- Plant selection, energy intake, and digestibility were studied in nonrepro- ducing Gerbillus henleyi (10 g; n = 9), Gerbillus dasyurus (24 g; n = 6), and Meriones crassus (90 g; n = 9). These gerbils coexist in the Negev Desert Highlands, have similar dietary habits (being primarily granivores), and do not require drinking water, but they differ greatly in body mass. Each species was offered different levels (0% and -40% and 80% of maintenance energy requirements) of millet seeds and ad libitum fresh vegetation consisting of Atriplex halimus, Moricandia nitens, Zygophyllum dumosum, Lycium shawii, Anabasis articulate, and Thymelaea hirsuta, the dominant plants in the area inhabited by the rodents. Drinking water was not available. We hypothesized that: (1) smaller animals would be more dependent than larger animals on highly digestible food items such as seeds, and (2) smaller animals would be more selective than larger animals for highly digestible and energy-rich plant parts. The minimum levels of millet seeds needed to maintain body mass were 97%, 94%, and 36% of maintenance energy requirements for G. henleyi, G. dasyurus, and M. crassus, respectively. Therefore, G. henleyi was the most dependent on seeds, and M. crassus the least, supporting hypothesis 1. At similar levels of seeds consumed, apparent digestibilities of dry matter (DMD) and energy of the total diet were highest for G. henleyi. In the absence of seeds, vegetation DMD averaged 88% in G. henleyi and 70% in M. crassus, yet G. henleyi lost body mass, as a percentage, twice as fast as M. crassus. The three rodent species were selective in consuming vegetation and made similar choices. A. halimus and M. nitens were preferred, comprising -48%, 64%, and 67% of total plant dry matter consumption in G. henleyi, G. dasyurus, and M. crassus, respectively. Plant species not selected were relatively low in water and protein contents and were high in phenolic compounds and condensed tannins compared to preferred species. Gross energy, ash, and fiber levels did not appear to affect selectivity. Selection of plant parts within each plant species, as indicated by the leaf: stem (L:S) ratio, differed among rodent species. The L:S ratio for A. halimus was 4.8, but the L:S ratios ingested were 1.7, 4.0, and 7.4 in G. henleyi, G. dasyurus, and M. crassus, respectively, indicating selection of stems by G. henleyi, selection of leaves by M. crassus, and basically no selection by G. dasyurus. The ratio of seed: green vegetation (S:G) intake was highest in G. henleyi and lowest in M. crassus. A similar inverse rela- tionship between body mass and S:G ratio within seasons was found in other Old World and New World free-living granivorous desert rodents.