We tested the validity of the three Larson relations in a sample of 213 massive clumps selected from the Herschel Hi-GAL survey and combined with data from the MALT90 survey of 3mm emission lines. The clumps have been divided in 5 evolutionary stages to discuss the Larson relations also as function of evolution. We show that this ensemble does not follow the three Larson relations, regardless of clump evolutionary phase. A consequence of this breakdown is that the virial parameter $alpha_{vir}$ dependence with mass (and radius) is only a function of the gravitational energy, independent of the kinetic energy of the system, and $alpha_{vir}$ is not a good descriptor of clump dynamics. Our results suggest that clumps with clear signatures of infall motions are statistically indistinguishable from clumps with no such signatures. The observed non-thermal motions are not necessarily ascribed to turbulence acting to sustain the gravity, but they may be due to the gravitational collapse at the clump scales. This seems particularly true for the most massive (M$geq$1000 M$_{odot}$) clumps in the sample, where also exceptionally high magnetic fields may not be enough to stabilize the collapse.