03 · What You Need to Know
The Problem Is Not Change Itself but Untraceable or Misleading Change
Research develops through revision. Data-entry errors are corrected. Code improves. Qualitative codes evolve. Instruments are amended. Derived variables are recalculated. Researchers add clarifications to earlier notes.
Requiring every research record to remain permanently frozen in its first form would make accurate research harder, not safer.
What matters is whether the record continues to represent its history faithfully.
Correction Is Different From Concealment
A legitimate correction acknowledges that an earlier entry was wrong and creates an accurate replacement or annotation while preserving appropriate evidence of the change.
Concealment instead makes the record appear as though the error, earlier value, or earlier decision never existed.
Correction
Changes inaccurate information while preserving appropriate evidence of the original entry, the change, and its provenance.
Improper alteration
Changes or removes information in a way that misrepresents the research history or prevents the original record and relevant change from being understood.
The boundary matters because falsification in the U.S. Public Health Service research-misconduct framework includes manipulating research materials, equipment, or processes, or changing or omitting data or results such that the research is not accurately represented in the research record. A correction that makes the record more accurate is conceptually different from a change designed to make the research history misleading.
Do Not Make the Original Information Disappear When Its History Matters
FDA guidance for computerized systems used in clinical trials provides a clear regulated example: a change to a required record should not obscure the original information, and the record should indicate that a change occurred while providing a way to locate and read the prior information.
That particular requirement belongs to an FDA-regulated context, but the principle is broadly useful. If an important record is changed, someone reviewing it later should not be misled into believing that the revised version was necessarily the original version.
Paper Records and Electronic Records Need Different Mechanisms
With a paper laboratory notebook, a correction might preserve the original entry visibly while adding the corrected information, date, and appropriate attribution according to the laboratory's procedures. Erasing, obliterating, or removing the original entry can destroy provenance.
Electronic systems require different controls. Depending on the system and applicable requirements, version history, timestamps, user identities, immutable logs, or formal audit trails may preserve the sequence of changes.
The objective is the same even when the technology differs: distinguish the original state from later modifications.
An Audit Trail Is More Than a Backup
FDA's guidance for computerized clinical-trial systems defines an audit trail as a secure, computer-generated, time-stamped electronic record that permits reconstruction of events relating to the creation, modification, and deletion of an electronic record.
In systems subject to the cited FDA requirements, audit trails record the date and time of operator actions that create, modify, or delete electronic records, and the audit trail must itself be protected from modification by personnel creating or changing the underlying records.
This should not be generalized into a claim that every spreadsheet used in academic research legally requires a Part 11 audit trail. It does illustrate what strong change provenance can look like.
A New Version Is Sometimes Better Than Editing the Old One
Not every research file should be edited in place.
Suppose you discover a mistake in the code used to create an analytical dataset. Rather than silently replacing the dataset and pretending the corrected version always existed, a stronger workflow may preserve the source data, correct the code, generate a new analytical version, document the reason for the change, and retain enough version history to establish which dataset supported which analysis.
This approach separates correction from historical revisionism.
Raw or Source Evidence Deserves Particular Protection
ORI's current guidance on research records emphasizes the evidentiary importance of raw data and associated metadata during research-misconduct proceedings. It notes that relevant evidence may include source files, instrument-generated data, spreadsheets, statistical files, figures, and other materials used to generate data, document results, and draw conclusions.
Preserving source evidence is important because later processing can legitimately produce many derived versions. Without the original or appropriate source record, however, it may become impossible to determine what changed.
This is one reason the difference between research data and the broader research record matters. The history connecting source evidence to later versions can itself be part of what needs preservation.
Adding Information Later Can Be Legitimate if the Timing Is Honest
A researcher may realize that an original record lacks important context. Adding an annotation can be appropriate.
The problem arises if the addition is backdated or presented as though it existed contemporaneously. A later annotation should be identifiable as later. If it reconstructs an earlier event, the basis for that reconstruction should be clear where consequential.
This protects both the researcher and the record. An honest late clarification is not the same thing as a contemporaneous observation, and the documentation should not pretend otherwise.
Corrections to Data Should Propagate Through the Research Workflow
Changing a value in a source or master dataset may have downstream consequences. Analytical datasets, tables, figures, statistical outputs, reports, manuscripts, repository deposits, or publications may need reassessment.
A good correction process therefore asks not only “Did we fix the record?” but also “What depended on the incorrect information?”
The separate issue of how corrections should be documented becomes particularly important when a change affects multiple stages of the research workflow.
Deleting an Incorrect Record Is Not Necessarily the Same as Correcting It
Researchers sometimes discover an obviously incorrect entry and instinctively want to remove it. Whether deletion is appropriate depends on the record, system, retention requirements, and circumstances.
If the incorrect entry forms part of the evidentiary history, deleting it may destroy information needed to understand what happened. In other situations, controlled deletion may be legitimate and recorded through an audit trail or records-management process.
The more specific question of whether it is acceptable to delete an incorrect research record therefore cannot be answered simply by saying that incorrect information has no value.
Version Control Helps When Research Materials Are Expected to Evolve
Some research records are naturally iterative. Analysis scripts, codebooks, protocols, instruments, coding frameworks, manuscripts, and data-processing workflows may pass through numerous versions.
Version control can establish which state existed at a particular time, who changed it, and which version was used for a particular output. This may involve specialized version-control software, electronic laboratory notebooks, repository histories, controlled document systems, or simpler structured naming and archiving practices appropriate to the project.
The important point is not to create dozens of files named “final_FINAL_revised2.” The point is to preserve an intelligible history.
Watch Out
Never alter an existing research record to make it appear that information was recorded earlier than it actually was. If you need to clarify or reconstruct an earlier event, preserve the chronology and identify the later entry as such.
