Quantitative information from electron spectroscopy for chemical analysis requires the use of suitable atomic sensitivity factors. An empirical set has been developed, based upon data from 135 compounds of 62 elements. Data upon which the factors are based are intensity ratios of spectral lines with F1s as a primary standard, value unity, and K2p3/2 as a secondary standard. The data were obtained on two instruments, the Physical Electronics 550 and the Varian IEE-15, two instruments that use electron retardation for scanning, with constant pass energy. The agreement in data from the two instruments on the same compounds is good. How closely the data can apply to instruments with input lens systems is not known. Calculated cross-section data plotted against binding energy on a log-log plot provide curves composed of simple linear segments for the strong lines: 1s, 2p3/2, 3d5/2 and 4f7/2. Similarly, the plots for the secondary lines, 2s, 3p3/2, 4d5/2 and 5d5/2, are shown to be composed of linear segments. Theoretical sensitivity factors relative to F1s should fall on similar curves, with minor correction for the combined energy dependence of instrumental transmission and mean free path. Experimental intensity ratios relative to F1s were plotted similarly, and best fit curves were calculated using the shapes of the theoretical curves as a guide. The intercepts of these best fit curves with appropriate binding energies provide sensitivity factors for the strong lines and the secondary lines for all of the elements except the rare earths and the first series of transition metals. For these elements the sensitivity factors are lower than expected, and variable, because of multi-electron processes that vary with chemical state. From the data it can be shown that many of the commonly-accepted calculated cross-section data must be significantly in error—as much as 40% in some cases for the strong lines, and far more than that for some of the secondary lines.