You spend a week getting a correction curve right on one deck, then a second identical-looking instrument arrives, or a partner lab wants to run your method, and the obvious move is to copy the curve across. It rarely survives the trip. The delivered volumes drift, the curve that was perfect is now subtly wrong, and the temptation is to blame a broken instrument. The real explanation is more interesting, and a long-running discussion among practitioners about standardizing liquid handlers circles it repeatedly: a correction curve is a description of one physical system, and physical systems differ in ways the numbers cannot carry with them.
What the numbers actually encode
A correction curve looks like abstract calibration data, but every point in it is the fingerprint of a specific machine doing a specific thing. The fit absorbs the syringe or plunger geometry, the exact tip and its bore, the pressure path from pump to nozzle, the labware, and the small mechanical quirks of that particular unit. When you copy the curve, you copy a description of all of that. Move it to a second instrument and every one of those variables has shifted a little, so the correction that inverted the first machine's behavior no longer inverts the second machine's behavior. The curve is not portable because the thing it describes did not come along.
This is why two units with the same model number on the same bench still want their own curves. Nominal sameness is not physical sameness, and the correction lives entirely in the physical detail.
The idea travels, the fit does not
It helps to separate two things that get bundled under the word transfer. The intent of a class travels fine: which liquid it is for, that it aspirates slowly and dispenses at the surface, that it needs a generous blowout, roughly where its curve sits. That knowledge is worth sharing and reusing. What does not travel is the fitted correction, the exact slope and offset or the exact point table, because that is machine-specific by construction. A useful transfer carries the intent and re-measures the fit, rather than pretending the fit is universal.
The standardization discussion makes the same point from the other direction, half in frustration: across labs almost nothing is truly common. As one participant put it, the one thing everyone actually agrees on is the SBS plate footprint, and everything above that is locally decided. If the community cannot agree on much beyond the plate format, a correction curve tuned in one shop is never going to drop cleanly into another.
What would make curves portable
The more forward-looking part of that discussion is not resignation but an idea: if the numbers cannot transfer because they encode physics, then model the physics and translate between machines. Instead of copying a curve verbatim, you would describe the behavior in physical terms, bore size, shear at the tip, the pressure response, and use those to approximately map a class from one instrument to another. It is harder than a copy and paste, and nobody pretends it is exact, but it is the only route that respects why the curve was machine-specific in the first place.
There is a second caution the same people raise. Gravimetric correction tunes for delivered volume, but volume is not the only thing that matters. The speed and acceleration of pipetting affect shear-sensitive biomolecules, and two instruments that deliver the identical volume can treat a fragile sample differently. So even a perfectly translated curve can leave a real difference behind if the assay cares about how the liquid was moved and not just how much arrived.
The habit that saves you
The practical stance is to treat a transferred curve as a starting point and never as a finished one.
- Carry the intent: move the liquid family, the mechanical settings, and the rough shape of the curve as a head start.
- Re-measure the fit: run replicates on the receiving instrument and refit the curve against its real behavior before trusting a single sample.
- Requalify per unit: do this for each physical machine, not each model, because two of the same model still differ.
- Watch beyond volume: for shear-sensitive work, verify the assay outcome and not only the delivered microliters.
A correction curve describes one machine, not a method. Share the intent, re-measure the fit on the instrument that will run it, and never trust a curve you did not measure where it will be used.