You are on page 1of 7


How do we know that we know

what we know?
Rick Marshall
Keele University, Keele ST5 5BG, UK

‘How Science Works’ is now the focus of the national science specifications
in English schools. This article is a brief introduction to the philosophy of
science from the Greeks to the present day that underpins this notion.

From September 2006, the emphasis for teaching title: ‘How do we know that we know what we
science at GCSE is on ‘How Science Works’. know?’.
This focus will also shape the revisions of the
AS and A2 specifications from September 2008. My truth is a dream unless my dream is
Ideas about how science should be done and why true; George Santayana
it works have changed over time. In essence,
Ridley in his book On Science [1] suggests that
this article is an introduction to the philosophy
there are many types of truth, which those of
of science that underpins the notion of ‘how
science stand beside:
science works’. The author is a physicist not a
the revealed truth of religion;
philosopher. The short bibliography at the end
the persuasive truths of the social sciences;
of the article represents the source of most of the
the demonstrable truths of mathematics;
ideas presented here.
and the magical truth associated with the
The article itself originated as a presentation
to students taking the International Baccalaureate
For the latter—poetry, music, the fine arts
(IB) diploma and formed part of their Theory of
etc—uniqueness is the key virtue. In contrast,
Knowledge course. It consists of a brief summary
the scientist focuses on the recurrent, and the
of the contributions of nine philosophers and/or
scientists to the debate: Aristotle, Roger Bacon,
Francis Bacon, Galileo, Newton, Wittgenstein,
Popper, Kuhn and Gödel. ‘All men by nature desire to know’,
The title of this article derives from Coperni- Aristotle (c 384–322 BC)
cus’s assertion that ‘To know that we know what Francis Bacon (who we shall look at in a moment)
we know, and to know that we do not know what effectively described Aristotle’s method of doing
we do not know, that is true knowledge’. The dis- science and finding what was true in the following
cussion will be less dogmatic than Donald Rums- way. Gather a group of clever people and
feld’s more recent version of Copernicus’s apho- encourage them to argue. If they are clever enough
rism, ‘. . . as we know, there are known knowns; then the truth must emerge!
there are things we know we know. We also know The Greeks seem to have been the first to pro-
there are known unknowns; that is to say we know pose that Nature is logical. Greek science however,
there are some things we do not know. But there was ‘non-instrumental’—they did not carry out ex-
are also unknown unknowns—the ones we don’t periments. Any evidence they used came from hu-
know we don’t know’. Hence the more tentative man observation and their conclusions relied upon

238 PHYSICS EDUCATION 42 (3) 0031-9120/07/030238+07$30.00 © 2007 IOP Publishing Ltd

How do we know that we know what we know?
pure thought/reasoning. They used deduction to the correct way to reason on the basis of the
establish the truth of the matter. Deduction de- evidence derived from the world about you. Thus
rives the consequence(s) implicit in premises; e.g., specific consequences (conclusions) were to be
First premise All teachers are mortal inferred from general observations (premises);
Second premise Susan Jones is a teacher e.g.,
Conclusion Susan Jones is mortal
Nothing is necessarily assumed about the (i) All students from (name your own school or
actual truth or otherwise of the premises—they are college) work very hard at their studies.
assumed to be true for the purposes of argument. (ii) Amongst this crowd of young people at a
There are obvious flaws in this type of reasoning. night club this is a student from said school
If you the reader are not a teacher, or like the or college.
author of this article, are no longer a teacher, then (iii) Therefore this student works very hard at his
you might further conclude that you are not mortal, or her studies.
i.e. you must be immortal!
The Greek philosopher Zeno came up with Francis Bacon (1561–1626)
other odd conclusions based upon the use of
logical deduction, summed up in his so-called Francis Bacon is noted for two contributions to
paradoxes. Perhaps the most famous concerns a our theme. First, he was a great publicist for the
race between Achilles and a tortoise. If Achilles central role of experiment, though it seems that
agrees to give the tortoise a head start, then Zeno Francis Bacon himself never indulged in much
can prove that Achilles can never win such a race. actual scientific work. He was reiterating more
He argues as follows. effectively scientists such as Bernardino Telesio
In the time it takes Achilles to get to where who is on record as advocating the carrying out of
the tortoise was at the start of the race, the tortoise experiments as early as 1509, and William Gilbert,
will have moved forward. Then, in the time it who in his book de Magnete published in 1600,
takes for Achilles to cover this extra distance, the was the first person to set out in detail the essence
tortoise will have again moved on some more, and of a scientific method based on ‘perform(ing)
so on. The argument can be repeated over and over experiments’ as well as ‘thinking’. It was a course
with the necessary conclusion that Achilles can of action he exploited to great effect. Francis
never catch up with the tortoise let alone overtake Bacon’s influence came to fruition 34 years after
it and thus win the race. This logical argument his death with the inauguration of the Royal
offends both our ‘common sense’ and experience Society in 1660 for the ‘promotion of experimental
when carrying out the experiment by effectively philosophy’.
arranging such a race. Although not as fast as Secondly, when it came to ‘thinking’ he re-
Achilles we could substitute for him. But so versed Roger Bacon’s inferring specific conse-
strong was the Greeks’ attachment to the power quences (conclusions) from general observations
of pure reasoning that when logic disagreed with (premises) and observed that all we can reasonably
experience, they concluded that their perception do is to generalize the particular. This is known as
of experience must be at fault, because their logic induction.
could not be faulted! This position can be summed An example of an inductive argument is as
up as ‘mind over matter’. follows.

