The problem with the answer given in both of the other two responses (so far) is that they claim that water has the unusual property of having transmitted red light filtered/absorbed so that the transmitted light is blue. I have zero problem with that. The authors of the J.Chem Ed. piece go on to talk about the looking at reflected light from a Colorado lake or the Caribbean. There, you have to be careful, reflected light is clearly not the same as transmitted light and the inclusion of those scenes is a distraction. The two tubes are clearly indicated to be left = water, right = air (due to the cost of D2O being above their budget). I also have zero problem with the assertion that the visible absorption spectra of heavy water is, essentially, flat (although I'd prefer a higher resolution spectra as well as a baseline which is level). (This statement ignores non-linear optical effects, which is quite reasonable.) However there's a clear difference between saying there's little absorbance and that the heavy water is colorless. Scattering will occur. Even in ultra-pure D2O, scattering will occur. see Rayleigh Scattering. It is inherent in the electrical nature of matter. So, blue wavelengths will be more scattered than the reds, regardless of other specific interactions. Heavy Water is colorless and the sky is not blue. If you're ok with that statement, then ok, we're done but perhaps you should take a look outside on a clear day, if you are lucky enough to live somewhere where we haven't fouled the air with pollution. Now, it takes miles of sky, I'm not sure how much liquid water it would take to have a noticeable scattering effect. Anyway, a reductive approach to the question of color should, sooner or later synthesize all of the various causes of color back into the (gesalt) whole. If X is a function of A,B,& C, then showing that one material's X is dominated by A while another material's X isn't doesn't say squat about what X is for the second material. You need to include all contributions. For what it's worth, being completely colorless would be a balancing act amongst all the factors (A, B, C, ..) and would be quite unusual. We usually just consider the color to be relevant to some expected optical path length, so color is typically context dependent. Reflected, transmitted, short path length, long, white light, colored light, etc., etc.