During 1972, a 2nd year of low herb production, there was a significant variation of seed reserves in the upper 2 cm of soil, with a doubling from Feb. to June and a density-independent halving by Oct. In Oct. 1972 there were 8 .times. 106 seeds/ha (5.3 kg/ha). After very high herb production in the spring, seed densities in Oct. 1973 were 10-16 times greater under shrubs and 23-27 times greater in exposed areas (max 187.5 .times. 106 seeds/ha, 84.3 kg/ha). The increase was principally by the winter annual grass, Festuca octoflora. In 1972, when there was a small difference in rodent density between 2 plots (0.8:1.0), there was no effect of rodents on seed density. In 1973, when there was a 1:17.8 ratio of rodent densities, there was a significant effect on seeds under shrubs. Then from Oct. 1973 to Oct. 1974, seed reserves in exposed areas between shrubs decreased by 20% in the plot with few rodents, and by 40% in the plot with many rodents. Eating of seeds by rodents accounted for 30 to 80% of the seed reserve decreases observed. Germination losses were no more than 25%. There is slight evidence that pocket mice selectively decrease abundance of the heavier species of seeds. Seed density was at least 5 times greater under shrubs than in exposed areas; density was significantly correlated with the size of shrub canopy. There was also a significant effect of the species of shrub on the density of seeds. During years of low production, under-shrub areas are a refuge for herb seed production, and in such years shrub seeds form a larger portion of the seed crop and new reserves. The difference in response of shrubs and herbs to weather increases the stability of seed reserves. In May 1973 the number of herbs per 100 seeds in the previous Oct. was 16.8 under shrubs and 43.6 in exposed areas; this implies a minimum germination of 24% of seeds over the whole habitat. It takes an exceptional coincidence of events, even in deserts, to cause a severe depletion of seed reserves.