Science Is Interpretation

By Neuroskeptic | January 4, 2014 1:07 pm

You don’t need new data to produce new science. A re-analysis or re-interpretation can be just as important and original as a new set of results.

I say this because there’s an interesting discussion going on over at PubPeer. In brief, British physicists Julian Stirling and colleagues have released a draft paper using reanalysis to criticize the idea of ‘striped’ nanoparticles.

Nanoparticles are tiny bits of a material, such as gold. They can be coated in various chemicals (ligands), which has important biological and medical applications. It has been suggested that some mixtures of ligands form regular stripes on the surface of gold nanoparticles, and that these stripes can be seen with an AFM scanning tunnelling microscope (STM).


Stirling et al argue that these stripes are nothing more than artefacts caused by technical limitations in the STM; their argument is that the stripes are no more real than, say, the dots, blotches and shadows that appear when images are converted to JPEG format. According to Stirling et al these artefacts have given rise to 23 peer-reviewed publications, many in top journals. For more on the background, see here and here.

textextAnyway, Stirling et al’s new paper is based largely on a reanalysis of the original data presented in support of the claims about striped nanoparticles. On PubPeer, the authors say that this was a problem when they tried to get it published:

Journals were unwilling to even consider papers which focused on reanalysis of published data (Something which I find very worrying…)

They quote an editor at an unnamed journal that had previously published some of the critiqued work:

However, [our journal] does NOT publish papers that rely only on existing published data. In other words. [our journal] does NOT publish papers that correct, correlate, reinterpret, or in any way use existing published literature data. We only publish papers with original experimental data.”

This is an all too common sentiment. It reminds me of the psychiatrists who, in response to criticism of their paper about bipolar disorder, wrote that:

“[Our critics] view their position as part of a ‘debate’ about the ‘ever-widening bipolar spectrum.’ We consider data, not debates, as central to the progress in the scientific understanding of mood disorders…”

Yet data is not science, or understanding, or even knowledge: it’s just data. All data needs to be interpreted before it can tell us anything. Sometimes an interpretation will be very simple (e.g. “this image accurately represents the subject”), but as Stirling et al point out, these simple interpretations can be mistaken.

Science is the process of understanding the world by drawing the right interpretations from the evidence. A paper that offers a new interpretation of old data is a new piece of scientific work, in every sense of the term, and publishers ought to treat it as such. Much of Darwin’s work was reinterpretation, as was almost all of Einstein’s – just to name a couple of famous scientific reinterpreters.

The idea that new science requires new data might be called hyperempiricism. This is a popular stance among journal editors (perhaps because it makes copyright disputes less likely). Hyperempiricism also appeals to scientists when their work is being critiqued; it allows them to say to critics, “go away until you get some data of your own”, even when the dispute is not about the data, but about how it should be interpreted.

ResearchBlogging.orgJulian Stirling, Ioannis Lekkas, Adam Sweetman, Predrag Djuranovic, Quanmin Guo, Josef Granwehr, Raphaël Lévy, & Philip Moriarty (2013). Critical assessment of the evidence for striped nanoparticles arXiv arXiv: 1312.6812v1

  • templeruins

    Interpretations often depend on how you are arriving at your “data”, and what type of data we are talking about. Purely mathematical and theoretical understandings of phenomena are often providing different interpretations to observed and experimental data of phenomena. Sometimes they cannot be reconciled. This has been happening in modern science since Newton. Look at the fundamentally different opinions on the qualities of light and colour between Newton and Goethe.

  • Guest

    Great piece, reminds me of the martian canals.

  • Rebecca Schwarzlose

    I remember an instance when Nature Neuroscience handled a similar situation well. Two groups had major problems with the analyses in a 2006 paper (Grill-Spector et al., 2006). NN gave both groups the opportunity to publish substantial critiques in the journal. We should definitely have more of that. Data mean nothing without the right analysis.

  • Pingback: Do dogs possess a poop compass? (And other weird science questions) | Welcome to our World()

  • Philip Moriarty

    Thanks, Neuroskeptic, for writing this helpful summary of the issues related to the “stripy nanoparticles” debate. As a co-author of the paper, I’d like to clarify two points:

    (i) The technique used is scanning tunnelling microscopy, rather than atomic force microscopy (although we use both techniques in our group).

