It has become familiar to interpret recent observations of high-redshift sources via the comoving volume-averaged history of star formation. This is the mean over cosmic time of the stochastic, possibly short-lived star formation episodes of individual galaxies, and follows a relatively simple dependence on redshift. Its latest version, uncorrected for dust extinction, is plotted in Figure 2 (left). The measurements are based upon the rest-frame UV luminosity function (at 1500 and 2800 Å), assumed to be from young stellar populations . The prescription for a ``correct'' de-reddening of these values has been the subject of an ongoing debate. Dust may play a role in obscuring the UV continuum of Canada-France Reshift Survey (CFRS, 0.3 < z < 1) and Lyman-break (z 3) galaxies, as their colors are too red to be fitted with an evolving stellar population and a Salpeter initial mass function (IMF) . The fiducial model of  had an upward correction factor of 1.4 at 2800 Å, and 2.1 at 1500 Å. Much larger corrections have been argued for by  (x 10 at z = 1),  (x 15 at z = 3), and  (x 16 at z > 2). As noted already by  and , a consequence of such large extinction values is the possible overproduction of metals and red light at low redshifts. Most recently, the evidence for more moderate extinction corrections has included measurements of star-formation rates (SFR) from Balmer lines by  (x 2 at z = 0.2),  (x 3.1 ± 0.4 at z = 1), and  (x 2-6 at z = 3). ISO follow-up of CFRS fields  has shown that the star-formation density derived by FIR fluxes (x 2.3 ± 0.7 at 0 z 1) is about 3.5 times lower than in . Figure 2 (right) depicts an extinction-corrected (with A1500 = 1.2 mag, 0.4 mag higher than in ) version of the same plot. The best-fit cosmic star formation history (shown by the dashed-line) with such a universal correction produces a total EBL of 37 n W m-2 sr-1. About 65% of this is radiated in the UV + optical + near-IR between 0.1 and 5 µm; the total amount of starlight that is absorbed by dust and reprocessed in the far-IR is 13 n W m-2 sr-1. Because of the uncertainties associated with the incompleteness of the data sets, photometric redshift technique, dust reddening, and UV-to-SFR conversion, these numbers are only meant to be indicative. On the other hand, this very simple model is not in obvious disagreement with any of the observations, and is able, in particular, to provide a reasonable estimate of the galaxy optical and near-IR luminosity density.
Figure 2. Left: Mean comoving density of star formation as a function of cosmic time. The data points with error bars have been inferred from the UV-continuum luminosity densities of  (filled dots),  (filled squares),  (filled pentagons),  (empty dot), and  (empty square). The dotted lin shows the fiducial rate, < * > = 0.054 M yr-1 Mpc-3, required to generate the observed EBL. Right: dust corrected values (A1500 = 1.2 mag, SMC-type dust in a foreground screen). The H determinations of , , and  (filled triangles), together with the SCUBA lower limit  (empty pentagon) have been added for comparison.