanfilippo (2019) conducted a bibliometric analysis on three international music psychology journals. They reported the average number of authors per paper (see Table 6 in their paper, column “average TA [total authors] per document”) which will be compared to the results of the present study. Figure A7 displays the mean number of authors from the present study (black dots with error bars) and from Anglada-Tort and Sanfilippo (2019; triangles). Here, we calculated mean and standard errors for the JBDGM data instead of quartiles (as in Figure 6) to make the present data comparable to the results from Anglada-Tort and Sanfilippo. The years 2012–2017 were included in both studies and the data from Anglada-Tort and Sanfilippo (2019) showed a higher mean number of authors per paper than the present study, especially for the earlier years.