The primary signal for Drosophila sex determination is the number of X chromosomes relative to the number of sets of autosomes. The present report shows that the numerator of this X/A signal appears to be determined by the cumulative dose of a relatively limited number of discrete X-linked genetic elements, two of which are sisterless-a and sisterless-b. This discovery regarding the nature of the sex determination signal grew out of previous studies of both the likely X/A signal target (the feminizing switch gene, Sex-lethal) and two positive regulators of that target gene (sis-a and daughterless). Combinations of genetic perturbations in these three genes had been shown to have synergistic effects. A model proposed in part to account for these interactions generated a large variety of strong predictions for sex-specific synergistic interactions that would be diagnostic for X/A numerator elements and could distinguish them from other components of the sex determination system. All these predictions, as well as other predictions for X/A numerator elements, are shown here to be fulfilled. The most compelling observations involve sexually reciprocal viability effects of duplications of wild-type genes: combinations of sis-a+, sis-b+ and/or Sxl+ duplications are lethal to males but rescue females from the otherwise lethal effects of changes in other components of the sex determination machinery. The many interactions described here illustrate an important principle that may seem counter-intuitive: perturbations of the sex determination signal for Drosophila generally will not appear to affect adult sexual phenotype. This principle follows from the fact that Sxl is involved in dosage compensation as well as sex determination, and from important aspects of the nature and timing of Sxl's regulation both by the X/A signal and by Sxl's own products (positive autoregulation). These factors mask potential effects on adult sexual differentiation by causing the premature death of cells and/or individuals. The fact that the vast array of results presented here conform to this principle is strong evidence in favor of a "binary state" model for Sxl regulation by the X/A signal. This model is favored over an alternative "multiple state" hypothesis that was proposed by others in a different study of the X/A signal. In that same study it was concluded that region 3E8-4F11 of the X chromosome contained especially potent X/A numerator elements.(ABSTRACT TRUNCATED AT 400 WORDS)