Contemporary science journalism is determined by the way resources, people and projects are combined to transform complex research into credible public information. In my portfolio, every task begins with a clear brief, a set of named sources and a documented trail of data, because viewers in the United Kingdom expect both correctness and openness. The following sections explain how I choose stories, coordinate collaborations, utilize specialist tools, handle ethical questions and keep records that endure editorial scrutiny. Audiences will discover concrete details on processes, typical costs and the real-world choices that shape each published piece.
Understanding Open Access and Data-Sharing Guidelines
Since the publication of the UK government’s 2021 open-access framework, most publicly funded research papers must be available under a Creative Commons licence within twelve months. I manage a bookmark folder of institutional repositories and the Europe PubMed Central feed, which together highlight approximately 70 percent of papers cited in my stories before they appear in print. When a paper remains behind a paywall, I utilize the author’s accepted manuscript, clearly labelled as such, and link to the version of record once it is released. For clinical trial data, I check the registry record on ISRCTN or ClinicalTrials.gov to confirm primary outcome measures before quoting efficacy percentages, because press releases sometimes emphasise secondary endpoints that were not statistically powered.
Financial Implications and Time Implications
Article processing charges for open-access journals typically range from £1 500 to £4 500, figures that appear on the journal website before submission. When a paper is only available through a read-and-publish agreement negotiated by Jisc, I record the licence type in the story’s reference list so readers know they can access the full text without payment. These administrative details rarely appear in the final article, yet they influence whether a claim can be linked directly to primary evidence or must be qualified with the phrase “according to a pre-print server”.
Essential Tools for Data Processing and Fact-Checking
Any dataset that is introduced into a story is saved into a local repository with file-naming conventions that feature the source URL, retrieval date and checksum katherineharmon.com. I employ R for initial cleaning when the file goes beyond 10 000 rows, and Python’s pandas library when joins across multiple tables are required. A separate plain-text log records each transformation step so that an editor or an interested reader can duplicate the figures. For statistical claims I consult the original paper’s supplementary information and, where possible, the analysis code archived in repositories such as Zenodo. When numbers appear in press releases without underlying data, I ask for the raw spreadsheet through the institution’s freedom-of-information route and note the reference number in the story’s endnotes.
Version Control and Handoff to Editors
Manuscripts are drafted in plain Markdown inside a private Git repository. Each commit message contains the working headline, the word-count target and the initials of any colleague who has reviewed a section. Branches are titled by story slug and close after publication, keeping a complete record of changes asked for by editors at outlets such as New Scientist or BBC Science. This system avoids accidental overwrites during late-stage edits and allows rapid retrieval of an earlier paragraph if a legal query comes up months later.
Preservation, Credit and Long-term Access
Published articles are downloaded as PDF/A files within 48 hours and kept in a local archive that is secured to an off-site encrypted drive. Each file is labelled with the ISO date, outlet and slug, allowing quick location when a reader asks for a citation years later. I also place supplementary material such as interview transcripts or analysis scripts in the Zenodo community “Science Journalism UK”, which issues a DOI and guarantees availability for at least twenty years. When a story relies on a dataset that cannot be disclosed for privacy reasons, I record a metadata record that details the variables and references the custodian’s access procedure.
Attribution Language for Secondary Sources
Secondary sources such as systematic reviews or government statistical releases are framed with a short clause that specifies the producing organisation and the publication date. For example, “Office for National Statistics data released on 14 March 2024 show…” This format enables readers to judge recency without searching through footnotes and lowers the risk of attributing a finding to an individual researcher when the work was in fact collaborative.
Establishing and Sustaining Source Networks
Trustworthy reporting depends on a recorded network of scientists, clinicians, statisticians and press officers who reply within set time frames. I maintain a contact file that records each person’s institutional affiliation, publication record and favored channel, whether that is an institutional email or a confirmed Signal number. Before any interview I send a one-paragraph description of the angle, the planned publication and the deadline, which reduces back-and-forth and demonstrates respect for the source’s schedule. After publication I append a short note on how the interview was used, allowing future queries to be precise rather than generic. The file is checked every quarter so that retired researchers are substituted and new group leaders at institutes such as the Francis Crick Institute or the UK Atomic Energy Authority are included while their work is still in preprint.
