Interannual variations of the Antarctic ice sheet due to surface mass balance (SMB) fluctuations are important for mass balance estimates and interpretations. To date, these variations are primarily assessed by global or regional atmospheric modelling. Satellite altimetry and satellite gravimetry over the ice sheet provide complementary observations of the related volume and mass effects, respectively. Yet, so far the interannual signal contents of these observations have not been extensively studied. We compare and jointly interpret ENVISAT radar altimetry (RA) and GRACE satellite gravimetry results, relying on RA products from the along-track repeat satellite RA approach and on the GRACE 10-d solutions by CNES/GRGS. RA results and GRACE results are expressed in terms of variations of ice sheet thickness, Δz(t), and ice-equivalent thickness, Δzice(t), respectively. In view of the different errors and limitations of both techniques and of differences between Δz(t) and Δzice(t) expected due to firn-related processes, our principal approach is a comparison of qualitative patterns in space and time. To adjust the spatial resolution of both data sets, we describe the spatial filtering inherent to the regularization of the CNES/GRGS GRACE solutions and apply this filtering to the ENVISAT RA height changes in a consistent fashion. After correction for glacial isostatic adjustment, the spatial patterns of linear trends seen by ENVISAT RA and GRACE over the period 2002 October to 2009 August agree well, not only for the extreme ice losses in the West Antarctic Amundsen Sea Sector but also for an alternating sequence of gains and losses along the East Antarctic coast. Our main focus is on interannual signals, which we represent by the low-pass filtered non-linear, non-seasonal components of the Δz(t) and Δzice(t) time-series. These components should reflect interannual SMB variations, apart from effects of changes in ice flow. We find an agreement between the interannual variation patterns from ENVISAT RA and GRACE with temporal correlation coefficients typically in the order of 0.8. This qualitative agreement, prevailing even in the theoretical absence of a simple proportionality between Δz(t) and Δzice(t), indicates that both observational signals primarily originate in common geophysical variations. Combining both geodetic data sets aids their interpretation and promises to be valuable for the reduction of SMB uncertainties. In a case study based on the analysis of both GRACE and ENVISAT RA, we identify a prominent interannual feature in West Antarctica in 2005 September/October as an event of excess snow accumulation with a mass effect of 82 ± 31 Gt.