We studied colonization of annual plants in small-scale disturbances and undisturbed soil for four years in the northern Negev desert. The experiment consisted of 24 patches of I m2: eight undisturbed patches, eight 30 cm deep pits and eight 20 cm high mounds. Disturbance removed the seed bank from the pits and mounds. Rainfall was average during 1992 and 1993 (163.5 and 157.0 mm), very low in 1994 (97.5 mm) and very high in 1995 (283.0 mm). For all ca 100 species in the assemblage, we measured abundance, as the average local density in patches occupied by the species, and incidence, the proportion of patches occupied. Abundance and incidence were positively correlated in the four years of the study (R2-= 0.71, 0.56. 0. 56 and 0.64). The relationship became steeper during the course of colonization. A minority of species colonized all patches rapidly and increased exponentially in abundance. These high incidence and abundance species responded little or not at all to variations in rainfall. Most of the species experienced frequent local extinctions and colonizations in few patches. These low abundance and low to medium incidence species responded strongly to rainfall by fluctuations in abundance. Differences per species between abundance and/or incidence in pits, mounds and matrix were evident in many species along the entire relationship. These were mostly in favor of pits. We propose that the positions and trajectories of the individual species in the abundance-incidence phase plane signify constraints on population growth along the abundance axis, and on colonization and population persistence along the incidence axis. Since species with lower incidence and abundance respond strongly to rainfall variation, we conclude that their local populations and spatial distribution are to a large extent site-limited. High-abundance and high-incidence species have few or weak constraints and experience density-independent population growth, which is only seed-limited. The causes of site limitation and its effect on population growth and on eotonization and extinction processes differ for different species. In a number of species, abundance and incidence may be limited by dispersal, in conjunction with dispersal mode and patch-specific seed capture. Based on our case study we suggest that the trajectories of the species in the abundance-incidence phase plane are a useful tool for investigating assemblage dynamics.