A study of non-anchored tanks found in Switzerland (Koller, 2003) revealed that the plastic rotation of the shell-to-bottom connection in four of the six tanks studied did not satisfy the 0.2 radians limit established by the Eurocode (Eurocode-8, 2005). The fact that tanks could undergo fracture in their shell-to-bottom connection prompted further investigation with funding from Carbura and OFEV. The investigation focused on determining the real capacity of the shell-to-bottom connection since the current limit given by the Eurocode was based on a series of assumptions which had not been verified experimentally and omit essential parameters such as the number of cycles undergone by the connection and the ductility of the material used. 24 experiments were carried out to investigate connections with different thicknesses, condition (new vs. rusted) and with different levels of membrane loading. Experiments were performed under constant amplitude cycles ranged from 0 to maximum angle rotations of 0.2, 0.3 and 0.4 radians. The results from the experiments allowed the creation of plots of rotation vs. number of cycles to failure (called RN curves). These RN curves provide the number of cycles that may be resisted at a given rotation angle. Or, if the number of cycles is known (estimated), the maximum rotation the connection can resist may be obtained from the RN curve. The number of equivalent cycles (i.e., cycles with a rotation angle equal to the maximum rotation angle obtained during a given earthquake) expected in the connection was obtained from a dynamic analysis of two tanks representative of large capacity tanks found in Switzerland. The analysis was carried out in the finite element software ABAQUS. For the dynamic analysis, a suit of 8 ground motions form earthquakes having magnitude and distance according to the de-aggregation (for a 2500 return period) for the site of Sion, Switzerland and with soil conditions representative of type C and D soil (Eurocode-8, 2005) was used. From this analysis the maximum number of cycles was found to be 4. The estimated maximum number of cycles is conservative since the de-aggregation for the city of Sion, which has the highest earthquake hazard in Switzerland, was used for the analysis. From the RN curve (95% confidence interval) and the estimated number of cycles, the maximum permissible rotation in the connection would be 0.59 radians. However, since the maximum rotation angle applied during the tests was 0.4 radians, this rotation limit should not be exceeded. By using the limit of 4 cycles at 0.4 radians, rather than 0.59, the confidence interval around the regression line was increased from 95% to 99.999% (µ - 4.7σ). Results from this research suggest that the rotational limit of the shell-base connection of tanks may be increased to 0.4 radians based on the limited number of tests carried out. This limit is twice the original limit given by the Eurocode. It should be emphasized, however, that a number of aspects should be addressed before the suggested new limit can be recommended for design. A number of aspects shall be addressed to confirm the results obtained: - Additional tests at rotation angles of 0.2 and 0.4 radians should be carried out since only two tests were performed at these rotation angles. A minimum of two additional tests should be carried out at each level (0.2 and 0.4 radians). - New connections made of S235 steel grade shall be tested - The number of tests was limited to rotation angles of 0.2, 0.3 and 0.4 radians. Additional tests should be performed at greater rotations (e.g. 0.5, 0.6 radians). - Specimens from old tanks were limited to one tank (Konolfingen). Samples from two or three additional old tanks are recommended. - The dynamic analysis was limited to two tanks and eight ground motions. One or two more tanks should be studied. Additional ground motions should be used (12 to15 ground motions in total).