Liquid-Handling Robot vs Human Pipetting

Liquid-Handling Session

Pipetting speed 6 wells/s

Human fatigue intensity 1.2×

Session totals

Wells pipetted

Plates completed

Human technician

Mean error

–

Variability (σ)

–

Automated robot

Mean error

–

Variability (σ)

–

How it works

Both lanes pipette the same nominal volume into every well of a 96-well plate, over and over, plate after plate, so the session accumulates hundreds and then thousands of repeated pipetting actions. Each action draws a random volume error from a Gaussian distribution.

human σ(n) = σ₀·(1 + fatigue·(1 − e^(−n/N)))
robot σ = σ₀_robot  (constant, ≈6× tighter)

Human technician — every pipetting action carries natural hand-tremor and meniscus-reading noise, and as consecutive repetitions n pile up over the session, both the spread and a small systematic bias grow toward a fatigued plateau — visible as the plate wells drifting from green toward orange/red later in the session, and as the rolling chart's human line climbing.
Automated robot — the calibrated gantry executes an identical motion on well #1 and well #10,000 alike; its error distribution never depends on repetition count, so its wells stay green and its chart line stays flat.
Pipetting speed — how fast the session advances (wells per second); does not change the physics, only how quickly you accumulate repetitions.
Fatigue intensity — how strongly session-length degrades the human's accuracy; at 0 the human behaves like a (noisier) robot, at 3 fatigue dominates quickly.

This is why large-scale assays — screening thousands of drug-candidate compounds across many plates — rely on automated liquid handlers: manual pipetting at that scale doesn't just add noise, its errors get systematically worse as the session goes on.

Low volume error Moderate error High error

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.

Rolling per-well volume error (%) HumanRobot