Various methods have been developed for direct observation of soil erosion in a watershed but rarely, they are truly satisfactory. Soil erosion can be observed using volumetric or dynamics methods. Volumetric methods such as erosion pins are simple but not sensitive. Improved volumetric methods such as contour plotting frame and laser scanner are more sensitive but they are also expensive and limited to very small sized plots. The major drawback of the volumetric methods, however, is their inability to provide samples for chemical or physical analyses. The dynamic methods, such as the runoff plots, are sensitive and providing samples for analyses. Runoff plots, however, are not really applicable to watershed studies under truly natural conditions. A simple mesh-bag method was developed sometime ago to assess soil erosion in field conditions. The mesh-bag method involves deployment of small mesh bags (e.g. 20x20 cm) in close contact with soil surface in plots of flexible sizes and shapes. After one or more rainfall events, the soil particles rested on and in the bags are collected for analysis. The principle behind the mesh-bag method is quite different from the traditional runoff plots in that it does not trap eroded soil but sample the redistribution of eroded soil after one or more rainfall events. The mesh-bag method, however, has not been compared with a traditional method or extensively studied. For example, what does the information obtained from the mesh-bag method tell us about soil erosion in a hill slope? How do we apply the method and analyze data properly under a given field condition? This paper discusses the principle, the proper use and misuse of the mesh-bag method with examples.