A lab manager reviewing a draft report finds a bar chart with tight error bars, an asterisk and the caption “n = 3.” Asked what the three were, the analyst explains that one stock solution was diluted once, pipetted into three wells on one plate and read together. Nobody did anything dishonest. But the chart implies the finding was reproduced three times, and it was really measured once, three ways. Whether that matters depends entirely on what the chart is being used to claim.
The difference between technical and biological replicates is one of the most common sources of overconfident results in laboratory science. It is easy to state and surprisingly easy to lose track of.
Two kinds of repeat, two different questions
A technical replicate repeats the measurement of the same prepared sample: several wells from one dilution, several reads of one plate, several sample introductions from one HPLC vial. The spread among technical replicates tells you how precise your measuring process is.
A biological replicate (sometimes called an independent or experimental replicate) is a separate instance of the system under study: a different culture, prepared independently, ideally on a different day and from a different passage, with its own fresh dilution of the test compound. The spread among biological replicates tells you how much the system itself varies.
Both are useful. Neither can stand in for the other. The mistake is not running technical replicates; it is counting them as if they were independent evidence about the system.
Define the unit before counting
The fastest way to sort out any replicate question is to ask: what is one independent observation in this experiment? The answer depends on the claim. If the claim is about how a population of cultures responds, one culture is the unit. Wells within that culture are subsamples of a single observation.
Statisticians call the error of counting subsamples as independent units pseudoreplication. The name is useful because it points to the fix. No clever test or adjustment rescues a pseudoreplicated analysis. The analysis has to be done at the level of the claim: average the wells within each culture first, then compute statistics across cultures. Where both levels carry real information, a mixed or nested model that represents the structure explicitly is the more complete approach, and simple averaging is its rough approximation.
Why the numbers look better than they are
Technical replicates share almost everything: the same preparation, the same dilution, the same analyst and the same few minutes. They will naturally agree more closely than independent repeats would. Counting them as the sample size produces two errors at once:
- the variability is underestimated, because the shared sources of error cannot show up within the group; and
- the sample size is overstated.
Both push the standard error down. Error bars shrink, and a significance test becomes more likely to cross its threshold. The result is a figure that looks precise and a p-value that looks convincing, neither of which says much about whether the effect would appear again in a fresh experiment.
Sorting common designs
Real experiments do not always fall neatly into one box. The useful question is which sources of variation the replicates share. This table covers the arrangements most labs meet:
| Design | Classification | What is shared |
|---|---|---|
| Three wells from one dilution on one plate | Technical | Everything except the pipetting of each well |
| Three plates from one dilution on one day | Technical with respect to preparation | Dilution error and any surface losses during preparation |
| One culture, compound added once, read three times over an hour | Technical | The entire biological and preparative history |
| Three cultures, each exposed to its own fresh dilution, on three days | Biological | Only the stock lot and protocol |
| Three independent cultures all exposed from one dilution made once | Partly independent | The compound preparation; one bad dilution affects all three |
The last row is extremely common. It is stronger than a technical triplicate and weaker than a fully independent one. The honest way to report it is to say plainly what the replicates had in common.
What to do with each kind
Technical replicates should be averaged into a single value per independent unit. Their spread is a quality control check on the measurement. If technical replicates disagree badly, that signals a problem to fix, such as pipetting error, bubbles, edge effects on the plate or instrument drift, rather than something to average away.
Biological replicates are the data. Their number is the sample size, and means, standard deviations and tests are computed across them.
A short way to remember it: technical replicates make each point sharper; biological replicates are what you count.
The habit of three
Triplicate is a convention, not a calculation. Three is the smallest number from which a variance can be estimated, and that estimate is rough. Experiments with low power tend to detect only the effects that happened to come out large, which inflates reported effect sizes. Adding biological replicates addresses this. Adding more technical replicates does not, because they add no information about the variability that limits the conclusion. When planning how much material an assay series needs, budget for independent repeats first; our guide to budgeting peptide for an assay series walks through that planning.
The same logic applies to analytical chemistry
Replicate structure is not only a cell biology concern. On an HPLC system:
- three sample introductions from one vial characterize the instrument’s repeatability;
- three solutions prepared from one weighing characterize the preparation step as well;
- three separate weighings capture weighing and dissolution error too;
- samples taken from different containers begin to characterize the material itself.
When a purity result is reported, knowing which of these it reflects tells you what the precision figure really covers. For background on the method, see reverse-phase HPLC for peptide purity, and for what independent testing does and does not mean, see what third-party tested means.
A reporting checklist
- State what one independent unit was.
- State whether n counts wells, plates, cultures or independent experiments.
- Note whether repeats were on separate days and used fresh dilutions.
- Say how technical replicates were combined before analysis.
- Describe anything all replicates shared.
Questions
Are technical replicates a waste of time?
No. They measure the precision of your assay and catch pipetting or instrument problems. They just should not be counted as the sample size for a claim about the system.
Can a statistical correction fix pseudoreplication after the fact?
Not by itself. The analysis has to be redone at the correct level, by averaging within each independent unit or by using a model that represents the nesting.
How many biological replicates are enough?
It depends on the expected effect size and the variability of the system. A power calculation is the proper way to decide; three is only a minimum convention.
Does this apply to HPLC purity testing?
Yes. Repeat runs from one vial describe the instrument, while independent preparations and samples describe the material.
Research use only. All products supplied by Battle Born Peptides are laboratory reference materials for in-vitro research and analytical use by qualified professionals. They are not drugs, foods, dietary supplements, cosmetics or medical devices; they are not approved by the FDA or any other regulator for use in humans or animals; and they are not intended to diagnose, treat, cure, mitigate or prevent any disease, or to affect the structure or any function of the body of humans or animals. Nothing in this article is preparation, handling or dosing guidance. See our full research-use terms.