Can science be applied to questions about the past?


A friend of mine recently wrote on his blog post that “we can’t apply the scientific method to the past” [1] .

I admit to sometimes being a little knee jerk with such comments: I often encounter them in religious circles, used as a way to dismiss particular scientific findings that appear to contradict religious doctrine. It is all too easy, by saying “science cannot go there” to throw out contradictory evidence with the flick of the wrist. In the process, one throws out a large number of topics that have long been considered science: geology, paleontology, cosmology, etc.

In any case, knee-jerk reactions are never worthwhile. His question is very interesting and a good teaching opportunity to review what science is, and to discuss why these subjects are indeed legitimately scientific. So, here we go.

Nobody Was Around Back Then

One argument often put forward is that we cannot make scientific statements about events that happened before anyone was around to witness them. This is a slippery slope: why stop with the past? It is true that nobody was around to witness the dinosaurs, but indeed nobody has ever seen a proton! Yet, I’ve never heard anyone question the existence of protons.

The vast majority of contemporary science happens far outside the realm of direct human experience (which covers an incredibly narrow band in the Universe). Most of science relies on instrumentation and techniques that extend our reach beyond the narrow world of our limited direct experience. These techniques have worked to great success, allowing us to see many orders of magnitude smaller than the wavelength of visible light, many orders of magnitude beyond our planet, and many orders of magnitude in the past. What makes any of these studies scientific, is not whether people are around to see them, but rather they are scientific because they correctly follow the methods and protocols of science.

So What Are the Protocols Of Science?

It is not my intention in this article to provide a full review of the scientific method. I happen to really like Peter Hadfield’s youtube video on the subject. Let’s just focus on some of the key points:

Is there data? Of course. We have fossils, mountains, volcanic rocks, erosion, radiometric abundances, ice cores, seasonal sediminent deposits, tree rings, etc.

Are there quantifiable, testable theories that explain the above phenomena? Of course: theory of plate tectonics, theory of evolution, nuclear theory,  etc.

A key point in science is the idea of falsification. Scientific theories must be capable of tests that can falsify them if they are incorrect. Is this the case for “historical” science? Of course. For example, there used to be many competing hypotheses regarding the extinction of the dinosaurs. As evidence piled in, the majority of these were falsified and the discussion focuses around a very limited number of hypotheses (for example, see The K-T ExtinctionThe Chicxulub Debate and Deccan Volcanism expressing several competing but much more circumscribed theories on dinosaur extinction) [2].

More importantly, have these theories made novel predictions a priori (without knowing the outcome) that were later tested and validated? Yes! Scientists have made countless thousands of predictions regarding evidence of past events that should be discoverable in the geological and fossil records. These predicted observations have since been independently and repeatedly confirmed.

A hypothetical example: a geologist discovers the site of a huge volcanic eruption. The date of the eruption is estimated from the location of the site in the geological column. Multiple samples are sent to several radiometry labs. Without knowing the origin of the samples, these independent groups consistently date the sample to a period in time that agrees with the geologist’s original estimate. Researchers look in the ice core record to find a dust layer consistent with aerosol particles from a volcano. Knowing when the eruption happened, they count back the layers of seasonal ice deposits to the range of years predicted by the radiometric dating. They find a large eruption in exactly the right spot. Similar observations are made in other geological records. Isotopic analysis shows all of these layers in multiple ice cores around the globe and in various sediment deposits to match the composition of material found at the site of the eruption. A real life example of this hypothetical described in more depth in this article on The Toba Super Eruption and Polar Ice Cores and this one on The Lake Malawi Sediment Chronometer and the Toba Super Eruption [3]. How is this not science?

But you cannot conduct controlled experiments?

There is a common misconception that to do science, you must be able to conduct laboratory experiments. By such an argument, one could say that “mountains exist” is not a scientific proposition, simply because it is impossible to create and observe a mountain in the laboratory. On the contrary, some sciences are based almost exclusively on field observation. This, of course, does present a number of unique challenges. Performing a “control run” is impossible with historical events, but it is still possible to implement “controls” against researcher bias. However, the unique challenges of historical science are understood and addressed by the experts in these fields. For an excellent in depth discussion, read this article [4].

