This thesis aims to analyse the conservatism in the current Dutch recommendations on the external friction angle 𝛿 of diaphragm walls: 𝛿 = 𝑚𝑖𝑛[𝜙, 20°] (for curved slip surfaces) (CUR/COB, 2010), by means of an experimental investigation. Based on literature, two sources of conservatism are identified: 1) filter cake contamination by excavated soil particles is not taken into account, which can lead to an increased 𝛿 of around 30° (Day et al., 1981; Henry et al., 1998; Arwanitaki et al., 2007); 2) a continuous shear plane through the filter cake is assumed, ignoring the influence of concrete roughness and filter cake thickness. Cernak et al. (1973) and Lam et al. (2014) found that the concrete roughness causes an increased shear strength compared to the filter cake below a certain filtration time (which controls the filter cake thickness). These aspects are captured in a conceptual model (hypothesis), which predicts the development of interface shear strength as a function of filtration time for the case of a contaminated filter cake compared to a clean filter cake: For a contaminated filter cake, the lower boundary shear strength, which is that of the filter cake, is higher compared to that of a clean filter cake. However, this lower boundary is reached after a shorter filtration time for a contaminated filter cake compared to a clean filter cake, caused by the higher filter cake growth rate due to slurry contamination. The first experimental phase focusses on the filter cake shear strength. A series of small-scale direct-shear tests (∅ 67 mm) on sand/filter cake (clean)/cement-mortar samples show a linear trend of peak shear strength of 18,3° for normal pressures in the range of 200 kPa - 400 kPa at a shear rate of 1,2 mm/min. This result corresponds to the friction angle of 19,5° by Deltares (2008), which lies at the basis of the current Dutch 𝛿 recommendations. Additional tests on clean filter cakes indicate a peak shear strength around 23° at a shear rate of 0,0072 mm/min. Test series on contaminated filter cakes show an increased linear trend of peak shear strength of 25,6° for normal pressures in the range of 200 kPa - 400 kPa at a shear rate of 1,2 mm/min. The filter cake contamination is based on an analysis of an in-situ slurry sample analysis from the Spoorzone Delft project. Increased friction angles of 28,5° and 32,8° are observed at 0,0072 mm/min shear rate. In the second experimental phase, the influence of filtration time on the interface shear strength for clean and contaminated filter cakes is investigated. In total, 7 direct-shear tests have been performed on sand/filter cake/concrete samples (170 mm x 170 mm), in which a realistic diaphragm wall concrete mix is applied (max. aggregate size 16 mm, class F5). The test results (normal pressure of 200 kPa, shear rate 1,2 mm/min) show a decrease of interface shear strength with increased filtration time for both clean and contaminated filter cake samples and it is shown that for the contaminated case the filter cake shear strength is reached after a shorter filtration time (around 12 hours) compared to clean filter cakes (around 24 hours), confirming the hypothesis. An analysis of the concrete surface textures (based on 3D laser scans) indicates an increased surface roughness with increased filter cake thickness. The obtained concrete surface roughness is comparable to in-situ conditions based on an analysis of in-situ data from the Spoorzone Delft project. In this comparison the influence of macro-roughness patterns is not taken into account. The experimental results of this thesis indicate that the main source of conservatism of the current Dutch recommendations on 𝛿 is the omission of the influence of filter cake contamination on the filter cake shear strength. In addition, 𝛿 could further be optimised as a function of filtration time. However, the time frame of increased interface shear strength from the filter cake shear strength towards 𝜙 is limited for contaminated filter cakes (around 12 hours in this research). It is therefore concluded that the recommended 𝛿 value(s) can best be based on contaminated filter cake shear strength, omitting the time effect. It is therefore suggested that future research should focus on the lower boundary shear strength for the case of in-situ filter cake samples. For this purpose the developed direct-shear set-up can be applied. It is also suggested to further investigate the influence of macro-roughness patterns, since it can not be excluded that macro-roughness patterns do not cause an extended time frame of increased interface shear strength.