I once read that even a humble medieval peasant merely had to sneeze to cause events hundreds of years in the future:

"Not just Napoleon but the humblest medieval peasant had only to sneeze in order to affect something which changed something else which, after a long chain reaction, led to the consequence that one of your would-be ancestors failed to be your ancestor and became somebody else's instead. I'm not talking about 'chaos theory', or the equally trendy 'complexity theory', but just about the ordinary statistics of causation. The thread of historical events by which our existence hangs is wincingly tenuous."

-Dawkins, 'Unweaving The Rainbow', p20

But I don't know if that is actually the case, and certainly in real life we don't say that a minor event in the distant past caused significant changes in the present.

For example, we don't usually say that Aristotle having one food instead of another for breakfast one morning more than 2000 years ago caused me to write this question. However, by the butterfly effect, if he had something else, then probably the whole world would be different, because there would exist different people in the world than actually does now.

So, my question really is: Is it legitimate to talk of an action in the distant past causing something in the present? And also, have philosophers written about causality specifically relating to the butterfly effect?

  • 5
    It literally means that in some deterministic iterated systems, inputs that are very close together may evolve to outputs that are very far apart. There is no notion of causality except in the pop science literature.
    – user4894
    Sep 30, 2022 at 2:00
  • The butterfly effect is meant to illustrate that future events may be sensitively dependent on current conditions. If you wish to repeat an action and have exactly the same outcome, then the conditions must also be exact (including a peasants sneeze) or the outcome may be wildly different.
    – user59124
    Sep 30, 2022 at 2:04
  • 2
    It uses the colloquial "but for" notion of causality, so-called cause-in-fact, proximate or legal cause, which courts of law also use. But for your peasant sneezing (keeping all other conditions fixed), there would have been no black plague (let's say) and the world would not have looked the same. This is backed up theoretically by sensitive dependence of chaotic dynamical systems on slight variations in initial conditions. It is plausible that dynamics of human societies is, indeed, chaotic.
    – Conifold
    Sep 30, 2022 at 5:03
  • @user4894 in an iterated system, earlier states of the system directly cause later states. That's what causation means: A causes B if when you start with A at an earlier time, you can derive that B must be the case at a later time.
    – causative
    Sep 30, 2022 at 6:39
  • This example could be seen as one reason to argue that traditional concepts of "cause and effect" have no place in science, which replaces them with the concept of dynamical laws--see my answer here.
    – Hypnosifl
    Sep 30, 2022 at 8:51

4 Answers 4


It's interesting he says "the ordinary statistics of causation", seemingly ignoring 'Correlation does not imply causation'. I presume he has in mind Bayesian Inference, which is a good tool as justified by experience, but cannot give us direct access to reality, and is prone to a range of errors.

Causation is really a set of cognitive tools, for abstracting useful information from experience and shedding non-useful complexity, in order to make tractable predictions. When rolling a dice or flipping a coin, we have a simple enough system to ignore differences between rolls/flips. When it comes to human decision making, the idea of being able to access alternate histories where everything is held the same except one thing, is a useful fiction. I make the case here causation really comes down to our cognitive bias to narrate subjective experiences of objects, because of the social-structuring around intersubjectivity of our neocortexes functioning: Is the idea of a causal chain physical (or even scientific)?

We definitely can't in general compare our timeline, to one with an added peasant's sneeze in medieval times, because of sensitivity of complex systems to initial conditions. In this passage of Dawkins, he gives the example immediately preceding:

"Napoleon started it all. If it weren't for him, I might not be sitting here now writing these words ... for it was one of his cannonballs, fired in the Peninsular War, that shot off the arm of my great-great-grandfather, James Morris, and altered the whole course of my family history." -Desmond Morris opening lines of his autobiography 'Animal Days' (1979)

The nature of narrating, of explaining, is premised on the idea of 'If x had been otherwise' to give x a role in the story. This is the process of abstraction, the sifting of experiences for what is transferable, intersubjective, explanatory. I relate this to causation in the terms of heuristic explanatory overlays, which group phenomena ultimately reducible to physics, into conceptual units, to make tractable predictions. So, the layer of character and intentions, is a far more efficient way to predict another human, even though we think full knowledge of their atomic states and their environment could with the right computation lead to more complete predictions. More here: Why do compatibilists believe that whether we act freely is independent of whether or not determinism is true?

