For railway frost protection- and subballast layers, requirements are given for both the material grain size distribution and the bearing capacity and density of the structure. In many cases, the required bearing capacity has not been possible to be met with materials that fulfill the grain size distribution requirements. Materials with grain size distribution curves close to the fine end of the requirement zone and materials with excess amounts of sand grains have been noticed to be problematic. The goal of this thesis was to find out, how the grain size distribution of natural materials affects the bearing capacity. Another phenomenon studied in this thesis was the effect of time and water content on bearing capacity. Bearing capacity of a structure has been seen to increase as time passes even if the structure has remained mostly untouched. Falling weight deflectometer tests on a railway structure were used to study the true distribution of bearing capacity values and the effect of time and grain size distribution on bearing capacity. Test embankments built from different materials were used to study the effect of grain size distribution, water content and compaction on bearing capacity and density. Tests conducted on the test embankments were plate load, falling weight deflectometer and Troxler test. The effect of grain size distribution, time and water content was also studied on a test structure indoors. In addition to evaluating the effect of grain size distribution with traditional methods, it was studied from the point of view of particle packing theory. In this thesis, a particle packing based calculation method was applied in identifying those grain sizes, that form the load bearing, and the grain sizes, that either support or disturb the skeleton structure. The stiffness of granular material is great when loads are transferred through the skeleton of large particles and the fine materials are supporting the larger particles. The calculation method can be used to recognize parameters from the grain size distribution that correlate with bearing capacity. Empirical equations defined in this thesis can be used to estimate the bearing capacity from grain size distribution by using the calculation method. The bearing capacity is affected greatly by the location of the grain size distribution curve inside the requirement zone. Higher bearing capacity can be reached with coarse materials than with fine materials. In the calculations, this shows up as the amount of materials that pass the 2 mm sieve. Bearing capacity can be lowered by a breaking of the skeleton structure. This shows as a low quantity of a specific grain size in the material. Bearing capacity of a structure increases as time passes due to suction i.e. apparent cohesion. This increase is significant on natural materials. Heavy watering can partly cancel this effect, but small amount of water doesn’t yet lower the bearing capacity.