This dashboard tracks how data-visualisation practices in scientific publishing have evolved over time, focusing on the use of bar charts versus more informative alternatives. Visualisation types are classified by barzooka, an automated deep-learning tool that screens PDF figures. barzooka identifies the following chart types used for continuous data:
- Bar charts — the conventional mean-and-error bar format, which can mask distributional features such as bimodality, skewness, and outliers
- Informative charts — bars with dots, box plots, dot plots, histograms, or violin plots; all of these reveal the shape, spread, and individual data points that bar charts conceal
Each screened paper is categorised as using only bar charts, only informative charts, or both types within the same article.
Data are organised across 12 Research Fields:
- Cardiac & Cardiovascular Systems
- Clinical Neurology
- Endocrinology & Metabolism
- Genetics & Heredity
- Immunology
- Neurosciences
- Oncology
- Orthopedics
- Pharmacology & Pharmacy
- Physiology
- Rheumatology
- Urology & Nephrology
Bar charts that reduce continuous data to a mean and error bar are widely criticised for concealing distributional features — bimodality, skewness, outliers — that are critical for interpreting and replicating results. Despite two decades of calls to replace them, bar charts remain the dominant format across many biomedical and life-science journals.
In response, a growing number of journals have introduced editorial policies that explicitly encourage or require authors to use more informative alternatives (e.g., box plots, violin plots, dot plots, or bars overlaid with individual data points). These policy adoptions create natural quasi-experiments: by comparing visualisation practices before and after a journal implements its policy — and against journals that never adopted one — it is possible to estimate whether editorial recommendations produce measurable changes in author behaviour.
This causal question has direct implications for how journals, societies, and funders design interventions to improve data transparency. Answering it at scale requires longitudinal tracking of thousands of articles across many journals, year by year. This dashboard provides exactly that view, stratifying all charts by Policy and Non-Policy journals so that the effect of editorial interventions can be visually assessed.
The pre-registered study protocol is publicly available at osf.io/tcyxg.
% only bar
Percentage of eligible papers in that year using this visualisation type for continuous data, as detected by the Barzooka screening tool.
% bar and informative
Percentage of eligible papers in that year using this visualisation type for continuous data, as detected by the Barzooka screening tool.
% only informative
Percentage of eligible papers in that year using this visualisation type for continuous data, as detected by the Barzooka screening tool.
Policy Year Marker
A vertical line marking the year a journal adopted an editorial recommendation on figure types. On per-journal plots it is a single solid line; on aggregated plots, semi-transparent yellow bands whose width reflects the proportion of journals that adopted a policy in that year.
Use the sidebar to switch between the GLOBAL overview, individual Research Fields, or this About page. Each field (and the GLOBAL tab) shows three aggregated charts — All journals, Policy journals, and Non-Policy journals — in a resizable grid. The Show charts dropdown in the top bar filters which aggregated card types are visible in the active tab. The Show journals selector in each field tab further filters individual journal cards. The Columns slider resizes the grids. Use Find journal to locate a specific card, and the ⠿ grip to reorder cards. The global toggles in the top bar show/hide any metric across every chart at once.