Examples
Worked examples
- Is an instance
A Snakemake pipeline producing identical FASTQ-to-VCF outputs across runs on the same container image.
- Is an instance
An R Markdown notebook re-knitting to bit-identical figures after a `renv::restore()`.
Counter-examples
Looks similar, but isn't
- Not an instance
Replication of a wet-lab assay (not computational).
- Not an instance
A new field study (different scope of reproducibility).
Editorial commentary
Computational reproducibility is the narrow technical sense of reproducibility: obtaining the same numerical outputs from the same data and code, run on a comparable computational environment. It is a necessary — but not sufficient — condition for the higher-order reproducibility claims covered elsewhere in this family.
Where it sits relative to the rest of the family
Computational reproducibility is the most mechanical, most narrowly technical member of the reproducibility family, and the one most amenable to tooling. It says nothing about whether the methods were reported clearly (methods reproducibility), whether the specific published numbers match what a re-run produces (results reproducibility — closely related but distinct, since a pipeline can run cleanly and still not reproduce the exact published table), or whether the underlying physical or biological procedure would give consistent results if redone (empirical reproducibility). A study can pass computational reproducibility — the code runs, the numbers come out — while still resting on an empirically fragile result.
Common, mundane failure causes
Computational reproducibility fails for reasons that are usually fixable with better practice rather than indicating anything wrong with the underlying science: a data file that wasn’t actually deposited, an undocumented random seed, operating-system-dependent floating-point behaviour, a library or package version that changed its default behaviour between the original run and the re-run, or hard-coded local file paths. This is precisely why it is the principal target of containerisation (Docker, Singularity), workflow-capture tools (Nextflow, Snakemake), and computational-environment-capture practices generally — the problem is largely an engineering-discipline one, distinct from the harder empirical question of whether the science itself replicates.
Sources
Stodden, Seiler & Ma, “An empirical analysis of journal policy effectiveness for computational reproducibility,” PNAS (2018).
Also known as
execution reproducibility · bitwise/numeric reproducibility
Machine-readable encodings
Use in your systems
<role vocab="credit"
vocab-identifier="https://casrai.org/dictionary/"
vocab-term="Computational reproducibility"
vocab-term-identifier="https://casrai.org/dictionary/term/computational-reproducibility" />{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"@id": "https://casrai.org/dictionary/term/computational-reproducibility",
"name": "Computational reproducibility",
"identifier": "https://casrai.org/dictionary/term/computational-reproducibility",
"description": "The narrow technical sense of reproducibility: obtaining the same numerical outputs from the same data and code, on a comparable computational environment.",
"inDefinedTermSet": "https://casrai.org/dictionary/domain/reproducibility#set",
"url": "https://casrai.org/dictionary/term/computational-reproducibility",
"sameAs": [
"execution reproducibility",
"bitwise/numeric reproducibility"
],
"license": "https://creativecommons.org/licenses/by/4.0/",
"publisher": {
"@id": "https://casrai.org/#organization"
},
"author": {
"@id": "https://casrai.org/#editorial-team"
},
"datePublished": "2026-05-21T02:22:49",
"dateModified": "2026-08-22T14:39:17",
"inLanguage": "en-GB",
"isAccessibleForFree": true
}
