The main objective of this research is to improve the comprehension of the hydrological behaviour of natural catchments. This study is based essentially on the observation and the measurement of hydrological processes on the Haute-Mentue experimental research basin. The main originality of this work is to associate different types of measurements in order to obtain a better vision of hydrological processes responsible for streamflow generation. First the hydrological behaviour is studied at the catchment scale by the application of environmental tracing. Then hillslope measurements are conducted so as to identify the form of water flows through the catchment. Before the application of the environmental tracing technique on the Haute-Mentue, an uncertainty analysis of hydrograph separation models is proposed. In order to improve their interpretation in terms of hydrological processes, it is essential to estimate systematically their uncertainty. Two types of uncertainty are distinguished: the model uncertainty, which is affected by model assumptions, and the statistical uncertainty, which is due to the temporal variability of the chemical tracer concentrations of the components. The statistical uncertainty is studied using a Monte Carlo procedure. The model uncertainty is investigated by the comparison of alternative hypotheses concerning the chemical concentration of mixing model components and their spatio-temporal variability. This analysis validates in a way the use of hydrograph separation for the study of catchment hydrological behaviour. In fact it indicates that despite the uncertainty, the flow sources which generate the stream flow are clearly identified. However, the precision and the coherence of hydrograph separations can be improved by taking account of the temporal and spatial variability of component chemical concentrations. After this analysis, hydrographs recorded at the outlets of four Haute-Mentue sub-basins are separated by the application of a three-component mixing model based on the silica and calcium concentrations of water. This model allows the distinction of the following three components: direct precipitation, soil water and groundwater. The analysis of hydrograph separations of these four sub-basins allows the study of the spatial and temporal variability of hydrological responses. Concerning the spatial variability, despite the uncertainty of hydrograph separation, it appears that the hydrological behaviour of each sub-basin is different. The spatial variability of hydrological responses, which is observed essentially at the soil water contribution level, seems mainly due to the catchment morphology and geology. In general, soil contribution decreases with the steepness of the soil and the permeability of geological formation. As for the temporal variability of hydrological responses, it seems essentially to be controlled by the catchment humidity. Soil contribution increases with increasing basin humidity. A statistical analysis based on the development of multiple linear regressions allows the confirmation of the influence of these physico-climatic factors on the spatio-temporal variability of hydrological responses. In order to discover the processes responsible for the important soil water contributions highlighted by the application of environmental tracing, two experiments, one with a TDR system and the other with a rainfall simulator, were conducted within the catchment. On the whole, these two experiments indicate that the water flow processes are spatially quite heterogeneous and depend on very local properties. In fact, a geostatistical analysis of TDR measurements shows that the spatial variability of soil water content is very important at a local scale (1-20 meters). The soil infiltration capacity, determined by the simulation of artificial rainfall, is also spatially quite variable. This variability depends on the physical properties of the soil and geology but also on very local properties as macropores. After these different experiments, all of the observations done at the local scale as well as at the catchment scale are combined in order to get the maximum information about hydrological processes which are at the origin of streamflow. First, the TDR measurements partially confirm the environmental tracing results indicating that the soil water is an important potential source of subsurface flows. In view of the important spatial variability of soil water content and more particularly of the drainage dynamics, it seems that the stream flow generation is essentially controlled by processes at the local scale. The experiment with the rainfall simulator confirms this observation showing that the capacity of soil to transmit a water flow varies strongly in space. According to this experiment it seems, in the particular case of the Haute-Mentue, that soil water flows contributing to the stormflow generation moves essentially through the macropore system. In order to identify the form of subsurface flows and indirectly to confirm the role of macropores, an artificial tracing experiment was conducted. The transit speeds highlighted during this experiment are quick enough to conclude that the water contained in hillslopes contributes to the stromflow generation. Considering the physical and hydrodynamic properties of soils, only preferential flows can explain these transit speeds. In fact, from this experiment, it is not possible to know the nature of these preferential flows, but according to previous results it seems to be macropore flows. Finally, according to the information collected in this study, a conceptual model is proposed in order to explain the hydrological behaviour of the Haute-Mentue. This model, which is based essentially on the extension of contributing areas and on the role of preferential flows, allows to explain in terms of mechanisms the hydrological and chemical responses highlighted by hydrograph decompositions while respecting observations done at the local scale.