Following the scientific consensus: how to be “the least wrong”

There are two important and common words that, when used scientifically, have a very different meaning than how we use them in everyday language: theory and consensus. These two words, in our commonplace usage, have meanings that imply a large degree of uncertainty, and enormous amounts of wiggle-room for how the reality surrounding these ideas could turn out to be vastly different from our current conceptions. A theory is merely a thought that anyone can put forward: a supposition, a wild guess, or even baseless speculation all count as “theories” in our daily conversations, where strongly validated ideas like gravitation and completely erroneous ones, like the Earth is flat, both get described with the same word: theory.

While most of us can recognize the difference between a scientific and non-scientific use of the word theory, this line is even blurrier when it comes to the notion of a consensus. Consensus, when we use it commonly, simply means, “most people believe this thing,” but that doesn’t necessarily mean such a thing is correct or true. Consensus could apply just as equally to statements like “the Earth is warming” as it could to those like “ninjas are deadlier than pirates,” even though the first one is well-quantified and the second one is not.

However, when a scientist talks about consensus, they are talking about something far more powerful: the least wrong approximation of reality supported by the full suite of evidence and the overwhelming majority of professionals in a particular field. Here’s the thought process behind why everyone would be better-served to follow the scientific consensus: in the end, it empowers all of us who do so, while imperiling all who reject it.

Vaccines, including the measles and HPV vaccines, are very safe, and most people don’t have any problems or side effects. Studies have shown that vaccines cause the rates of illness and infection decline precipitously, with the HPV vaccine leading to a 64 percent reduction in HPV rates among teenaged girls ages 14 to 19, and 34 percent among women ages 20 to 24. Yet vaccination rates remain low for HPV and many other preventable diseases, causing thousands of unnecessary illnesses and ill effects.

Theory: this is the starting point of it all. If we ever want to understand what it means to abide by or reckon with the scientific consensus on an issue, we have to go back to an understanding of this key term: a theory. Specifically, here are four things that “theory” is not.

• It isn’t the colloquial definition that we’re used to, which is any proposed explanation for why some phenomenon occurred. (E.g., flat Earth theory, hollow Earth theory, young Earth theory, etc.)
• We also want to be careful to stay away from the mathematical definition: a self-consistent set of axioms or postulates that allow the construction of a framework. (E.g., set theory, number theory, even string theory, and so on.)
• We mustn’t fall prey to the trap of using “theory” to talk about a speculative extension to the mainstream, accepted theories that we have that don’t have adequate supporting evidence behind them. (E.g., supersymmetry theory, Kaluza-Klein theory, technicolor theory, composite Higgs theory, etc.)
• And finally, “theory” in this context also cannot refer to an idea that was once viable, until it failed to explain key pieces of evidence, conflicting with a key measurement or observation. (E.g., Lamarckian evolution, the Sakata model for mesons, the SU(5) grand unified theory, etc.)

Instead, when scientists most frequently talk about theories, they talk about the accepted theories that are overwhelmingly supported by the evidence: the starting point for modern science. General Relativity is our theory of gravity; the Standard Model is our theory of elementary particles; germ theory is our theory of how infectious diseases work; genetics and Darwinian evolution are our theory of how living organisms pass on their traits to future generations; etc. When scientists typically mention a theory, they’re discussing the prevailing theory of the day: what’s already been robustly established as the outlined framework for all current and future discussions.

The Standard Model particles and their supersymmetric counterparts. This spectrum of particles is an inevitable consequence of unifying the four fundamental forces in the context of String Theory, but supersymmetry, string theory, and the presence of extra dimensions all remain speculative and without any observational evidence. They are not part of the scientific consensus.

The novel phenomenon: ideas like “scientific consensus” rarely come up unprompted, or in a vacuum. Instead, they usually come up in discussions surrounding an issue because something new, important, unexpected, or not fully explained has been observed to occur. Here are some observations that might appear to be those novel phenomena.

• We observe that the carbon dioxide concentration in the atmosphere is rising, that the pH of the oceans, globally, are acidifying, and that extreme temperatures are being recorded more frequently all over the world.
• We observed that an astrophysical cataclysm occurred some 130,000,000 light-years away, with gravitational waves and gamma-rays first arriving from that event in 2017. When we detected them, we found that the gravitational wave signal ceased arriving ever-so-slightly before the very first electromagnetic signal showed up: by 1.7 seconds.
• And we observed, in late 2019, the emergence of a novel disease in humans. When we sequenced the genome of the pathogen responsible for it, we found that its genetic sequence is similar to, but evolutionarily divergent from, other known disease-causing agents in the same family.