Aristotle's breakfast, would be a poor choice of imagined variable, that would lack explanatory power. Humans have historically made many bad choices fir explanatory variables. Athenians considered the patterns in a sacrificed animal's entrails a critical predictor of the future outcome of a battle. We can see the Pathetic Fallacy as the overspill of brain being adapted to understand emotional and intentional states on to the world. And we can see science, as the systemising of attempts to minimise cognitive bias, post hoc reasoning, and to ensure consilience and convergence of evidence.

You ask "Is it legitimate to talk of an action in the distant past causing something in the present?" and I would say, it can be, if we make good abstractions. Our telling of history is not simply an account of the past, but a gathering of regathering from information about the past, to bring insight into now. Every generation has to seek it's own insights. Discussed here: Do historians have responsibility in how they decide to depict something? Good history seeks valid inferences from the past, which can inform us now.

I make the case emergentism is a mode of causal narrative, opposite in character to reductionism, here: What's the "opposite" of emergence? The difference in the outcomes from complex systems with slightly different initial conditions is chaotic - the interaction of three similar blackholes can cause different outcomes based on different initial conditions below the Planck scale, fundamentally limiting determinism. But emergence is powerful because it seeks likely similar outcomes regardless of initial conditions. Human character is one such, we abstract things about behaviours that we think will cause similar outcomes in many possible futures. Complexity is the whole field, that includes both chaos and emergence.

A butterfly flap is a much poorer predictor of a hurricane than ocean surface temperatures, regional solar gain, and measures like that which will tend to lead to similar outcomes from many initial butterfly states.


The "Butterfly effect" that you're referring to is an thought experiment from chaos theory/complex systems of how tiny disturbances in the initial conditions of a system can cause vastly different outcomes.

Now Dawkins pretends to just talk about snowballing effects over time, like how idk a 1% interest rate over 400 years would mean ~50 times the initial value or ~20.000 times the initial value for 1000 years and for 4000 years it would be ~ the initial value. So if you wait long enough even small difference accumulate. But what he describes might be a bit fast and still be an example of chaos theory and/or complex systems.

Or idk picture a satellite that is just slightly too slow to keep up it's stable orbit and is slowly descending down to earth in spirals. It might take thousands or more revolutions around the planet before it hits the surface and even the tiniest difference in it's initial condition can lead to making a rotation more or less meaning it's impact point might be right next to one with similar conditions or at the other end of the globe.

Now you can simulate such systems and control for any but one parameter, so that you can put the blame on this one parameter. So theoretically you can make the argument that the butterfly has caused the tornado because without the butterfly the tornado would not have formed. On the practical end of things however you'd be completely lost as to what has actually caused it because once the changes in the initial conditions get so tiny the culprit could literally be anything and it's not that you could blame them for events caused by their action, which they had not the faintest chance to foresee and mitigate.

So either way you'd probably would need to ignore the "original culprit" and look for something that is closer to the actual result. Where you can identify that something "goes wrong" and where you can identify someone or something that is instrumental in it and able to affect the result and to that agent you'd tag the "cause".

So technically speaking the big bang is to blame for everything that is happening in the universe but identifying this "cause" isn't really any helpful.


Dawkins doesn't quite understand the principle here. The Butterfly Effect refers to a property of non-linear deterministic systems (NLDS): sensitivity to initial conditions. Sensitivity to initial conditions means (quite literally) that very small differences in starting state can create progressively significant differences in later states of the system, so that it becomes impossible to predict the future state of a system without absolutely perfect knowledge of the initial state. So in that sense, yes: if we don't know that a butterfly has flapped its wings in China, we cannot predict what the weather will be like on some future day in Kansas. That tiny movement of air changes the initial state of the system.