    (ii) There are a number of sources of artefacts and improper data acquisition/analysis. One of these is somewhat related to the pixelation described in the article but there are quite a few others. For me, a fascinating aspect of the stripy saga is the strong influence of observer bias, where patterns have repeatedly been picked out of noise or randomly arranged features. The comment from “Guest” re. the Martian canals is very appropriate.

    Philip Moriarty

    School of Physics and Astronomy,
    University of Nottingham

    • Neuroskeptic

      Thanks for the comment. I’ve fixed that microscopy bit!

      • Philip Moriarty

        Thanks, Neuroskeptic, both for fixing the STM/AFM bit, and for writing the article. Much appreciated.

        All the best,


      • Philip Moriarty

        P.S. You may be interested in this blog post re. seeing patterns in noise/random distributions of features:

        About two-thirds of the way down the post is a really neat comparison of spatially correlated and randomly-distributed (Poisson process) points.


  • Pingback: Science Is Interpretation | Dr. Goulu |

  • Uncle Al

    One could employ surface ligands that were inherently self-ordering, such as liquid crystals or A-B immiscible-block co-polymers. The object is not to produce segregative surface patterns as such, but to produce different patterns that cannot have a common basis in Moiré fringes, diffraction, or whatever.

  • Pingback: Science Is Interpretation - Discover Magazine (...()

  • Pingback: Science Is Interpretation : | Welcome!()

  • Pingback: Science Is Interpretation |

  • Pingback: Why we are very unlikely to see an admission that the 2004 paper is based on an artifact, let alone retracted | Rapha-z-lab()

  • Pingback: Whither stripes? | Rapha-z-lab()

  • Pingback: Reproducing science: there are no stripes, but FGF-1 IS the universal ligand | Ferniglab Blog()

  • Pingback: Postpublication "Cyberbullying" and the Professional Self - Neuroskeptic |

  • Pingback: Postpublication "Cyberbullying" and the Professional Self :: School Bullying Council()

  • Pingback: Friday Links | Meta Rabbit()

  • FalseBeard

    It’s genuinely depressing for a number of reasons. These guys have apparently never heard of the Nquist theorem. But it doesn’t stop there.
    At first, second, and third glance, the lines look precisely like digital aliasing. In sampling you have to be very precise in knowing what is contributing to the sample resolution. STMs and ATMs use stepper motors, that is a digital signal in itself. Digital noise tends to manifest in patterns, unlike analog noise which is more random. Anything below the correct resolution frequency has to be filtered or ignored.
    The other thing that gets me is the interpretation. The stripes are not appearing on the surface of the nanoparticles. The surface is a pseudo surface. The TMs, measure the field density – or, the stylus is reacting to changes in the field density. The images are processed, along with a false shadow, that give a very strong impression – which is literally an illusion – that the particles have a discrete surface. Yes, the images are nice. But I can see a severe limit in making any interpretations from the images themselves.

    • FalseBeard

      Specifically, the stripes look like Moire patterns.

  • Pingback: Found while foraging (March 18, 2013) – Nouvelles et satellite scientifique()

  • Pingback: Can Science Work Without Trust? - Neuroskeptic |

  • Pingback: Can Science Work Without Trust?Fresh News Today | Fresh News Today()

  • flocculent

    I disagree with your definition of science: “Science is the process of understanding the world by drawing the right interpretations from the evidence.”

    Science is a process of hypothesis refinement. Progress quite often arises when people draw the WRONG interpretations from the evidence, spurring the proposal of testable hypotheses.

    Still, I think your point still stands. I am hopeful, since the recent fetish for “big data” will inevitably create demand for reanalysis, and monetization will soon follow. I predict that publishers will be all over that style of publication once they find out how to make money from it.

  • Pingback: Stripy Signals (A tutorial on signal averaging) | qqq porn()

  • Pingback: Three little (nano) controversies and their morals | Rapha-z-lab()



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar