Suffusion is an internal erosion mechanism that occurs in embankment dams when fine grained particles in the dam core are washed out by seepage. Initiation of internal erosion depends mainly on three major factors: grain size distribution of the soil, stress conditions and hydraulic gradient; whilst its continuation depends on the properties of the filter. Broadly graded moraines, as glacial tills, are more susceptible to internal erosion by suffusion than other types of soils used in dams. Most embankment dams in Sweden consist of a central core of glacial till built more than 50 years ago. At the time of its construction, the available guidelines did not include specific grain size boundaries for the core and the filter related to internal erosion susceptibility. Today, several Swedish embankment dams have experienced incidents of internal erosion such as leakages and sinkholes, making internal erosion an important safety issue. This circumstance leaded to the question: what are the conditions triggering internal erosion by suffusion in embankment dams? This research aims to contribute to the assessment of dam safety by increasing the knowledge on glacial till soils regarding: i) the optimum empirical method to evaluate the susceptibility to suffusion; ii) the effects of boundary and testing conditions in the experimental evaluation of suffusion, and iii) summarize reference values of the hydraulic gradient triggering the initiation of suffusion. The thesis includes a literature review on the existing methods to evaluate soil’s susceptibility to suffusion, a comparison among the difference testing and boundary conditions applied in the experimental assessment of soil’s susceptibility to suffusion, and an experimental study aiming to determine the critical hydraulic gradient for suffusion to initiate in glacial till soils (ic). The experimental study includes three glacial till soils with particle size distributions representing different initial conditions in the core material, e.g.: internally stable, internally unstable and critically internally unstable. The boundary conditions considered in the test program are: initial void ratio, type of filter and specimen size. Test were performed with three different ratio of increase of hydraulic gradient and three different time interval to increase the hydraulic gradient. Results show that the Rönnqvist (2015) adaptation of the Kenney and Lau (1985, 1986) method modified by Li and Fannin (2008) is an accurate empirical method to evaluate the susceptibility of glacial till soils to suffusion. It was also concluded that the critical hydraulic gradient triggering suffusion is not a unique value but depends on the testing conditions, such as axial loading, rate of increase of hydraulic gradient (Δi) and time interval to increase hydraulic gradient (Δt). The higher the axial load the higher the critical hydraulic gradient needed to initiate suffusion. High ratio of increase of hydraulic gradient applied with short time interval can lead to ic - values higher than the obtained in tests with low Δi and long Δt. Tests should be performed with low rate of increase of hydraulic gradient and each hydraulic gradient should last long enough for the specimen to adapt/respond to the new hydraulic conditions. Nevertheless, a general observation is that the critical hydraulic gradient of internally unstable soils tends to be lower than 5 when tested without axial load.