Verifying Credentials and Evading Conflicts
Each new contact is cross-checked against the institutional website and recent peer-reviewed papers shown in PubMed or Web of Science. If a researcher has secured funding from a company whose product shows up in the story, that fact is noted in the same file and revealed in the article. I also log when a source has previously counseled government committees, because such experience can add context but must not be framed as independent commentary. This running ledger prevents the common error of citing the same small circle of voices and guarantees that readers obtain a balanced range of perspectives.
Principled Decision-Making in Daily Practice
Ethical issues surface most commonly around patient narratives, commercial sponsorship and the use of unpublished pictures. For patient accounts I obtain written consent that details the level of identification permitted, and I file the form separately from the interview transcript. When a company provides early access to a device or data set, I demand a written statement that no editorial control is associated; the statement is stored with the project notes and outlined in the published article if the material impacts the central claim. Photographs shot at laboratories are authorized with the press office, and any image that depicts an individual’s face is accompanied by a model-release form even if the person is a researcher rather than a patient.
Handling Corrections and Clarifications
Errors are logged in a public corrections area on my website, with the date of the original article, the incorrect statement and the corrected language. If the mistake alters a numerical value by more than 5 percent or shifts the direction of a finding, I also alert the publisher’s corrections team and seek that the change be noted in the online version. This method has resulted in three published clarifications over four seasons, each fixed within ten days of the original publication date.
Choosing Projects That Meet the Needs of a UK Readership
Project selection begins with an assessment of relevance to readers who monitor policy decisions issued by Westminster and research programmes funded by UK Research and Innovation. I keep a rolling list of fifteen topic areas, covering antimicrobial resistance surveillance by the UK Health Security Agency to offshore wind deployment targets defined by the Department for Energy Security and Net Zero. Each candidate story is rated for timeliness, availability of primary data and potential to impact daily life, such as changes in energy bills or medicine supply. Stories that score above an internal threshold move to a one-page treatment outlining the central claim, the evidence required and the expert voices needed to validate it. This scoring sheet travels with the assignment from pitch to publication, providing editors a transparent audit trail.
FAQ
How do you determine which scientific papers deserve coverage for a UK audience?
I evaluate each paper on applicability to current UK policy, presence of primary data and possible impact on readers’ daily lives. Only papers that meet a minimum score progress to a one-page treatment that details required evidence and expert voices. The scoring sheet remains with the assignment from pitch to publication.
What actions do you take to verify a new source’s credentials?
Each contact is reviewed against their institutional page and recent publications in PubMed or Web of Science. Any industry funding or advisory roles are recorded in a private file and disclosed in the article if relevant. The file is reviewed quarterly to keep details current.
What software do you use for data cleaning and version control?
Large datasets are cleaned in R or Python’s pandas library, with every transformation recorded in a plain-text log. Manuscripts are drafted in Markdown inside a private Git repository so that editors can access the complete change history and earlier versions can be recovered if needed.
How do you address corrections after publication?
Errors are registered on a public corrections page with the original wording and the correction. If the change goes beyond five percent or changes a finding’s direction, the publisher’s corrections desk is also informed. Most clarifications are addressed within ten days.
What becomes to interview transcripts and datasets after a story is published?
Articles are archived as PDF/A files and backed up off-site. Supplementary material such as transcripts or analysis scripts is deposited in Zenodo with a DOI. Datasets that cannot be shared for privacy reasons obtain a metadata record referring to the custodian’s access procedure.
Do you require payment readers to access your published articles?
No. All articles are published in outlets that offer open access or are placed behind a publisher paywall that libraries can manage through Jisc agreements. I do not run a separate subscription site or position articles behind an additional paywall.