In the construction of roads or railways, the capping layer is the last layer of the earthworks phase. This layer can be made of non bound aggregates or by a treatment of a soil with lime and/or hydraulic binder, such as cement or hydraulic road binders. In the latter case, the in situ testing of the capping layer performances should encompass treated soil modulus measurements, but sampling such materials is not often satisfactory, because of the risk of material degradation by the sampling itself. Consequently, a non destructive method, aiming at measuring the modulus of the treated materials, could be very useful. For this reason, we propose to study the feasibility of the seismic surface or guided waves dispersion analysis in order to recover the depth and the S wave velocity of the subgrade. Previous works provided results and analysis of the dispersion curves concerning the pavement auscultation (Ryden et al, 2004). However, in these cases, the subgrade was an underlying layer in the global zone of interest that includes the upper pavement layers where the measurement surface is the thin asphalt layer. In the present study, we focus on the subgrade layer in the case of under construction roads, before the shallower pavement layers are built because it should help to qualify the project acceptance concerning this earthworks phase. In this context, the issue deals with a two layers medium case where the investigated subgrade, whom the top is the measurement surface, lays above a low velocity zone, i.e. the natural soil. As described by Ryden et al. (2004), the resulting dispersion curves in the case of a high velocity upper layer should be typical of Lamb waves dispersion curves and could bring out higher modes that could be difficult to pick. In this case, they advocated the entire dispersion diagram inversion to avoid any subjective picking in the data (Ryden et al, 2006). However, the treated soil can contain heterogeneities unfavourable to the assumption of homogeneous layers presupposed to the dispersion diagram calculation. Thus a first feasibility stage, i.e. a field experimental data acquisition, was conducted to define the ability of seismic data to provide a coherent dispersion diagram in the spectral content required. The dispersion curve have been extracted and inverted with an iterative weighted least squares local minimization method (Hermann, 2002). In order to consider the possibility of inverting the entire dispersion diagram, a second feasibility stage consisted in analysing all the events that possibly occur in the dispersion diagram. For that, the measurement experience was reproduced at reduced scale in laboratory as a perfectly controlled experimental modelling approach. These data and more precisely the dispersion diagram is compared in one hand to the theoretical curves associated to the leakage attenuations and in an other hand to the theoretical dispersion diagram numerically calculated with an original method taking into account the source effects.