Partial melting and melt extraction in the Earth's upper mantle can significantly modify the composition, and therefore the thermophysical properties, of the solid residues (restites). Seismic velocities have been by far the most used observable to infer such melt depletion anomalies in the lithospheric mantle. However, three fundamental issues that make such interpretations dubious are: i) previous studies addressing melt depletion effects have been based on isobaric batch (equilibrium) melting experiments at near-solidus temperatures instead of more realistic polybaric near-fractional models at different equilibration pressures, ii) the use of indicators such as VP/VS is strongly influenced by anelastic attenuation and phase stability, and iii) compositional effects tend to be of the order of uncertainties in seismic models.This paper presents the results of a systematic exploration of all these effects on seismic velocities in peridotites. Our results show that polybaric vs isobaric models of partial melting result in similar VP and VS trends with degree of melting (F) when they re-equilibrate at sub-solidus temperatures, but they differ from previous parameterizations based on isobaric experiments at near-solidus temperatures. For deep melting paths (i.e. high potential temperature), solid residues show maximum VP and VS increases of ~1% and 1.7%with respect to their fertile counterparts. Shallow melting paths produce significantly smaller effects, even at F~0.25. Moreover, in the case of VP, these effects are not linear and ∂VP/∂F can be either positive or negative depending on the melting path. We find that while the ∂(VPVS -1)/∂F of residues is strongly dependent on the melting history, ∂(ΡVS -1)/∂F remains nearly constant regardless of the melting path or extent of melting. This confirms that Ρ/VS is not only a reliable indicator of compositional variations in natural samples, but also of degree of melting. This indicator could therefore be used to inform joint inversions of gravity and seismic velocities to test plausible depletion effects in the lithospheric mantle.