The Jar Test Priced in Time and Error: Settling Times, Layer-Reading Errors, and the Protocol Stokes' Law Actually Supports for Soil Texture by Sedimentation

The soil texture jar test — shake soil in water, let it settle, and read the sand, silt and clay layers against a clock — is usually taught with two fixed reading times, about 1 minute for sand and about 2 hours for silt, with clay taken as the remainder. This paper computes, from Stokes' law with stated particle sizes, densities and water properties, what those times actually deliver. At a settling depth of 100 mm and 20 °C, a particle at the sand/silt boundary (50 micrometres) needs 44.5 s to fall the full depth, while a particle at the silt/clay boundary (2 micrometres) needs 7.73 h, and the finest clay fraction (0.2 micrometres) needs 32.21 days, so a clay layer read directly is never complete on any practical timescale. Because a shaken jar starts well mixed, layer boundaries are blurred by particles that began near the bottom: reading the sand layer at 1 minute overstates the sand content of a loam (40 percent sand, 40 percent silt, 20 percent clay) by 8.1 percentage points, and reading the silt layer at 2 hours overstates its clay content by 2.68 percentage points while 9.5 percent of its silt is still suspended and 5.6 percent of its clay has already settled into the wrong layer. Water temperature moves the silt/clay reading materially: the same 2-hour reading in 10 °C water overstates loam clay by 4.23 percentage points, against 1.38 points at 30 °C. No single silt reading time suits every soil — the error-cancelling time is 3.67 h for a sandy loam, 3.25 h for the loam and 1.50 h for a clay soil — so the protocol the arithmetic supports (100 mm depth, sand read at the computed boundary time of about 45 s at 20 °C, silt read at 3.25 h, clay always by difference, temperature stated) carries an honest precision for clay of about ±11 percentage points in the worst case across the soils examined. The jar test is a coarse classifier, not a measurement of texture to the nearest percent. This study is simulated, not measured.