||The use of New Austrian Tunnelling Method has become more popular in recent years. The construction has been observed to cause significant influence on the stress regime and ground deformations around the tunnel. However, current understandings on the construction of the NATM twin tunnel are still limited. The aims of the current research are to investigate the construction of the NATM tunnel in soft ground and to study factors that govern the ground responses associated with the construction. The study is done by performing a series of three-dimensional (3D) numerical analyses. Computed results are compared with field measurements, empirical solutions as well as two-dimensional (2D) numerical analyses results. Different construction sequences have significant effects on the stress re-distribution of the ground, resulting in different magnitude of ground deformations, magnitudes and distributions of lining thrusts. In all the construction sequences analysed in this research, stress paths show that substantial yielding of soil happens at the invert due to high lateral stress at the invert, which caused the collapse of the Heathrow Express Tunnel. Moreover, the length of the unsupported span (LU) is an important parameter in tunnelling in a way that significantly different ground surface settlements and lining thrusts are resulted. The behaviour of the single tunnel construction is investigated three-dimensionally and compared with the results of 2D analyses. Results show that ground surface settlement profiles can be estimated reasonably well using the method of nodal force reduction. In additions, the amount of nodal force reduction required for matching 3D analysis results is related to LU and the initial Ko conditions. Nevertheless, the distribution of lining thrust resulting from 2D and 3D analyses are very different. Construction of two parallel tunnels has been modelled three-dimensionally. Load redistribution mechanism of parallel tunnels depends on the lagged distance of the tunnel (LT). For small LT (LT = 0), the loadings associated with twin tunnel construction are shared by the ground in the vicinity of both tunnels and the shotcrete lining. Stress paths show that the longer the LT, the more is the load attracted to the leading tunnel, resulting in different proportions of load sharing. Due to different load redistribution mechanisms, the bending moment in lining varies with LT. For the leading tunnel, the bending moment in the lining increases with LT, especially at the invert regions. On the other hand, it is seen that the maximum bending moment of the lagging tunnel decreases with LT. Furthermore, the resulting maximum ground settlement and the offset of the settlement trough varies with the LT. The maximum ground surface settlement shows a slightly decreasing trend because increasing amount of loading associated with the excavation of a neighbor parallel tunnel is transferred to the nearby lining of the leading tunnel with larger LT. As a result, at the end of the construction, the ground settlement trough is shifted towards the lagging tunnel. The offset increases with the lagged distance of the tunnel. Moreover, simple superposition of the single tunnel solutions is an inappropriate way to estimate the solution of the twin tunnel settlement profile, which underestimates the results from a twin tunnel analyses. On the contrary, LT does not seem to have significant effect on the excess pore water pressure at equilibrium.