One interesting point made in Ref [4] is that establishing whether an event happened is an easier task than deriving a generalizable causal model for a phenomenon. Saying “B happened” is much easier than saying “phenomenon A generally causes phenomenon B through mechanism X”. The many trials and the control runs in laboratory science are necessary in order to disentangle the possible causes of an effect. These are unnecessary in establishing the particularities of a single historical event. A criminologist must study the profiles of many serial killers and many normal individuals in order to understand what makes a serial killer. A forensic detective need only establish that a pattern of murders is consistent with a single serial killer. There is no arguing that both of them are studying the same phenomena. The difference is that one looks forward in time (the criminologist) and the other looks backward (the forensic detective).

What if the laws of nature have changed in time?

As I discussed in my post on coherence between many lines of evidence, scientists don’t just look at a single measurement or a single technique. We compare different measurements to each other. The more and varied these measurements are, the more constrained is the system we are trying to understand. To go back to that previous example: If the laws of radioactive decay had changed significantly over the last hundred thousand years, we would see them diverge from the tree ring or the ice core record. One would be hard pressed to come up with a scenario where completely independent physical processes such as seasonal melting patters in ice, seasonal sediment deposits in lake beds, chemical processes, etc would vary perfectly in concert so as to render the change in radioactive decay rates undetectable. Rather, if all of these independently corroborate the story told by a radiometric measurement, then I can rest assured that the decay rates must have been pretty constant with respect to a wide variety of other physical processes. These sorts of calibrations (including volcanic eruptions) have been performed many thousands of times by many thousands of scientists for nearly a century. The majority of evidence from our past works out to be very consistent with science as we know it, and across all major fields of science.

But You Are Taking on Faith That Different Scientific Principles Should Behave Coherently?

It is not necessary to take the coherence of physical law on faith. The proof is in the pudding. Science is an empirical endeavor. We value a model on: does this model explain the data? Can this model effectively elucidate new observations that we hadn’t thought to look for? If the model works, we have a good theory.

The coherence of natural principles in the past is demonstrated by the fact that our scientific theories are able to make coherent statements about them. If natural law were incoherent in the past, or even inconsistent with modern science, then our theories would not be able to make sense of the observations: something as simple as comparing carbon dating with tree rings would just break down and give divergent or nonsensical answers.

At the end of the day, it is always true there are certain implicit assumptions built into a scientific analysis. This is as true for science of present phenomena as it is for past phenomena. But the goal is:

1) Minimalism in assumptions
2) Hyper-sensitive awareness of those assumptions and their possible consequences.
3) Constantly testing those assumptions: either through direct measurement or by working out their consequences. Never taking those givens for granted. Not accepting them on faith or just because they “feel right”.

Few non-scientists will ever appreciate the degree to which scientists take these principles seriously. We spend our every day thoroughly and systematically seeing these principles applied. I wish more people could experience the work it takes to perform a precision measurement in science.

Maybe the laws themselves all changed coherently?

Imagine a society with a gold-based currency. Now imagine I told you that the value of gold per-gram has doubled. Now imagine further that the prices in this economy simultaneously doubled across the board. The price of an apple would be twice as expensive, but the value of a gold coin would also be twice as big. If an apple cost 3 gold coins before, it would still cost 3 gold coins. Since all of the relationships between money and price remained the same, a person would be unable to detect any meaningful change.

As in the above metaphor, science studies the interrelationships between various natural phenomena. If all of these phenomena change in a way that preserves those interrelationships, than at some point it becomes meaningless to say that anything was different. At the end of the day, if the science of today can make predictive statements about evidence from the past, then one can say that the science of today is perfectly good at describing said evidence.  One can say that science is still valid, and it is merely a matter of philosophical speculation about “the nature of reality” to delve any further. After all, our brains could be in “The Matrix”, in which case all of our science (past and present) is purely fictional.  Such speculations may be valuable themselves, but they go beyond science and into the realm of metaphysics.

