Introduction: not only OB stars contribute to the ultraviolet emission which heats the dust. Non ionizing UV emission is also important in late B and Early A stars, especially in "normal" star forming galaxies and this UV emission is still related to the recent SFR over ~10^8 years, so relevant for the discussion. --I do say "much of the radiation" originates from OB stars; I don't say "all" of the radiation. Do you think I should say something different, like "OB and some early A stars"? Your reference list on the IR/UV analyses is quite large but perhaps can you add some new works which (I hope) shed a new light to this issue: Burgarella et al. 2005, MNRAS, 360, 1413 and Inoue et al. 2006, MNRAS 370, 380, which focused on the importance of the dust properties on the link between the UV slope and the IR/UV ratio. Another paper (Panuzzo et al. submitted) will be accepted very soon (I hope.., the revised version is submitted) and shows that geometrical distribution of stars and dust as well as the adopted IMF can also modify the irx-beta plot. I guess that the latter reference will be available before the acceptation of your paper, I can send a version to you as soon as it is accepted. As a consequence the star formation rate may not be the only factor which reddens the UV colors of (normal) galaxies selected in UV. --References added. section 4: I understand that your fit of the NIR stellar continuum with a curve obtained after 900 Myr of continuous star formation is a base line for comparison but when you say that you will analyse the deviations from this fit in UV and optical, one might understand that the scenario is good for the NIR which is probably not the case --True. I anticipate that the reader will understand this since the stellar curves are fitted to the 2MASS data. When you calculate the UV emission as nu*fnu(1500)+nu*fnu(2300) have you checked it is a good estimator of the UV emission? In my new paper on IR/UV comparison (G2 paper, submitted to ApJS, IR/Uv analysis of IRAS/GALEX samples, focused on luminosity functions) I have calculated that for a constant SFR over 100 Myr L(1200-3600A) = 0.8 * (nu*Lnu(1500)) and that for an extinction of 3 mag in FUV and a Calzetti attenuation law the relation becomes L(1200-3600A) = 1.3 * (nu*Lnu(1500)). Adding NUV and FUV luminosities may lead to a too large UV emission (but I did not make the exact calculation for this addition) --Alessandro raised a similar point. I use TIR/FUV in plots versus UV slope and H band luminosity (a new plot Alessandro suggested). I use TIR/FUV in these plots since they allow easy compares to figures in the literature. I also use FUV in computing the SFR. I decided to sum FUV and NUV in the other figures since I thought this would help to increase the UV S/N. A quick calculation shows a median value of 1.3 for (L_FUV + L_NUV) / [1.3 L_FUV]. In other words, FUV+NUV is ~1.3x larger than using L(1200-3600A) for 3 mags of FUV extinction. I do explicitly say "FUV+NUV" and not "TUV" in the plots. Probably because of my poor knowledge about the PCA I do not understand the conclusion of the analysis in 4.3 and the physical meaning of the eigenvectors, does it mean that the main parameters controlling the SED is the IR/UV ratio (what is the meaning of infrared-to-ultraviolet spectrum (e2) as compared to infrared to UV ratio (e1)?) --I have revised this section in an attempt to be more clear. section 5: 5.1 I'm glad to see that you find almost no effect with the inclination. I always found no or little effect of the inclination on the extinction without clearly understanding why. Do you have any idea? --My intuition tells me that it only shows up in a significant way when the galaxies have extreme disk inclinations. This is also consistent with the curvature displayed in the Tuffs et al. curve in Figure 11. 5.2 As you certainly know the overall variation of the IR/UV ratio as a function of the morphological type (excluding E-S0) is classically related to the heating by older stars as the galaxy type becomes earlier. Indeed it is found very easily when calculating an IR/UV ratio with different scenarii of star formation that for a given dust attenuation this ratio increases for quiescent galaxies (i.e. the figure 1 in Buat et al. 05, see also Kong et al. 2004). Perhaps can you add few words on this point, at least to be complete. --Done. Thanks. The case of early type galaxies is difficult: you have liners and seyferts and for them both the UV and the IR emissions are difficult to interpret. For the "star forming systems", the dispersion is very large given the low statistics... For irregular galaxies it is well known that extreme cases exist in the nearby universe (M82 has a very large extinction for example). I guess that the dispersion you find is also due to the way the SINGS sample has been selected and the results may well be different for a UV selection for example. Indeed in Buat Xu (1996, A&A 306, 61) with a selection close to a pure UV one we found a low mean extinction (i.e. IR/UV ratio) for all the late-type galaxies of our sample), only M82 was very discrepant. I suspect we have the same trend with our new GALEX samples. I realize it might be interesting to compare the trends according to the selection. I have to collect the types for my samples... --I hadn't thought of this before, about the dependence of IR/UV dispersion on sample selection. I would like to see such a comparison! 5.3 Is there a link with the morphological type in your figure 13, the y axis is related to it (cf figure 12) and I guess that some link also exists for the x axis? The figure 14 is much clearer and the effect obvious. Great and very new result! --The dependence with optical morphology in Figure 13 is not obvious. A question related to this very very interesting plot: what is the consequence on the calibration of the IR/UV ratio in terms of dust attenuation? In our calculations (i.e. Buat et al. 05) we found that a clumpy medium does not modify the calibration a result also found by Karl in 2000 (if I remember well). Is it still true? I guess that Karl has some ideas on the question. --Karl says: Yes. Clumpiness does affect the IR/UV ratio, but it does not affect the conversion of IR/UV to tau_eff. This finding was basically the main motivation for the Gordon et al. (2000) paper. Basically, if you rearrange the same amount of dust from a homogeneous distribution to a clumpy distribution, the IR/UV ratio will go down, but so will tau_eff. 5.4 Perhaps I am wrong but I do not see a very new result in this plot as compared to the previous GALEX (and pre-galex) studies. Don't you try to see where the galaxies dominated or not by a nuclear emission in IR are located in this diagramm? The trend found for "starburst galaxies" was obtained with IUE data on the central parts of the galaxies... --I included this plot because I simply thought such a plot would be expected from a SINGS UV+IR paper -- where do SINGS objects lie on such a famous diagram? I don't think it does present anything new in terms of normal galaxies, except for maybe that most of the deviations from the starburst curve are due to Sa types and earlier. To answer your question, I have added a plot to my paper webpage that displays TIR/FUV vs FUV/NUV with the symbols plotted according to the ratio of nuclear-to-total 24um emission. I do not see a clear trend with symbol size.