Within most scientific disciplines, it is traditional to wait to publish a study until the results provide a strong answer to a research question. However, there is active debate as to whether this is the ideal workflow. Under this model, an estimated 40% of data go unpublished [3], with a strong skew against publishing negative results [4]; a source of bias in the scientific literature referred to as “The File Drawer Problem.” Emphasizing a publishable “story” may also encourage questionable research practices, such as the flexible “dredging” of data to achieve statistically significant results (P-hacking), adapting hypotheses to match results (HARK-ing), and the selective reporting of findings (cherry picking) [5].

In response, there has been a growing movement for “modular publishing,” or sharing smaller component parts of a study as they are produced [6]. Within this model, components can be publicly released, either in specialized repositories as non-peer-reviewed interim products, or published separately in dedicated peer-reviewed journals (Table 1). There are also formats that encourage such staging, allowing components to be formally linked or released together, sometimes with opportunity for postpublication peer review and revision (e.g., Open Science Framework (OSF), Octopus.ac, ResearchEquals, and Research Ideas and Outcomes). Individual complete experiments can also be documented in short formats instead of being grouped together to build a larger “story.” In general, publication in specialized repositories is free and rapid (on the scale of days), whereas traditional journal publication is often accompanied by article processing charges and is much slower due to the editorial and peer review process, which may involve reviewer-requested experiments, or rejection and resubmission.

Online tools are available to help match publishing solutions to the needs of the researcher. The online tool Fiddle can match scientists with peer-reviewed and non-peer-reviewed options, taking into account individual financing and indexing constraints. The online tools Jane and Journal Finder recommend journals using the abstract text and/or title, and Transpose provides an easy way to search for journals with particular policies that can expedite publishing [7], including peer review transfers between journals (although this database is no longer actively maintained). ASAPbio also hosts a searchable database of preprint servers, and re3data, FAIRsharing, and DataCite have extensive guidance regarding field-specific data repositories.

There are reasons not to publicly release partially finished projects. Some research may be incompatible with these formats owing to ethical constraints on the public release of human subjects data or sensitive animal data. There may also be concerns that publication of research components may hinder later attempts at traditional publication. To this end, most major journals have begun to welcome, or even encourage, the separate release of study components [8], as this relieves the final publication of the burden of providing the full study documentation necessary for research reproducibility. The vast majority of journals also welcome manuscripts that have been released as preprints [7], with some even accepting direct transfer of preprints (and sometimes their reviews) to journal submission systems. In the meantime, these modular research outputs can be reported on CVs and grant applications to illustrate productivity, skills, and dedication to open science practices.

Another concern is that “unfinished” projects may lack replication or validation work. The publication of such data without acknowledging its exploratory nature may lead to wasted future effort. However, if limitations are properly acknowledged, contributing exploratory work to the scientific record can benefit future meta-analyses and encourage replication attempts.

Although the options described in Table 1 for modular publishing or preprinting are faster than traditional scientific publications, preparing these products still requires effort and time. All formats require gathering coauthor approval and conflict of interest disclosures, which can take several weeks, and peer-reviewed products will still require both editorial and peer review, and often revision and re-review (for individual journal time estimates, see SciRev.org). Furthermore, the products should be carefully prepared and cleanly formatted, as they will be joining the permanent public record and their utility will depend on the quality of their description.

To this end, scientists who still have access to resources may consider offloading some work onto a professional science writer, paper reformatting service (e.g., Editage) or scientific editing service (e.g., those offered by SpringerNature or Elsevier). If so, it is important to select a writer or service with suitable training and qualifications, and to properly define, oversee, and acknowledge their involvement, as ghost authorship is ethically prohibited in most contexts. Moreover, substantial contributions to interpretation may require consideration for full authorship [9].

Artificial intelligence (AI) tools can also help with browsing literature, summarizing rough notes, digitizing protocols, organizing code or data for release, improving data visualizations, improving grammar, clarity, tone, or conciseness of text, and helping reformat a paper to fit journal requirements [10]. Researchers need to proceed with AI tools carefully. Some tools will store and reuse inputted data, which is incompatible with sensitive data (e.g., clinical data). It is also important to double-check AI output for inaccuracy, inappropriate use or misinterpretation of source material and hallucinations [11]. Moreover, the paper must still center on the authors’ own work and ideas, and all AI usage must be carefully documented and transparently disclosed, following journal requirements [9].

Many of the publishing outlets in Table 1 are free, but some charge fees. Ideally, costs such as article processing charges would be charged to grants, as it is in the funders’ best interest for all research products to be publicly shared. If funding is unavailable, it is worth contacting your institution’s library or grants office, if available, to determine if any journals or publishers have an open access agreement or fee waiver for your institution or research society (e.g., the Big 10 Academic Alliance). Many journals also offer fee waivers or reduced charges for researchers from low- and middle-income countries (e.g., Research4Life), or in exchange for reviewing papers.

What about the cost of preparing the publications and research products? At this point, there are few sources of support for scientists interested in cleaning out their file drawers. The American Association of University Women offers grants to finish dissertation writing or prepare manuscripts for publication, and publication grants are occasionally available at the institution level. Offering grants to defunded or transitioning researchers for rapid research product release and manuscript preparation would be a high-impact investment for institutions and research foundations, yielding high returns for minimal upfront costs.

Due to a variety of circumstances, including sudden policy shifts and common personal challenges, many science projects will never be finished, despite years of invested resources and effort. By carefully and strategically documenting scientific work achieved, components of unfinished projects can be salvaged and preserved to benefit future researchers. Moreover, by thoughtfully considering options for modular publishing before they are needed, scientists can be better prepared for unexpected changes to their trajectory and careers.