Selection Bias

In Cosmology by Brian Koberlein12 Comments

There’s a new paper in the International Journal of Modern Physics which presents evidence that the universe is not expanding. You heard that right. If true it would overturn decades of cosmological theory. It’s the kind of revolutionary find that wins Nobel prizes. It’s gotten a bit of attention in the popular press, but don’t throw your old astronomy books out just yet.

The paper focuses on a cosmological test known as the Tolman test.  It was first proposed in the 1930s by Richard C. Tolman, and it is done by comparing the surface brightness (brightness per area) galaxies to their redshifts. If the universe were perfectly static, then the surface brightness of a distant galaxy should be the same as that of a close galaxy.  This is because while distant galaxies appear dimmer, they would also appear smaller by the same amount, since apparent brightness and apparent area both follow an inverse square relation. Of course, we know that most galaxies have redshifts, and this would serve to dim a galaxy even further. But it turns out that the amount of extra dimming is proportional to the redshift. So for a static universe with redshift, the ratio of surface brightness to redshift should be constant.

The Tolman test has been done several times before, with thousands of galaxies, and the results have agreed with the expanding universe models. So what’s different about this paper? It doesn’t look at a traditional static universe where redshift is due to the motion of galaxies relative to us, but rather an alternative static universe where light “ages” by some unknown mechanism so that it redshifts over time.  In this way, distant galaxies have a redshift due to their “tired light” rather than relative motion or cosmic expansion. For such a model, the surface brightness should dim proportional to redshift.

So the authors compare the surface brightness and redshifts of both low redshift and high redshift spiral galaxies, and find that the results agree with their static tired light model (what they call the SEU model).  They then conclude that while the result is not in itself sufficient to overturn the standard expanding universe model, it shows that more research should be done on the SEU model.  Hence the headlines “Universe is Not Expanding After All, Scientists Say”.

But not so fast.

If you’re going to test an alternative model that requires the introduction of some unknown mechanism for redshift, you should probably compare your results to an expanding universe model to see if yours works better. In particular, you should probably compare your data to the ΛCDM model (the standard dark energy/dark matter/expanding universe).  Do they do this? No.  In their own words, “In this paper, we do not compare data to the ΛCDM model. We only remark that any effort to fit such data to ΛCDM requires hypothesizing a size evolution of galaxies with z.”  Apparently hypothesizing a size evolution for galaxies is bad, but introducing an unknown tired light mechanism to preserve a static universe is okay.

There’s another reason they didn’t compare their results to ΛCDM, and that’s because it would match the data equally well.  Such comparisons have been done before, and they support ΛCDM.  In fact, a paper earlier this year used a similar cosmological test known as the Alcock-Paczynski test, and compared the results to a range of models. The test eliminated all but two models, ΛCDM and static tired light.

Of course tired light is excluded by other observational evidence. If we lived in a static tired-light universe, distant galaxies should appear blurred (they don’t), distant supernovae shouldn’t be time-dilated by cosmic expansion (they are), and we shouldn’t see a cosmic microwave background (we do).

So does this paper show that the universe might be static? No. They didn’t compare the results to a range of static and expanding universe models to show the static model works better. They assumed a fringe static model specifically designed to match the standard expanding model.  They even state this in their paper: “In this paper we reconsider this subject by adopting a static Euclidean Universe with a linear Hubble relation … resulting in a relation between flux and luminosity that is virtually indistinguishable from the one used for ΛCDM models.” In other words, they chose the tired light static universe model because it matches the standard expanding universe model for the Tolman test.

This is not how science is done.


Paper: Eric J. Lerner et al. UV surface brightness of galaxies from the local Universe to z ~ 5. Int. J. Mod. Phys. D, (2014). doi: 10.1142/S0218271814500588

Paper: López-Corredoira M. ALCOCK-PACZYŃSKI COSMOLOGICAL TEST. ApJ. 2014;781(2):96-.


  1. Dear Brian

    the conclusion of your post “This is not how science is done.” is completely unjustified. The paper you are commenting on, did present the analysis of a set of data (real observations not simulations) and discuss then on the contest of a specific model. The reason for not discussing the same data within the framework of LCDM is that, far from fitting the data, LCDM supporters simply ASSUME galaxies follow whatever evolution on both size and luminosity is needed to describe the data. Modern cosmology ASSUMEs the existence of dark matter, dark energy, and inflation just to make sense of observations. While in this paper the authors assume light is loosing energy over time (choice imposed by the existence of redshift), and implicitly that the cosmic microwave background is a local phenomena (unrelated to the big bang), nothing else.

    By the way, to be fair, one should have mentioned that the time stretching is NOT observed when using QSO instead of supernovae, casting serious doubt on that result.
    Also, the paper present a large section devoted to re-analyse the data used in previous implementation of the Tolman test showing that in ALL cases they agree with the static Universe model.

