How Do We Know? The Value of Scientific Models.

Last month, astronomers announced the prediction of a new giant planet in our solar system dubbed Planet IX, a genuine ninth planet with ten times the mass of Earth.  The announcement lead to some confusion on the Internet as to the whether the planet had actually been discovered.  In fact, no direct observation of this planet has been made. Rather, the planet has been predicted by a model, a simplified description of a system which often incorporates hypothetical elements to explain the variance in data.  Because many models use equations to describe a system, a model can often be thought of as a theory with a mathematical backbone.

Planet IX Nasa
Existing models do not predict what Planet IX would look like, but it is probably a cold, dark, lonely place.

Watson and Crick Double HelixBut when Watson and Crick published a paper describing the double helix structure of DNA as a mechanism for genetic information and inheritance in 1953, was this a model prediction or a discovery?  History has largely remembered this event as a discovery.  But the original manuscript reads, “We wish to propose a structure for the salt of deoxyribose nucleic acid (D.N.A.),” the base-pairing of which “suggests a possible copying mechanism for the genetic material.”  The language is very hypothetical and cautious, as things generally should be in science before very strong evidence has been gathered.  To be sure, strong evidence for Watson and Crick’s model was also published in the very same issue of Nature.  Yet this classic paper is a great example of the blurred line between prediction and discovery.

Because the inner workings of the brain are often difficult to observe directly, models are particularly useful in neuroscience.  This is especially true in clinical research when the patient’s brain may only be studied non-invasively.  For instance, EEG signals recorded from the scalp reflect underlying brain activity, yet it is impossible to know which brain structures generated these signals without opening up the skull.  For this reason, neuroscientists use inverse-source localization, a technique that attempts to model scalp EEG using cortical sources, which explain the variance in scalp recordings.  Some papers more-or-less treat these inverse models as fact, boldly stating that such and such a brain structure generated an EEG signal.  In reality, an infinite number of models can explain variance in any scalp EEG recording.  While some models are better than others, the model chosen by the research often reflects what he or she wishes to find in the data.  This doesn’t mean that such models are worthless — rather, it means that due skepticism should be applied when considering the validity of such models, especially when the model is presented almost as fact.

What makes a good model?  Generally, a good model should not only explain variance in data, but also make new, testable predictions.  Without testable predictions, a model is pure theory with no science.  In 1952, Hodgkin and Huxley developed what is arguably one of the best modeling feats in neuroscience: a mathematical description of the action potential.  After inferring the existence of ion channels in the squid giant axon from careful experiments, the Hodgkin and Huxley model described gating particles — particles which cause the ion channel to open and close in response to an electric field — before such channels had even been cloned and sequenced!


So when is a model considered as truth?  Both Watson and Crick’s double helix and Hodgkin and Huxley’s gating particles have been confirmed with molecular evidence.  But what if a better model comes along?  Indeed, the Hodgkin and Huxley model has been modified many times, and new elements, such as additional ion channel populations, have been added to our understanding of excitable membrane physiology.  In other disciplines, Newtonian physics has been replaced with quantum mechanics and Einstein’s theory of relativity, both of which may soon be replaced by a grand unified theory.  What does it mean, then, to “believe” in something like ion channels or relativity?  The scientific worldview holds that absolute certainty is impossible.  Scientific theories can never be proved in the same sense that mathematical theorems can be proved.  A theory accumulates more and more evidence until it is either generally accepted or a better theory replaces it.  Truth concerning the physical world may exist, but as humans, we wear the blindfold of agnosticism.

Let’s return to astronomy’s newest model, Planet IX.  The discovery of Planet IX will not be announced until light from the hypothetical planet has been observed through a telescope.  For most people, this counts as direct observation.  But is it really?  No human eye will directly see the planet.  The point of light seen through the telescope might be an artifact created by the signal processing algorithm used to generate the image.  If the position of the artifact in the image is seeded by the time on the computer’s clock, the artifact might even drift across the image each night as one would expect from a planet.  This scenario is highly unlikely, and a good astronomer would know how to rule it out with a large degree of certainty.  Nonetheless, even if such light is directly observed, Planet IX is still, in some sense, a model explaining variance in data.  For this reason, independent replication of findings is an important part of the scientific process. 

French philosopher and mathematician René Descartes was among the first to think deeply about how we know things. His famous quote, “I think, therefore I am,” summarizes the view that one can doubt the existence of anything except one's own existence.
French philosopher and mathematician René Descartes was among the first to think deeply about how we know things. His famous quote, “I think, therefore I am,” summarizes the view that one can doubt the existence of anything except one’s own existence.

Can we trust science then, you ask, if its models are seemingly delicate, fragile things, always at risk of being discarded and replaced by new evidence?  Well, chances are you trust your brain, which also uses models to explain variance in sensory input in much the same way as a scientists uses models to explain variance in data.  Each time you fall victim to an optical illusion, your brain has used an incorrect model to explain input from the retina.  Have you ever had the experience of thinking you recognized someone in a crowd, only to discover that the person was a stranger?  Again, the brain attempts to fit visual experiences to a model, which may later be rejected after gathering new “data.”  One important model used by the brain is called theory of mind, the idea that other people have minds and internal experiences similar to your own.  A theory of mind emerges in children’s brains early in development, and it is hypothesized that individuals with autism have difficulty applying this model.  Sometimes, we wrongly apply theory of mind, such as when we assume the characters in our dreams are real people like ourselves.

Absolute certainty is not just a problem for science.  It is a facet of the human condition.  Science is the tool we use to build a working understanding of the world around us from experiment and observation.  While the only thing you may be absolutely certain of is your own existence, this is not a carte blanche for ignorance.  Rather, understanding the fluid nature of scientific knowledge allows us to be both skeptical of existing dogma and open minded to new theories.  Without such fluidity, science would not exist.



Batygin, Konstantin, and Michael E. Brown. “Evidence for a Distant Giant Planet in the Solar System.” The Astronomical Journal 151.2 (2016): 22.

Hodgkin, Alan L., and Andrew F. Huxley. “A quantitative description of membrane current and its application to conduction and excitation in nerve.”The Journal of physiology 117.4 (1952): 500.

Pascual-Marqui, Roberto Domingo. “Review of methods for solving the EEG inverse problem.” International journal of bioelectromagnetism 1.1 (1999): 75-86.

Watson, James D., and Francis HC Crick. “Molecular structure of nucleic acids.” Nature 171.4356 (1953): 737-738.

Joel Frohlich

Joel Frohlich is a postdoc studying consciousness in the lab of Martin Monti at UCLA. He is interested in using brain activity recorded with EEG to infer when a person is conscious. Joel earned his PhD from UCLA in 2018 studying EEG markers of neurodevelopmental disorders in the lab of Shafali Jeste. You can also check out Joel's blog Consciousness, Self-Organization, and Neuroscience on Psychology Today. For more about Joel's research and writing, please visit Joel's website at

One thought on “How Do We Know? The Value of Scientific Models.

  • February 25, 2016 at 10:05 am

    The professional behaviors that are labeled as “science” are so varied as to make the label mainly rhetorical – mainly by anti-science folks. A more useful and accurate phrase is experimental knowledge. Actually, all we have is statements. True statements predict measurable events in the future – period. All other statements refer to magical and make-believe claims. However, it seems that 99% of human language is statements referring to imaginary things. For example, “I decided to to order coffee.”, etc. There is no “I” and no conscious decision making.

Comments are closed.