Breeding for drought tolerance is one of the main concern of rice breeding. This investigation aimed to evaluated some rice (Oryza sativa L.) genotypes under in vitro conditions and to study their genetic behavior under three different concentrations of PEG as a tool to asses their drought tolerance. The genetic materials used in this investigation were four parental varieties named and their six F₁ hybrids, which were produced from half diallel crosses mating design. The parental varieties were Giza 177, Sakha 101, Sakha 104 and IRAT 112. Results revealed that genotypic mean squares were highly significant for all studied in vitro traits, indicating the presence of real differences among genotypes. Furthermore, the variation due to PEG levels and genotypes x levels interactions were also highly significant for all in vitro traits except for green point percentage. This indicates that these genotypes gave different responses at different PEG levels. The magnitude of general combining ability (GCA) effects for each variety indicated that the varieties Sakha 104 and Sakha 101 were the best, while Giza 177 and IRAT 112 were inferior for studied traits. The crosses Sakha 104 x IRAT 112 and Sakha 101 x IRAT 112 were the best combinations for all in vitro studied traits except for green point percentage. The magnitudes of dominance genetic variance were larger than their corresponding variances of additive for all in vitro traits, indicating that the major role of dominance genetic variance in the inheritance of these traits. The results were emphasized by dominance degree ratio, which were more than unity for all studied traits, revealing the importance of over dominance in the genetic expression of these traits. It could be concluded that most of studied in vitro traits are mainly controlled by non-additive genes in addition to the minor role of additive gene action under drought stress levels. Thus, it could be suggested that one of the best methods of breeding to improve these genotypes for tolerance to drought is the recurrent selection program.