The aim of this prospective study was to compare implants placed in grafted and normal non-grafted maxilla by means of resonance frequency analysis (RFA), clinical stability and implant failure. Twenty-nine patients with severe atrophy of the edentulous maxilla were treated with autogenous bone grafts as onlay (24 patients) or as interpositional grafts in conjunction with a Le Fort I osteotomy (five patients) 6 months prior to placement of 222 implants. Ten non-grafted patients treated with 75 Brånemark implants in the edentulous maxillae served as a control group. RFA was performed at implant placement, abutment connection and after 6 months of bridge loading. Seventeen (8%) implants were lost in the grafted bone and one (1%) in normal bone. RFA revealed a similar pattern in both grafted and normal maxillae, i.e. increasing resonance frequency (RF) with time (Wilcoxon Signed Rank test for paired data). Twenty implants that were rotation mobile (low primary stability) at the time of insertion showed a significantly lower value at implant placement according to RFA (Mann-Whitney U-test, P = 0.020). The RF for the failed implants revealed a tendency towards lower values (Mann-Whitney U-test, P = 0.072), compared to the successful implants. It is concluded that implants placed in grafted bone when using a two-stage technique achieve a stability similar to that of implants placed in normal non-grafted bone.