Munsell Soil Colour Charts are widely used by soil scientists as the most unbiased way for determining the colour of soil samples. However, there are only a few studies on the variance in people's assessment of the colour of soil samples on this scale and the factors that determine this variance. In this work, we analyze the colour estimates of > 100 undisturbed soil samples by 20 people. For diverse soil samples collected from Anthrosols, Arenosols, Histosols, Podzols and Retisols at the Smolenskoye Poozerye national park, the fraction of votes that support the mode differs from 0.15 to 0.63 with mean of 0.29. The agreement of people’s estimates is significantly less pronounced for hue than for value (p = 0.0001) and chroma (p = 0.002) in dark and muted, but not in light and saturated samples. On average, females vote for mode in more fraction of samples than males (p = 0.026). Moreover, there are people who vote for and against mode more frequently that it is expected by chance. Within-sample variances of estimates calculated after conversion into RGB and CIELAB systems correlate weekly with uncertainty of estimates calculated as fraction of votes that do not support the mode. Soil horizons have different variances in colour estimates. The uncertainty in colour estimation with a Munsell Soil Colour Chart is highly dependent on both sample characteristics and people. The agreement between participants in colour assessment increases substantially when neighboring Munsell chips are treated as equal.