    Finally, you present the choice of the flux-luminosity relation made by the authors as an ad hoc assumption to mimic LCDM. This is just the opposite of what the authors have done. They assumed the Universe is Euclidean and static, because it is so here on the solar system, and a linear Hubble relation at all redshifts, because it is so locally. The flux-luminosity relation is then a necessary consequence of these two assumption (with no fudging factors). What is really strange is that modern cosmology after several adjustment of their relation (adding dark matter and dark energy at will) ended up with the Euclidean one!

    So, who is doing science right?

    1. Author

      My comment was completely justified. The data fits both static tired light and LCDM. Instead of noting this, you excluded LCDM in order to fit it to a particular tired light model that is already known to fit the data. From this you argue that the results are evidence of a static universe. They are not, since they support both a static tired-light model and LCDM. You then went through previous results (which had been focused on a traditional static universe) and argued that they support your tired light model. They do, but they also support LCDM.

      You could have been upfront and noted that your data supports two models. Others have, for example, noted that results support both LCDM and tired light. You chose to ignore that to portray your results as only supporting a stationary universe. Bad science indeed.

  2. Thanks for a good article.

    But not sure about your saying it’s “bad science” either – it does seem striking that the “tired light” model with a rather simple approach just happens to match what you get with more complex assumptions of different factors compensating for each other in the standard model. Not conclusive or convincing particularly, especially if it can’t explain the 3D background radiation – and didn’t claim to be a complete model – but striking – and it’s also perfectly okay for scientists to be biased in their papers towards their own ideas – it’s for others, reviewers and other scientists who respond to the paper to present other points of view and science advances through that interplay.

    I think that’s more an issue of the media response than the scientists and the way they present the results.

    1. Author

      It isn’t tired light that is the issue, but rather asserting that observation supports tired light over expanding models in an article that doesn’t compare the two. Tired light has been shown to be in disagreement with observations.

      1. Yes, just that – I think that’s more the media reports did that than the original paper which just said it’s an interesting observation that the tired light idea gives same results – just for red shifts – as the standard model and didn’t go as far as to say that it supported any tired light model, which to my mind is still within the realm of good science, controversial, but good. Having said that I’m just going by the abstract and the reports of the paper as it’s behind a paywal for m.

      2. A bit more about it -it’s a different stage in the scientific method. You have stage of rigorous final theories where you make a prediction and it stands or falls by it, which you do when you have nearly all the pieces of the jigsaw puzzle in place already.

        But theories that disagree with observations can still have something interesting to say, especially when competing theories are incomplete – in this case the main competing theory postulates dark matter and dark energy which nobody has yet observed. So you couldn’t possibly say it is in any way a complete theory yet. It’s a theory with a big unknown at its heart.

        To take a well known example – Einstein’s first explanation of the gravitational bending of light gave the wrong prediction. If they’d been able to do the experiment then he’d have been disproved. But he was on the right lines, just needed to do more work on his theory.

        Here, clearly the simple tired light theory is incorrect if it doesn’t predict the three degree background. But it’s a striking observation that it gives a simpler explanation of red shifts and surface brightness. So – is that a possible lead into another theory? Not tired light obviously in its simple form but something? Or is it just a remarkable coincidence that has no significance?

        It’s well within the field of good science to point that out. And to point out that the competing theory is incomplete with many gaps in it. Not drawing any conclusions as they do – too early to do that, but drawing attention to something striking and interesting which might possibly come to have significance later on.

        Does that make sense to you?

      3. > It isn’t tired light that is the issue, but rather asserting that observation supports tired light over expanding models in an article that doesn’t compare the two.

        Except the paper you linked doesn’t say this anywhere. This was your own editorializing. Other posters are right, the paper didn’t make any claims that weren’t supported by their arguments. They only reported that the observations they investigated matches their model, thus tired light isn’t discounted given this evidence like it’s widely believed, and it might be worth another look.

        1. Author

          As I noted in the post, the authors made no comparison to other static universe models. Their “tired light” approach matches expansion because it’s specifically designed to match expansion. The paper in no way supports their argument that their tired light is worth further investigation.

  3. Author

    Several papers show the Tolman test does indeed agree with an expanding universe, and is in disagreement with tired light. Tired light is not a simplifying hypothesis, since it is in disagreement with a range of different observations and requires an unknown mechanism for redshift. Yes, I realize you think dark matter/dark energy is a worse idea, but both agree with most observations, and both have made successful predictions about the universe. Static tired light models have not. Of course you know this, which is why in your paper you argue that your new Euclidian tired-light approach can be fit to observations. This is hardly compelling evidence that the universe is not expanding.

  4. Thanks for this post. I was thinking along the same lines but have no formal scientific training. I was glad to find my thoughts weren’t that far off after all.

Leave a Reply