Although these may seem like wildly disparate examples from a variety of scientific fields — the climate change problem in the context of environmental and geological/atmospheric sciences, the astrophysical neutron star-neutron star merger observed in both gravitational waves and electromagnetic radiation, and the origin of SARS-CoV-2 in the context of virology, disease ecology, and epidemiology — scientists take the same approach in every instance. In fact, the first step in all of these cases is simply as follows.

This figure shows the structure of the spike protein in SARS-CoV-2. Panel A shows the spike homotrimer in its open configuration, while panel B shows the cleavage sites on the spike protein. Note how the configuration of a protein, and how it folds in its environment, controls many aspects of its functioning. Even an identically structured protein isn’t going to perform the same in different environments, which is research that would be outlawed under the proposed Dangerous Viral Gain of Function Research Moratorium Act.

Identify the null hypothesis: this is an unspoken step that any scientist will recognize when it comes to their own field, but that simply doesn’t occur to most people in most endeavors of life. When we say “the null hypothesis,” what we mean is to look at the novel phenomenon and ask, “what explanation for this new, observed phenomenon would indicate that its emergence is already accounted for by the known laws, theories, and frameworks that are already in place to explain our reality as we know it?”

The null hypothesis would mean that, sure, you’ve discovered a new phenomenon, but no new rules or outside influences need to be invoked to explain it. It’s simply a novel manifestation of the rules, laws, models, and frameworks that have already been established.

The null hypothesis sometimes means, “things are behaving as they’ve always behaved, and what we’re observing is within the realm of natural variation/variety.” Numerous announced discoveries that were later overturned occurred because of a statistically unlikely result in the data that didn’t hold up over time, especially when more, better data was acquired. Ruling out the null hypothesis, however, can be an incredibly powerful achievement. In the case of the temperature of the Earth, going all the way back to the earliest global temperature records in the early 1880s, the null hypothesis is now ruled out at greater than 5-sigma confidence, with less than a 1-in-3.5 million chance of it being a mere statistical fluke.

This three-panel animation shows: (1) what the actual CMB looks like, including in both E-mode and B-mode polarizations, (2) the expected modification from an anisotropic or rotating Universe, and (3) the CMB as it would appear if it contained such a modification. The data is most consistent with no rotation and no anisotropy at all: the null hypothesis.

So, we’ve found something’s new. Now what? Again, there’s an unspoken step that scientists take that’s rarely discussed. Scientists often ask themselves an important question, particularly when a novel phenomenon crosses the threshold of ambiguity and whose existence can now be considered non-controversial. That question is, simply, “is there a way to explain what we’re seeing in the context of what’s already known, without introducing anything novel and extraordinary?” Going back to our earlier examples:

• The Earth is warming, the oceans are acidifying, and the carbon dioxide concentrations have been rising, too, all together, over the past couple of centuries.
• The arrival time of gravitational waves and electromagnetic signals have been accurately measured and their origin point has been confirmed to be identical, and yet the gravitational waves still get there 1.7 seconds earlier, even though both should travel at the same speed: the speed of light.
• And the novel coronavirus SARS-CoV-2 did, in fact, emerge in humans in late 2019, even though the precise origin of how this virus found its way into the human population remains obscure.

What we typically do in this situation is resort to what some scientists also call “the null hypothesis” but which I prefer — to distinguish it from our earlier “nothing to see here” example — to call the default hypothesis: the idea that everything needed to explain this emergent phenomenon is already known, but that we just need to correctly identify the important contributors. Can we explain this novel phenomena with the default hypothesis? That’s the question we should always be asking before attempting to resort to anything extraordinary.

Illustration of a gamma-ray burst, thought to occur from the merger of two neutron stars. The gas-rich environment surrounding them, as well as the matter from the neutron stars themselves, could delay the arrival of any electromagnetic (but not a gravitational) signal, explaining the observed 1.7 second difference between the arrivals of the gravitational and electromagnetic signatures. This is the best evidence we have, observationally, that the speed of gravity must equal the speed of light: to approximately 1 part in 10¹⁵ (a quadrillion).