A person can always invoke “unknown unknowns”, or speculate as to the existence of some mechanism for fooling the science. Such claims are often made without doing any of the hard work to provide a mechanism or demonstrate it through evidence based inquiry. As a scientist, I always leave room for the “unknown unknowns”. But, I have to be pragmatic and I have to live in a world where some things are better known than others. We understand many events of the past better than we know of many happenings in the world today.


The fact that a topic is based on field observation rather than laboratory experimentation does not mean it isn’t science. Moreoever, direct eye-witness human contact is not a precondition for science. Science is a about building precise, numerical theories that explain what is known and make verifiable predictions of phenomena that were previously unknown. The latter of these two criteria is essential. The ability of a theory to predict phenomena that nobody previously thought to look for is the key to making sure that said theory is not merely being tweaked or fudged to match the known data alone.

Most scientific frameworks for explaining past events satisfy these criteria in spades. While cosmology of the early Universe is still subject to big unknowns, natural history going back a few hundred million years is well known not only in broad strokes, but even in terms of exact details: periods of glaciation at exactly the calculable times of wobbles in the earth’s orbit,  mass extinctions, asteroid impacts, and volcanic eruptions.

Scientific paradigms in biology, astronomy, physics, geology, chemistry, and climatology are able to precisely and consistently describe phenomena in the geological record on multiple levels of complexity: ranging from the fundamental (eg nuclear decays or thermoluminesence or rates of fossilization) to the macroscopic (eg tree growth, statistical changes in fossil morphology, glacial growth, or solar evolution). Having read some of the primary literature (though I don’t claim to be an expert), it is impossible for me to understate how vast the evidence is. Anyone who tells you otherwise is either lying to you or deceiving themselves. I urge my readers not to take me on my word. Learn about it. It’s really cool stuff! And, I’m glad to help you out with references or explanations on technical matters (to the best of my ability), and to refer you to experts when I’m in over my head.

Further Reading

[1] Patently Jewish – Folly of Faith, Folly of Reason

[2] R. Cowen- The K-T Extinction , and G. Keller – The Chicxulub Debate , Deccan Volcanism

[3] The Toba Super Eruption and Polar Ice Cores and The Lake Malawi Sediment Chronometer and the Toba Super Eruption. Check out the primary sources for further reading.

[4] C.E. Cleland, “Historical Science, Experimental Science and the Scientific Method“, Geology (Nov 2001) v. 29; no. 11; p. 987–990. Click here to read the full article.



  1. Michael Feigin, Esq., NY, NJ Patent Attorney

    I can’t say I really disagree with you, but I argue the further out we go in any observable dimension, the more difficult it is to arrive at truth. This is true of both space, time, and even more so … for other less observable dimensions. Your mountain example is an example which is readily observable in time (we can see it now and see it continue to exist now) and space (it’s rather large and we can observe it with our own eyes). Now your proton example – that’s more difficult because of it’s size (requires special equipment) though it’s observable in time (it exists while we do). But our methods of observation, even for a proton, become somewhat indirect. Our method of observation of a quark, less direct … but repeatedly testable, no?

    Now, extrapolate out time and we have a much greater difficulty because there is nothing observable but the remnants. Yes, we explore this, and no that’s not bad science, but it is limited by what we can observe. The further away in this dimension, and no matter how well we increase our tools of observation, we still have a large problem. Some theories would require testing over a billion years (e.g. unguided macroevolution), others are simply not repeatable in our known universe (e.g. big bang).

    Worse, the further out in time or dimension beyond what we (and I suppose, our tools) can observe, the less points of light we have and the more building on already existing approximates of theories. That is, the less observable, the more we build on extrapolation and interpolation. As someone who didn’t know in a 6th grade experiment that the metric system set the mass and volume of water equal and who extrapolated out from a calculation for the weight of 20mL, I can tell you that my 100mL measurement was way off and I learned a valuable lesson.

    I further argue that your faith and reason should coincide, or, at the very least, we should be able to live with the contradiction until we find more evidence which explains the contradiction. For me, Dr. Gerald Schroeder’s work provided the path towards this.

