Replication Type Declaration

Treating technical replicates as independent biological replicates is pseudoreplication —one of the most consequential and common errors in experimental biology. It inflates n, deflates standard error, and produces artificially narrow confidence intervals. The result is not just misleading: it is unreproducible by design.

STEP 1 —The Pitfall

Technical replicates are repeated measurements of the same sample. Biological replicates are measurements from independent biological units (separate animals, patients, donors, or cell lines). Combining 3 mice x3 technical replicates per mouse to get n=9 —and then treating these as 9 independent data points —is pseudoreplication that inflates statistical power by approximately 3-fold.

The confusion is easy to make when data arrive as a spreadsheet with no column-level metadata. A column called "n" tells you how many rows there are; it does not tell you how many independent biological units they represent.

[Image placeholder: Two analysis trees. Left: 3 mice x3 technical reps xn=9 (pseudoreplication). Right: same data correctly analyzed as n=3 biological replicates with technical reps averaged or modeled as random effect. Shows the change in SE and CI.]

Pseudoreplication (left) vs. correct analysis (right): n=9 vs. n=3 from the same experiment. SE shrinks by approximately 3xin the pseudoreplication case.

STEP 2 —Journal Requirement

ARRIVE 2.0 requires explicit statement of what constitutes a biological replicate vs. a technical replicate. Nature Methods and eLife require "n represents the number of biologically independent experiments" or equivalent. Journals routinely require revision when technical and biological replicates are conflated in a figure legend.

STEP 3 —Licklider Solution

Input

Subject / animal / donor / batch ID columns declared via Variable & ID Mapping
Observation unit declared via Observation Unit Declaration
Replication type declaration: biological / technical / mixed

Output

Biological replicate count (unique subject IDs) vs. total row count displayed per group
Verification: if subject ID column exists and n_rows > n_subjects, a pseudoreplication warning is raised
Suggested handling: average technical replicates per subject, or model subject as random effect
Validity Score dimension: Fail if n used in inference = n_rows when n_rows > n_subjects and no subject-level aggregation or mixed model is declared
"n represents biological replicates" text auto-generated for figure legends

Guard

If subject ID columns exist and the ratio of n_rows to n_subjects exceeds 1 (multiple rows per subject), the guard raises a mandatory dialog asking the user to declare the replication structure. Proceeding without resolution results in a Fail on this Validity Score dimension.

[Image placeholder: Licklider replication type dialog showing the detected structure: 3 subjects x3 rows = 9 total rows. Options: "Average technical replicates per subject (n=3)", "Model subject as random effect (mixed model)", or "These are independent replicates" —with the last option triggering a confirmation dialog.]

Replication structure dialog: Licklider detects repeated measures per subject and requires explicit resolution.

STEP 4 —Draft Output (Draft / Needs review)

Data represent n = 6 biologically independent experiments (animals),
each measured in technical triplicate. Technical replicates were averaged
per animal prior to statistical analysis. Statistical comparisons were
performed on the biological replicate means (n = 6 per group).