I often wondered: What are the most basic assumptions I have to make before I can even start thinking about life, universe and the rest?

So far I have boiled them down to three:

  1. There is a world, a reality.
  2. I am part of this world.
  3. My senses provide me with a not entirely wrong, arbitrary yet mostly consistent, representation of reality.

Everything else should follow from those three.

My question now is: Are these three really all the assumption everyone must make before one can conduct science? Have I omitted something else?

For example, must I assume that if I do the one thing (e.g. let a rock fall) that this will be repeatable, or is the fact that it is repeatable already an insight into reality?

  • 1
    Given point 3, aren't 1 and 2 superfluous? (BTW: 3 is "non-conspiracy"?) PS: The example seems not OK to me, as QM says that it is not repeatable (with probability 1).
    – user3164
    Commented Apr 11, 2013 at 20:56
  • @Gugg 2 is there to state that I do not assume to be part of a computer simulation within a computer simulation which then is part of an unknowable reality. Without it, three would be false, as the input would be entirely wrong about reality.
    – k0pernikus
    Commented Apr 11, 2013 at 21:01
  • If 2 "is" required for 3, doesn't 3 then imply 2? And, therefore, wouldn't 2 then be superfluous?
    – user3164
    Commented Apr 11, 2013 at 21:13
  • 4
    More to the point, are any of your premises necessary for science? Suppose you are in a computer simulation part of an unknowable reality. Why would the practice of submitting hypotheses and subjecting them to falsifying experiments not classify as science, simply in virtue of the unknowability of the world outwith the computer simulation that the hypothesising agent resides in? See, e.g. Bas van Fraassen's Constructive Empiricism: plato.stanford.edu/entries/constructive-empiricism
    – Paul Ross
    Commented Apr 12, 2013 at 0:42
  • @PaulRoss I am not asking: "Would science stop being science if we were in a computer simulation." And even so, my answer would be "no, it still is science". The simulation then would be our reality. Yet the concept of computer simulation presupposes a concept of reality; so while I do not need to assume that I am in a computer simulation, I first have to assume that there is a world out there. Or to rephrase my original question in terms of your comment: What do I have to assume to be able to submit hypothesises and test them via falsifying experiments?
    – k0pernikus
    Commented Apr 12, 2013 at 1:11

4 Answers 4


The core of the scientific method is to have an observer, something to observe, a mechanism for generating and evaluating predictions about future observations, and a way to either select similar future observations or generate new similar events to observe. You then observe, model, test, observe, model, test, etc., as a way of improving your predictor's performance.

The constraints are therefore extraordinarily weak, though not exactly the ones you describe. Conditions need to be temporally stable enough so that you can run your observe-model-test loop many times before the rules completely change (slow drift in rules is okay, if you know to expect it); outcomes need to be sufficiently reproducible so that there is something to predict (but broad distributions are okay). Both randomness and extreme complexity can frustrate reproducibility; the more sophisticated of an observer/modeler you are, the more complexity you'll be able to tackle.

Vastly stronger than the constraints on the rest of the world are the constraints on the observer and modeler (possibly the same entity, though there is no reason it needs to be; the observer can use a modeler-oracle). Between the pair of them they need to be able to translate events into a representation of those events, detect which differences are purely stochastic and which are regular, and devise some sort of compact representation of such things that can be used to make future predictions. This is an immense amount of computational work, and it seems unlikely that in a badly chaotic time-varying universe that such entities could exist.

So the answer is probably very close to: if you exist and have adequate capacities to attempt to follow the scientific method, you can probably use it to find out at least some things.