    • Matt

      Hey Mike,

      My point is that neither a T. Rex nor a proton can be observed directly, and no amount of being present “in time” will make these observations any more direct. We know about both of them only through indirect evidence. But, there is a ton of said evidence…to the point that I would feel ridiculous denying either of their existences. I would suggest that being contemporaneous has little to do with my confidence. All that matters is the quantity of the evidence and my confidence in the methodology.

      Repeatability is no less a component of historic science than it is for laboratory science. I can observe a proton again and again, but I can also dig up more and more dinosaur bones. In the time since Darwin first articulated his theory, the number of discovered fossils has increased many orders of magnitude (arguably factors of tens of thousands). And the number of transitional species discovered is quite remarkable. We’ve reached a point where paleontologists can actually numerically curve-fit for rates of speciation in the fossil record (though only for some clades).

      The only difference between lab science and historic science with respect to repeatability is that I can create new data in the lab, so I can collect as much data as I want. The total number of fossilized dinosaur bones is ultimately finite (though quite large) and beyond my control. This does limit historical science somewhat, but you only need as much statistics as your target precision. And, the quantity of historical data is more than enough to support many of the observations being claimed.

      I agree that the “further out we go in any observable dimension”, the harder is to make a precise statement. But this is a somewhat vague idea. How far is too far? Such statements are often based on faulty intuition (or motivated by a posteriori rationalizations). Frankly, 100,000 years ago is nearer to our direct experience than the 15 orders of magnitude that separate us from the proton.

      More importantly, this is only *ONE* aspect of the measurement problem. Other aspects are:
      (1) how much data do we have?
      (2) how many different independent kinds of measurements with non-overlapping assumptions validate the observation and?
      (3) how good is the technique and methodology?

      The more data we have, the better we can know something. The more constrained the system is by very different lines of evidence, the better we can know it. As an experimental scientist I look for the combination of all three of these. A volcanic eruption may have happened a hundred thousand years ago, but if I can match the isotopic composition of samples from the eruption site to hundreds of samples of volcanic dust in ice cores and other sources dated to the same time, if I can correlate these with a measured cooling in the tree-ring proxies, if all of these things identically match the signatures of modern volcanoes in the same record then I’m pretty darn confident it happened.

      Your 6th grade experiment anecdote is one measurement by one person. When I describe a scientific finding as being robust, I am typically referring to something supported by unimaginably many observations, repeated and scrutinized by many different researchers over many decades, and demonstrated through different and complimentary measurement techniques. THAT is how you achieve scientific confidence. THAT is how you catch mistakes. THAT is how science works.

    • Matt

      Hey Bri,

      Sorry about that… 🙂

      I was only trying to make a transitive argument. Even if you want to evoke “unknown unknowns” in cosmology, where there are “known unknowns” like dark matter/energy, you’d be hard pressed to cast doubt on the last 100k years of the earth’s history where we know glaciation periods like clockwork.

      I hope that didn’t come off too dismissive. I think our understanding of early Universe cosmology is pretty darn impressive. I just meant to say, that if there is a surprise lurking anywhere, I have a hard time fathoming that it’s lurking in the recent geological past. I’m glad to reword, if you have any suggestions.

      BTW, I read your article on Hume. Very interesting, been meaning to comment on it. I owe you a call soon.

  2. bapowellphysics

    No, you were not too dismissive — I’m just having fun. Your point is well taken, and in fact cosmology is a great example of a “historical” science. The big bang only happened once, and we’ve got only one CMB to observe. This leads to all sorts of fun statistical difficulties: for example, the largest scale fluctuations in the CMB can only be observed through 5 independent measurements (the largest scale fluctuations are given by the quadrupole moment in a multipole expansion, for which m takes on only 5 values). The relatively large standard error imposes a fundamental limit on our ability to accurately measure the size of the largest scale CMB fluctuations — we simply can’t do better because we’ve got only one CMB sky to observe. Much of the challenge of modern cosmology is working to deal with these kinds of puzzles: what the universe is telling you might be just noise, and we need to be careful not to interpret it as signal. So, yeah, using science to study the past comes with its unique challenges.

    I’m happy you gave the Hume article a read. Any comments or criticisms are always welcome!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s