Regulation of gene expression is usually separated into cis and trans components. The separation may become artificial if much of the variation in expression is under multigenic and epistatic (e.g., cis-bytrans) control. There is hence a need to quantify the relative contribution of cis, trans, and cis-by-trans effects on expression divergence at different levels of evolution. To do so across the whole genome, we analyzed the full set of chromosome-substitution lines between the two behavioral races of Drosophila melanogaster. Our observations: (i) Only Ϸ3% of the genes with an expression difference are purely cis regulated. In fact, relatively few genes are governed by simple genetics because nearly 80% of expression differences are controlled by at least two chromosomes. (ii) For 14% of the genes, cis regulation does play a role but usually in conjunction with trans regulation. This joint action of cis and trans effects, either additive or epistatic, is referred to as inclusive cis effect. (iii) The percentage of genes with inclusive cis effect increases to 32% among genes that are strongly differentiated between the two races. (iv) We observed a nonrandom distribution of trans-acting factors, with a substantial deficit on the second chromosome. Between Drosophila racial groups, trans regulation of expression difference is extensive, and cis regulation often evolves in conjunction with trans effects.K nowledge of the genetics of complex traits is fundamental to modern medicine, agriculture, and evolutionary biology. Among all complex traits, gene expression as phenotype may be most amenable to genetic analysis. The first question about expression regulation naturally is whether there is a cis component and how strong the cis component is. (cis regulation refers to the control of expression by the gene itself whereas trans regulation refers to the influence of the genetic background.) Many studies have addressed this question at various levels of divergence (1-9). For example, one may measure the expressions of two alleles at the locus of interest in a common genetic background (usually F 1 s) (1, 2, 5). Because the collection of trans-acting factors in the same cellular environment is assumed to affect the two alleles equally, asymmetric allelic expression implies differences due to cis-regulatory divergence. Similarly, expression quantitative trait loci (eQTL) mapping permits inference of cis regulation if the eQTL is mapped closely to the expressed gene itself (6-10).A second question is how strong cis regulation is relative to trans regulation. In the extreme case where most expression variation is controlled by cis-trans interactions (e.g., joint actions of cis elements and transcription factors), the question would not be very meaningful because cis and trans components are both indispensable. It is desirable to explicitly model expression regulation to include cis, trans, and cis-by-trans control. Many kinds of data allow such explicit modeling. The use of large numbers of recombinant strains for express...