  • The system of logical, semantical, methodological, and attitudinal ideals constitutes the institutional rationality of science even though individual scientists may more or less often fail to behave rationally. However biased the individual scientist may be, the institutional values are the basis for the institutional objectivity of science. Rationality is present only at the level of the institutionalized community. Objectivity, then, is a characteristic of a institutionalized community’s practice of science rather than of an individual’s. Commented Apr 12, 2013 at 17:46
  • There are conditions that a institutionalized scientific community must satisfy in order to qualify as rationally developing scientific knowledge:There must exist within the community recognized and approved forums or avenues for the criticism of theories, evidence, experiments, assumptions, and inferences. Commented Apr 12, 2013 at 17:48
  • Don't you need basic assumptions to surpass the Münchhausen trilemma about the impossibility to prove any truth? Commented Apr 12, 2013 at 18:00
  • 1
    @RicardoBevilaqua - Proof is too high a standard, so the trilemma isn't directly applicable. I believe any assumptions that generate what I stated will work. Also, the institutions of science serve as a check that the individual scientists are doing what they ought to. Almost all the work is done by individual scientists. I'm not sure why you think all the objectivity comes in when some people look over things at a high level and say, "Yeah, looks like you did things the way you were supposed to." Of course this helps the robustness, but you seem to be identifying it as the key step.
    – Rex Kerr
    Commented Apr 12, 2013 at 18:24
  • @RicardoBevilaqua - It's worth pointing out that if one wants the observer-modeler-environment system to be capable of doing enough philosophy to reassure itself that it actually is doing science, there are additional constraints (possibly including a satisfying resolution to Agrippa's trilemma). But the question didn't include this as a criterion.
    – Rex Kerr
    Commented Apr 12, 2013 at 18:45

I think requiring too much to to "start" conducting science would be too restrictive and even dogmatic, which science, by its very nature, tries to avoid. Assuming that what is asked is not what really distinguishes scientific practice from others (say, from philosphy), then we can really have very loose starting assumptions:

I need no more than "I" myself with the ability to use a capable language to help myself "understand" and build on.

All the rest would be arbitrary restrictions:

  • Why would you necessarily separate things as "I" and the rest? How can be so certain about sepration or independece? How does it help us to start doing science?
  • If there is such thing as reality, remember that, it will be covered (hopefully exhaustively, we don't know yet) as a subject matter of the scientific practice when I start to conduct it. I am not saying that reality, whatever it is, should be the sole interest of the science. But it will certainly be one of them. During the practice, I need the freedom to redefine reality (or any other concept) or drop it at all during the course of the practice.
  • Certain required attributes of "I" who will conduct science: Many of them, including the one particularly mentioned can be dropped as well. Do I really nead senses enabling a consistent respresentation of the reality? I think not as I beleive that I still have the chance to cope with, say, inconsitency. If to require a constraint or attribute on "I", i would choose freedom only. Being capable of conducting science does not count as a requirement in my opinion. However, ability to use a language, understand and develop/build-on imply certain attributes that "I" has to have. All this things suggest to me that we should focus on the qualifications and abilities of "I" so that "I" can conduct science.

Note that, many other practices other than science would require as much or the same set of basic assumptions as science does. Is this a problem? We are interested in starting point-only with this question... What separate science from non-science is a totally different question.

  • The question is “Are these three really all the assumption everyone must make BEFORE one can...” “BEFORE one can conduct science..” I think that the question is really about the foundations of reality, "all the rest" is not a detail in this question. Commented Apr 12, 2013 at 21:21
  • I wanted to stress out arbitrariness or potential restrictiveness rather than being "detail"... "Foundation of reality" is expected to be an outcome of the practice not its pre-condition I beleive... Or i get your remark wrong...sorry then...
    – mami
    Commented Apr 12, 2013 at 21:32
  • The tags are metaphysics and epistemology too, in this question they aren't arbitrary restrictions. Commented Apr 12, 2013 at 22:00
  • Can there be language with only "I"?
    – labreuer
    Commented Jul 17, 2014 at 19:43

The assumptions that you start with shape the future of your thought. Different assumptions different futures. There isn't a neccessary single starting point.

Your headline question talks about Science. In Greek antiquity science was done by natural philosophers. They didn't use the modern scientific method but at least supposed that the world was objective and rationally explicable as per your assumptions.

When in the body of your question you say how to comprehend 'life, universe and the rest'. It is clear that you think that the scientific method is the only way to go. But plenty of philosophers have disputed this whether theologically inspired (Buddhist & Islamic Philosophy) or rationally inspired, for example the phyrronic skeptics.


I have been reading the answers and have gotten VERY confused. This conversation clearly beyond me, but I do have know one assumption necessary for empirical science.

A coin if flipped n times, with the coin always landing on heads. The chance of this happening with an unbalanced coin is 1. However, the chance of this happening with a balanced coin is 1/(2^n). No matter how large n is, the chance that this coin is balance will never reach 0. Therefor, before an experiment can be preformed, an assumption must be made that probabilities lower than a certain level are considered impossible.

As I am reading this over, I am starting to wonder whether this is needed for science or if this is only needed for something to be proven by science. Either way, it is something you should consider.

You must log in to answer this question.

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