Aqueous classes are forgiving because water more or less does what a liquid handler expects. Glycerol, DMSO, ethanol, and detergents do not, and the correction curve you build for them looks different: larger, steeper, and more sensitive to how the rest of the class is set. The mistake is to treat the bigger correction as the fix for a hard liquid. It is the last step, not the first, and getting the order wrong produces a curve that papers over problems the flow rate should have solved.
Why hard liquids need more correction
A viscous liquid clings to the tip and lags the plunger, so more of what you aspirated stays behind and the delivered volume runs short by a margin that grows with viscosity. A volatile liquid loses material to evaporation and pressure changes in the tip, and its first transfer often behaves differently from the rest as the tip equilibrates. Both produce a larger systematic shortfall than water does, which is exactly what a correction curve exists to cancel. Opentrons makes this explicit in its liquid handling for viscous liquids: the recommended classes pair a slower flow rate with a distinctly larger correction by volume than the aqueous default. The bigger correction is real and expected.
But the curve comes last
Here is the ordering that saves you. A correction curve cancels a steady, repeatable shortfall. If the shortfall is not steady, because the liquid is trailing inconsistently or the first shot is an outlier, the curve is correcting a moving target and will never settle. So the mechanical settings have to go first, to make the behavior repeatable, and only then does the curve have something stable to correct.
- Flow rate: slow the aspiration and dispense so a viscous column has time to follow the plunger instead of lagging differently each time. Some classes drop to a small fraction of the water rate.
- Settling time: lengthen it so the column equalizes before the tip lifts, which tightens the shortfall into a consistent one.
- Dispense mode: prefer a surface dispense over a jet for clingy liquids, so the liquid is laid against the wall rather than fired through air it will bead in.
- Blowout: increase it for both viscous and volatile liquids, so the tip clears fully and vapor has room and force to leave.
- Pre-wetting: for volatiles especially, wet the tip once so the first transfer matches the rest rather than standing out as an outlier the curve cannot represent.
Do these first and the shortfall becomes a stable, sizable bias. Now the larger correction curve has a fixed thing to cancel, and it works.
Shape, not just size
Hard liquids also bend the curve, they do not just shift it. The fraction held back by a viscous liquid changes with volume, and a volatile liquid loses a different proportion at small volumes than large, so a single slope across the whole range fits worse than it does for water. The practical response is more points, more closely spaced, especially at the low end where the proportional error is largest. You are tracing a curve that genuinely curves, not fitting one straight line to a liquid that refuses to be straight.
Do not forget temperature
Both viscosity and vapor pressure move with temperature, which means a curve fitted on a cold morning can be wrong by afternoon. Fit and verify at your real working temperature, note that temperature with the class, and treat a seasonal or bench-to-bench temperature swing as a reason to re-check the curve rather than a mystery to puzzle over later.
For hard liquids the correction is bigger and the curve is bendier, but both come last. Make the shortfall repeatable with flow rate and technique first, then correct the steady bias that remains.