Identifying what matters: a lot of people have this misconception that science is wedded to what we’ve already established, and that scientists are incredibly resistant to new ideas. But this is not how science works (or how the overwhelming majority of scientists behave) at all. Although you can certainly find people — even a few scientists among them — who feel that this is the case, the truth is far less exciting.

The reality is that what’s already been established, scientifically, provides us with an incredibly strong and versatile foundation to accommodate almost any new phenomenon that’s going to be observed. Our understanding of the natural world is powerful and far-reaching, if not quite comprehensive and complete.

The default hypothesis, in practically any case we encounter, is going to be this: there is a completely mundane explanation for this novel phenomenon that only relies on correctly applying the science of what’s already known to the situation at hand. The default hypothesis is the least radical suggestion of all: that you might need to add only a small, additional twist, ingredient, pathway, or component in order to get the full story out. The assumption is that once you apply the underlying scientific rules correctly, you can wind up fully explaining everything that has been observed.

This graph shows the global average temperature anomalies relative to the 1850-1900 baseline. The red line shows the multi-year moving average of global temperature, while the dotted green line shows a linear fit to the warming from 1974-2022. As recent years show, the warming trend has accelerated in recent years, with 2023 and 2024 marking severe (hottest-ever) departures from the late-20th century trend.

Recognizing alternatives for what they are: sometimes, but not frequently, they can be equivalent or even superior explanations for these novel phenomena. After all, sometimes there really are novel rules that come into play, and oftentimes our first clue that our current theoretical framework needs modification comes exactly in the form of an unexplained observation. However, elevating the alternative explanation to the status of leading explanation requires something more: a demonstration that the default hypothesis is somehow insufficient.

This has happened numerous times throughout history, of course. Whenever the default hypothesis fails to explain what we see, it’s eventually led to a scientific revolution in our conception of reality.

• The fact that Mercury’s orbit around the Sun couldn’t be explained by Newtonian gravity led scientists to hypothesize an unseen, inner planetary companion to Mercury: Vulcan. Only when Vulcan failed to turn up was the alternative hypothesis — that Newtonian gravity needed to be superseded — explored and eventually validated, culminating with Einstein’s General Relativity.
• The fact that the Earth is, geologically, billions of years old seemed incompatible with the Sun’s current power levels sustaining its energy output over billions of years. The mechanism of gravitational contraction could only sustain the Sun for tens of millions of years; it wasn’t until decades later that the secrets of nuclear physics would pave our way for understanding how the Sun actually worked.
• And the fact that galaxies are zipping around inside galaxy clusters at speeds far too great to be consistent with the amount of matter present inside them led to the idea that some “dark” form of matter was present throughout our Universe. Only after decades of robust observations confirmed that there was no form of normal matter that could account for these motions — and additional observations (of individual galaxies) independently confirmed the cluster problem — was dark matter accepted into the mainstream.

A cross-section of the Wealden Dome, in the south of England, which required hundreds of millions of years just to explain the erosion features observed. The chalk deposits on either side, absent in the center, provide evidence for an incredibly long geological timescale required to produce this structure: longer than any contemporary explanation for the Sun’s energy could have provided in the late 19th century. This was noted by none other than Charles Darwin in the mid-1800s, and would present a puzzle that would not be resolved until the process powering the Sun, nuclear fusion, became understood.

However, these examples are exceptional and noteworthy, not just for leading to scientific revolutions in the 20th century, but for how rare events like these are. Far more frequently, the default hypothesis is going to be the one that carries the day. It’s important, as a scientist, to entertain the possibility of alternative explanations for any phenomenon you might have observed, but to relegate them to the status of both speculative and unproven until you establish the insufficiency of the default hypothesis.

And establishing that insufficiency, perhaps unfortunately, is tremendously difficult to do.

