Abstract : Most of the Earth coasts recede and 80 % are rocky. Prediction of sea-cliff recession is essential to anticipate future risks for coastal development. However, it is difficult to understand this recession because many parameters control it. In addition, both the space and time scales are too big for the different mechanisms of cliff erosion to be fully analyzed. Experiments in a small-scale wave flume were conducted in which a massif made of wet sand is submitted to wave attack. The aim is to understand how cliff erosion is wave-controlled. The technique of shadow graph measurements was used to detect the time evolution of sand and water surfaces. We have analyzed the influence of wave forcing (F, ξ) (where F is the incident offshore wave energy flux and ξ is the surf similarity parameter) on the cliff recession rate and on collapse event size. The cliff recession rate increases linearly with the wave energy flux F. The eroded cliff materials change the bottom morphology; the types of bottom morphology strongly depend on the surf similarity parameter at the breaker point, or the Dean parameter Ω. Bottom profiles characterized by unsteady self-sustained sandbar oscillation were observed. In addition, we studied how sand granulometry change the system evolution. Finer the sand is, more cohesive is the cliff and bigger are cliff collapses. Contrary to what was expected, cliff recession is more important for finer sand: this could be due to a more dissipative bottom morphology built by fine sands. The sand volume within the system changes following cliff collapses and a sandbar removal during particular experiments. The cliff recession rate is constant when the sandbar is removed and decreases with cliff height. It seems that the unsteadiness of the bottom profile is activated when the volume of eroded sand exceeds a threshold value.