Scientific models are falsifiable

A scientific model has no chance of being right unless there’s a possibility it’s wrong.

This is one of ScienceOrNot’s Hallmarks of science. See them all here.

In short…

For a model to be regarded as scientific, there must be some way of testing whether it is false. A model remains in contention as long as no falsifying evidence appears from such tests.

To be scientific, a hypothesis has to take a risk, has to “stick its neck out.” If a theory takes no risks at all, because it is compatible with every possible observation, then it is not scientific.

Peter Godfrey-Smith, Australian-American philosopher, 2003

What does it mean for a model to be falsifiable?

Scientific models must have the potential to be refuted or falsified. This means that it must be possible to do a real-world test that could show the model is incorrect.  The test would produce data that, according to the model, should not occur.

How to tell that a model is falsifiable

A scientific model should specify what evidence would falsify it. For instance, the Darwinian evolutionary model says that advanced organisms developed from simpler ones. This could be falsified by finding fossils of advanced organisms in rocks that are older than those their ‘ancestors’ are found in – “fossil rabbits in the Pre-Cambrian era”, as the biologist J B S Haldane put it.

Most scientific models are presented in a straightforward way, and it’s not difficult to see how they can be falsified. If a model is so vague and all-encompassing that is compatible with all possible evidence, then it cannot be falsifiable. It cannot, therefore, be scientific.

Why models must be falsifiable

It may seem paradoxical, but it’s because scientific models are falsifiable that we can have confidence in them.  Scientists accept a model when it is supported by evidence, but there is always the possibility that falsifying evidence may be found. Good science, as well as looking for evidence in support of a model (see Scientific models are tested against the real world), actively checks to see whether falsifying evidence exists, .

If, as time goes on, supporting evidence accumulates and no falsifying evidence is found, scientists have increasing confidence in the model. However, no scientific model can be ever considered ‘proven’; all scientific models are tentative to some degree (see All scientific models are tentative).

Note that if a model IS falsified, it doesn’t have to be totally abandoned. It may be possible to alter some aspects of the model so that it accounts for the new evidence.

Bogus science typically invents models that are not falsifiable and then works on finding confirmations. Since there is no possibility of ever finding evidence to contradict such models, the only evidence we can ever expect to find will be either supportive or neutral, which is meaningless. The models might be true or false, but there is know way of ever knowing.

Examples

• The luminiferous aether model was falsifiable and has been falsified. Scientists once felt that light could not travel through empty space, and they proposed that there was a medium called the aether through which light travelled . Its existence was shown to be false, most famously by the experiments of Michelson and Morley.
• The model of parallel universes proposes that there are other universes existing in parallel with ours. The model is not falsifiable because there is no way we can receive information from any other universe. It cannot claim to be scientific.
• Reincarnation is not falsifiable because there is no conceivable way of showing that souls, if they exist, cannot be passed from one mortal being to another. Reincarnation is not a scientific model.
• The biological evolutionary model is falsifiable – as explained above – but has not been falsified. Jerry Coyne has more observations that could falsify evolution here.
• The Intelligent Design model for the existence of the universe is not scientific because it is not falsifiable. An intelligent designer could overrule our notions of true or false, so nothing could prove the model is incorrect.

Further reading

Applying falsifiability in science at coelsblog.

Peter Godfrey-Smith’s quote is from Theory and Reality, p58.
The Pre-Cambrian rabbit image is by Tochtli-Rabbit-Conejo.jpg: DoriPrecambrian_sea.JPG: Ghedoghedoderivative work: Obsidian Soul [CC-BY-SA-3.0], via Wikimedia Commons. 

This is one of ScienceOrNot’s Hallmarks of science. See them all here.

Updated: 2014/07/22