We have estimated the source parameters of interplate earthquakes in an earthquake cluster off Kamaishi, NE Japan over two cycles of M~ 4.9 repeating earthquakes. The M~ 4.9 earthquake sequence is composed of nine events that occurred since 1957 which have a strong periodicity (5.5 ± 0.7 yr) and constant size (M4.9 ± 0.2), probably due to stable sliding around the source area (asperity). Using P- and S-wave traveltime differentials estimated from waveform cross-spectra, three M~ 4.9 main shocks and 50 accompanying microearthquakes (M1.5–3.6) from 1995 to 2008 were precisely relocated. The source sizes, stress drops and slip amounts for earthquakes of M2.4 or larger were also estimated from corner frequencies and seismic moments using simultaneous inversion of stacked spectral ratios. Relocation using the double-difference method shows that the slip area of the 2008 M~ 4.9 main shock is co-located with those of the 1995 and 2001 M~ 4.9 main shocks. Four groups of microearthquake clusters are located in and around the mainshock slip areas. Of these, two clusters are located at the deeper and shallower edge of the slip areas and most of these microearthquakes occurred repeatedly in the interseismic period. Two other clusters located near the centre of the mainshock source areas are not as active as the clusters near the edge. The occurrence of these earthquakes is limited to the latter half of the earthquake cycles of the M~ 4.9 main shock. Similar spatial and temporal features of microearthquake occurrence were seen for two other cycles before the 1995 M5.0 and 1990 M5.0 main shocks based on group identification by waveform similarities. Stress drops of microearthquakes are 3–11 MPa and are relatively constant within each group during the two earthquake cycles. The 2001 and 2008 M~ 4.9 earthquakes have larger stress drops of 41 and 27 MPa, respectively. These results show that the stress drop is probably determined by the fault properties and does not change much for earthquakes rupturing in the same area. The occurrence of microearthquakes in the interseismic period suggests the intrusion of aseismic slip, causing a loading of these patches. We also found that some earthquakes near the centre of the mainshock source area occurred just after the earthquakes at the deeper edge of the mainshock source area. These seismic activities probably indicate episodic aseismic slip migrating from the deeper regions in the mainshock asperity to its centre during interseismic periods. Comparison of the source parameters for the 2001 and 2008 main shocks shows that the seismic moments (1.04 x 1016 Nm and 1.12 x 1016 Nm for the 2008 and 2001 earthquakes, respectively) and source sizes (radius = 570 m and 540 m for the 2008 and 2001 earthquakes, respectively) are comparable. Based on careful phase identification and hypocentre relocation by constraining the hypocentres of other small earthquakes to their precisely located centroids, we found that the hypocentres of the 2001 and 2008 M~ 4.9 events are located in the southeastern part of the mainshock source area. This location does not correspond to either episodic slip area or hypocentres of small earthquakes that occurred during the earthquake cycle.