However, another equally important property of NLDS is self-similarity, usually across scales of measurement. Even though small differences in initial state produce significant differences in later states, the system as a whole will run through some recognizable pattern of states that is itself comparatively stable and predictable. For example, we cannot predict when and where a specific tornado will land in Kansas — because a few butterflies are flapping their wings behind our backs — but we know there is a season when tornados will land in Kansas, almost to a certainty. Weather systems keep cycling back through the same patterns, and even though the pattern manifests differently in each cycle, it's still a pattern.

Dawkins talks about the issue in purely stochastic terms, as though a random peasant's sneeze might have any outcome whatsoever: i.e., it might result in aliens invading the earth, or dinosaurs coming back to life, or even (unlikely as it seems) causing Dawkins to write serious philosophy. But no... Human life (individually and collectively) is cyclical and self-similar. A random peasant's sneeze will make the exact state of future human existence unpredictable, sure, but the warp and woof of human society and civilization will not change greatly on a macro-scale. We might have had different wars, different rulers, different plagues, or different paces or locations of innovation, but wars, rulers, plagues, and innovations would have occurred in recognizably similar ways.

It annoys me that Dawkins has the temerity to dismiss mathematical facts as 'trendy ideas'. Just to be clear about it...

  • "Dawkins talks about the issue in purely stochastic terms, as though a random peasant's sneeze might have any outcome whatsoever" How do you figure? The quote just says a different person than you might have been born, i.e. every fertilization event from medieval times to today would not involve the same sperm and egg in the world where the peasant sneezed vs. the world where he didn't. There's nothing there about "any outcome whatsoever", it doesn't rule out the possibility that various broader statistical patterns (say, population growth) might likely have been about the same in both worlds.
    – Hypnosifl
    Oct 1, 2022 at 2:16
  • @Hypnosifl: I'm just reading what he said in the given quote. The 'ordinary statistics of causation' means stochastic processes, which are (mathematically speaking) unbounded and prone to runaway accretions. Chaotic behavior means (mathematically speaking) recursive, self-similar, cyclical processes. Since he explicitly rejects the second, he can't mean anything other than an open ended-random process. He may not understand the math involved here (he's a biologist by training), but that's not really an excuse. Oct 1, 2022 at 4:18
  • He doesn't "reject" chaos theory, he just says you don't need to consider it to get the conclusion that you probably wouldn't have been born in an altered history where something small changed long ago. And his "ordinary statistics of causation" statement doesn't say larger statistical patterns (like population growth) would be totally different if something small changed, he's very specifically talking about the issue of whether a person with your exact genetic code would still get born.
    – Hypnosifl
    Oct 1, 2022 at 9:30
  • @Hypnosifl: Dawkins said what he said, and he doesn't need an apologist. He's merely wrong. Being wrong is excusable; let it go. Oct 1, 2022 at 14:30
  • It seems straightforward he's saying "the ordinary statistics of causation" are sufficient to conclude that any small change in medieval times would mean "one of your would-be ancestors failed to be your ancestor", do you actually disagree with that statement the way I wrote it, or do you agree with it but think Dawkins meant something different than my statement? If the latter you should be able to defend your reading by pointing to some phrase or sentence that suggests he's talking about something else, you can't just point to the phrase about "ordinary statistics" since mine used that too.
    – Hypnosifl
    Oct 1, 2022 at 14:52

To add slightly to Ted's response, note that the cyclic patterns he mentions are ones in which the outcome of the previous cycle perturbs the development of the next cycle i.e., the system possesses feedback. It is this property which allows perturbations to grow as the cycling proceeds, and renders the system chaotic after a certain number of cycles when the perturbations "blow up".

Note also that for those perturbations to persist and grow, there must be almost no dissipation in the system, or else those perturbations will get "erased" at the end of each cycle- and they will not get an opportunity to propagate and grow.

Now note that a natural process like a wing flapping in air is, on the scale length of the earth as a whole, heavily dissipative: air has viscosity, and after a short while and over a short distance the perturbations created in the air by the butterfly are rubbed away into nothingness long, long before those perturbations have any opportunity to propagate and grow by feeding upon themselves.

In this sense the "butterfly wing effect" is a canard.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .