The central part of the portuguese continental passive margin is extensively studied by portuguese and french teams. On the shelf, many seismic lines have been realised and the data on submarine geological structure allows to reconstruct very precisely the history of sedimentation on the sea floor, and to link it with some phases of canyon incision. On the main land, DEM of the landforms have been constructed and some models of evolution have been proposed. On the shore line, a survey of paleo sea levels, including submarine sea stands, has permitted to establish a chronology os eustatic movements and of uplift rates. This set of data provides a good information about erosion rates, extended along various time span. The seismic profiles are used to calculate an average speed of retreat since the end of the Tortonian (7,1 millions years), because we can determine where the coast line was at this time. The average rate is 0,7 mm/year. The Eemian sea level, together with some Weichselian stands may be used to determine an average rate of coastal retreat for the last 120000 years. The rate is 5 mm/year. A very peculiar sea stand, at -7 m depth, is used to calculate a rate of retreat for a 1000 years long period, during the Holocene (5500-4500) with a very high rate of 10 mm/year. Some data is also given by present day rate of retreat, observed and measured between 1990 and 1994. The average is 1 mm/year with a great variability in space. The present day rate o retreat, when interpolated, doesn't fit with any of the other figures, neither does the holocene rate fits with the 120000 one. Though, all these rates could be found, today, on other sites of the Portuguese coast, in the same limestone material. They are commonly observed on exposed cliffs. We may propose that successive morphogenic systems, along the coast we study, have not been very different than the one we observe elsewhere today. A first discussion tries to expose why linear interpolation is impossible, in time, even if processes have remained more or less the same. A non linear model of evolution is then proposed. It relies on an arbitrary average speed of retreat which is calculated as an average during the longest know time span. If evolution were linear, all data should then be on one single straight curve. The difference between the theoretical position of data and its actual location away from the curve is used to calculate a threshold: it gives a measurement of the relative efficiency of morphogenic processes during a short period, compared with the average efficiency of all processes during long term. The originality of a morphogenic crisis may be demonstrated. Some explanations are proposed. A model is constructed and presents how sea level variations are related with the speed of coastal retreat. The speed of the sea level rise is much more important than the total amount of vertical displacement. Though, the landforms which are carved by this rise are not conserved in the morphology for very long (in this case: 100000 years). Most of the observed present landforms are a testimony of periods where erosion intensity was low. The question of morphological heritages is discussed. Tectonics are know to be quite important in this area. A comparison is attempted, but a precise quantification is at this stage of our work, not very reliable and we propose no explanation for the observed correlations. Our conclusion makes clear the difficulties in using todays figures for modelling the land forms evolution.