Purpose: Three previously published datasets of high-mountain soil variation in proglacial valleys in the Swiss Alps (80 soils) and a new dataset of high-mountain soil variation in a formerly glaciated valley in the Colorado Rocky Mountains (9 soils) are used to test the validity of the chronosequence approach and to study divergence and convergence of soil properties. Materials and methods: Standard field-based soil observations were done, complemented with simple laboratory measurements of pH and soil organic matter. Results and discussion: The mean values of soil properties change over time, as well as their standard deviations and coefficients of variation. Variation in soil properties between soils of the same age is significant. Although sampling was performed at locations that are assumed to be geomorphically stable, the observed variation in properties casts doubt on this assumption. Depending on the valley and the soil property, standard deviations and coefficients of variation increase over time whereas in other cases, they decrease. This indicates divergence and convergence of soil properties over centennial and Holocene timescales, respectively. Both dynamics are explored quantitatively. Conclusions: Divergence is observed in settings that are unaffected by outside (hillslope) influences and presumably caused by vegetation differences and small-scale (diffusive) redistribution of the fine earth fraction. Convergence is observed in settings where soil formation is disturbed by outside influences. In the Swiss Alps, this influence is the provision of material from surrounding hillslopes. Chronosequence studies should sample and average multiple soils per age group, to characterize soil variation and minimize the uncertainty in the estimation of soil properties. © 2014, Springer-Verlag Berlin Heidelberg.