• The default hypothesis is that the Earth’s temperatures are warming, its climates are changing, and its oceans are acidifying because humanity has significantly modified the contents of our atmosphere, largely through the burning of fossil fuels for energy.
• The default hypothesis is that gravitational waves arrive before electromagnetic waves because the light that’s generated from a neutron star merger must travel through matter — which slows down light — before arriving at our eyes, while the gravitational waves simply pass, unimpeded, right through that same matter.
• And the default hypothesis is that SARS-CoV-2 emerged in humans through zoonotic spillover, contemporaneous (or possibly just prior to) the superspreader event at the Wuhan market, likely through some form of animal agriculture, farming, or encroachment of human activity into previously wild territory.

The central idea of the lab leak hypothesis, that the virus spilled over from the Wuhan Institute of Virology, is only possible if the virus from which SARS-CoV-2 originated was actually ever inside the institute itself. If the virus originated naturally, with parts of it found in animals that were located in a wild population in Laos, which genetic sequencing uncovered in 2021 indicates, the lab leak hypothesis is ruled out as a possibility. You cannot create something through gain-of-function research that will have an identical genetic code to something that came about in the wild through natural processes such as recombination.

Consensus. So, now let’s assume that we’ve done our homework. We’ve learned everything that humanity knows about the knowledge relevant to this particular scientific issue, just like all the leading scientists in a particular discipline try to do. Now, the critical moment comes: we’re trying to synthesize together everything that we know and obtain a scientific consensus.

What does a “scientific consensus” mean, in this context, and how do scientists get there?

A scientific consensus — where the overwhelming majority of experts agree on the explanation for something that’s been robustly observed — isn’t for every problem. It can (and will) only be achieved if:

• a single framework explains all of the legacy puzzles as well as the novel phenomenon,
• no unproven, evidence-free conjectures need to be true for the explanation to hold,
• when the full suite of evidence is considered — scientifically admissible evidence, as opposed to speculation or circumstantial evidence — there are no “dealbreaker” puzzles that remain,
• and if the overwhelming majority of expert-level professionals actively working in the field all draw the same conclusion: that this one, favored, consensus picture is the best explanation for everything we’ve observed.

Any consensus we achieve is always provisional, of course; any one of the alternatives could always turn out to be true. But if you are to truly compete with a consensus opinion — the Standard Model, dark matter, cosmic inflation, Darwinian evolution, human-caused global climate change, the natural origin of SARS-CoV-2, etc. — you have to identify where and how the consensus opinion breaks down, and to demonstrate where your preferred alternative not only succeeds where the consensus fails, but to demonstrate its success in every place where the current consensus also succeeds. Otherwise, the “alternative” position is pure contrarianism, which sometimes, as in the case of pursuing alternatives to universal vaccination campaigns, comes along with deadly consequences.

In February of 2025, noted anti-vaccine, anti-fluoride, and anti-GMO crusader Robert F. Kennedy, Jr., despite the objection of qualified scientists and medical professionals, was sworn in as the director of Health and Human Services in the United States. In the one month since that event took place, vaccination rates have dropped, and vaccine-preventable diseases have resurged, claiming lives unnecessarily.

Over the course of human history, what was once a consensus opinion among scientists has been found to be insufficient on one or more accounts. When this occurs, the “old consensus” doesn’t suddenly become wrong, but rather gets demoted to a mere approximation — or special case — of a more comprehensive framework: a new, superior scientific consensus. Our current consensus is not evidence of groupthink, but rather is the culmination of our modern scientific enterprise. It is the best approximation of reality that the full suite of evidence, in the context of our most successful scientific theories, can possibly put forth.

As in all things, many of today’s consensus positions will no doubt be found to be lacking in some key way, and will someday be regarded the same way we regard Newtonian gravity: revolutionary for its time, accurate and useful under certain conditions, but only an approximation of a deeper, more fundamental description of reality. That is not a flaw in the scientific method nor in our way of thinking today; that is the nature of science. A scientific consensus is not just valuable; it is the foundation for creating an evidence-based, reality-based society and civilization that can thrive.

When we interrogate the Universe in just the right fashion, we must keep our minds open to the possibility that a deeper truth may yet be revealed. However, it won’t render what we currently think useless or untrue; it can only, at most, refine our conception of how things are. The key to advancing human knowledge is to understand the limitations of the current consensus position and to identify the criteria necessary to overthrow it. Unless that’s precisely what you’re doing whenever you consider an alternative, you’re arguing against the colloquial, and not the scientific, meaning of consensus.

A version of this article was first published in July of 2021. It was updated in 2025.