Transient aseismic slip events (that do not emit elastic waves) were first discovered on the San Andreas fault in central California in 1960 (Steinbrugge et al., 1960, Tocher 1960) and were later confirmed by the development and installation of GPS stations. Aseismic slip can occur on continental faults as well as on subduction zones. However, the interactions between aseismic and seismic slips are not fully captured yet. Understanding the mechanisms at stake on fault planes is fundamental since several large earthquakes were preceded by aseismic slip episodes.For example, the setting of a slow slip event before the 1999 Izmit earthquake in Turkey was evidenced by the observation of repeating signals at one station (Bouchon et al., 2011). Also, by studying repeating earthquakes before the 2011 Tohoku earthquake, two slow slip episodes that migrated towards the rupture intitiation were observed (Kato et al., 2012). Other transient slips preceding large earthquakes occurred, including in Mexico, before the M7.3 Papanoa earthquake (Radiguet et al., 2016) and in Chile, before the 2014 Iquique earthquake (Ruiz et al., 2014).The aim of this thesis is to better characterize the interactions between seismic and aseismic slip that can occur in a subduction zone largely instrumented: Japan.The very high density of the japanese monitoring networks(seismic networks, both at short and long periods, and GPS network), allows the detection ofseismic and aseismic slip events with low intensity / size, and is amenable to the use of arraymethods to improve the signal-to-noise ratio. The search for episodes of transient deformation will be based on several observables andtechniques, and is complemented by the analysis of the changes in earthquake rates concomittent to these episodes. This work thus gives new clues on how aseismic deformation can -or not- accelerate the occurrence of strong earthquakes, and on what are the conditions (tectonic, dynamic) for this to happen.