Natural selection on codon usage is a pervasive force that acts on a large variety of prokaryotic and eukaryotic genomes. Despite this, obtaining reliable estimates of selection on codon usage has proved complicated, perhaps due to the fact that the selection coefficients involved are very small. In this work, a population genetics model is used to measure the strength of selected codon usage bias, S, in 10 eukaryotic genomes. It is shown that the strength of selection is closely linked to expression and that reliable estimates of selection coefficients can only be obtained for genes with very similar expression levels. We compare the strength of selected codon usage for orthologous genes across all 10 genomes classified according to expression categories. Fungi genomes present the largest S values (2.24-2.56), whereas multicellular invertebrate and plant genomes present more moderate values (0.61-1.91). The large mammalian genomes (human and mouse) show low S values (0.22-0.51) for the most highly expressed genes. This might not be evidence for selection in these organisms as the technique used here to estimate S does not properly account for nucleotide composition heterogeneity along such genomes. The relationship between estimated S values and empirical estimates of population size is presented here for the first time. It is shown, as theoretically expected, that population size has an important role in the operativity of translational selection.