(i) The Sun has risen on every day in recorded

‘Cease to be ruled by dogmas and history.
authorities: look at the world’, Roger (ii) Tomorrow is another day.
Bacon (13th century) (iii) Therefore the Sun will rise tomorrow.
To avoid the peculiar conclusions of unbridled
reasoning alone Roger Bacon’s (not to be confused In other words the fact that the Sun has risen
with his namesake Francis Bacon) succinct every day so far, and there are no known counter-
command reversed the Greeks’ position of ‘mind instances or exceptions, and we have no reason to
over matter’ to become ‘matter over mind’. suspect it will change the way it behaves, implies
However he still advocated logical deduction as that it will always rise on every day in the future.


R Marshall
There has been one morning when the Sun so-called Classical Physics was developed. With
did seem to be behaving in a very odd way [2]. the advent of quantum physics on the one hand
It was on 16 July 1945 when a truck driver in and relativity theory on the other at the turn
New Mexico, USA, reported that he saw the Sun of the twentieth century, philosophical problems
starting to rise at 4 a.m. Then it went down again, reasserted themselves once more (although the
and came up an hour later. What the truck driver problem of induction had never really gone away).
had unknowingly at the time witnessed was of
course the testing of the first nuclear explosion. ‘You see how deceptive vocabulary is
There is no logical justification for an . . . and how our way of speaking colours
inductive leap of faith. Maybe inductive arguments our way of thinking’ [3]
work because they have always worked up to now.
Physics since 1900 has raised many new philo-
This line of reasoning was dubbed the ‘scandal of
sophical questions, not least the problem as to
philosophy’ by Bertrand Russell, as it sought to
whether there are limits to how much we can know
justify induction by using induction! So the use
about the Universe in which we live. In 1921
of induction is strictly illogical, but this does not
Wittgenstein asserted that there must be something
necessarily mean it is irrational—just non-rational
in common between the structure of a sentence and
the fact that the sentence asserts.
The outcome of inductive arguments can be
So for Wittgenstein the limits of thought
just as nonsensical as deductive ones. At the time
are the limits of its language, and we can only
of writing this article, at every birthday I have had
understand our Universe by exploring what can
so far I have been less than sixty-four years old.
be said, and what cannot be said, about it.
Using an inductive argument I would conclude that
George Orwell in his novel Nineteen Eighty-Four
I will always be younger than sixty-four. I have
(published in 1949) understood the implications of
proved again that I am immortal!
this insight. He realized that if language could
The success of induction as a method of change, then previous thoughts could no longer
scientific reasoning implies that scientific laws do be thought. The whole aim of ‘Newspeak’ was
not vary as time goes by and it is this regularity to narrow the range of thought. In the end the
that is at the heart of being able to make reliable new political order could make political dissent
predictions (‘the Sun will rise tomorrow’) using (‘thought crime’) literally impossible, because
inductive arguments. there would be no words in which to express it.
Thus different languages imply different
Galileo (born 1554) and Newton (1642 (the knowledge. For the poet Ezra Pound, ‘The
year Galileo died) to 1726) sum of human wisdom is not contained in any
Galileo’s name is associated with the crucial one language, and no single language is capable
importance of the controlled experiment—the ‘fair of expressing all forms and degrees of human
test’ beloved of GCSE course work. This is comprehension’ [4].
quite a change from simply observing Nature There are more than 6500 different languages
to actively interrogating Nature. Galileo, and, around the world today, in other words, 6500 ways
par excellence, Newton, described what was of knowing about the world. But according to
going on in mathematical terms—they devised a research being carried out at the London School
mathematical model of the behaviour. Science of Oriental and African Studies Dying Language
strives to be precise, unambiguous, logical and Research Unit, they are being lost at an average
universal. Natural language is none of these rate of one every two weeks.
things; hence scientists have become addicted Max Born in his textbook Atomic Physics
to mathematics. In passing, this raises two explained why the new quantum physics was so
interesting questions: Why should the world be baffling. ‘Ultimately the origin of the difficulty lies
mathematical? And why should our brains be in the fact (or philosophical principle) that we are
capable of doing mathematics? compelled to use the words of common language
Experimentation and mathematical modelling when we describe phenomena; not by logical or
facilitated the explosion in knowledge during the mathematical analysis, but by a picture appealing
two centuries from 1700 to 1900 AD in which to the imagination. Common language has grown


How do we know that we know what we know?
by everyday experiences and can never surpass apparent strength is in fact a fatal weakness. If
these limits. Classical physics has restricted itself every aspect of historical change can be given
to the use of concepts of this kind; by analysing a Marxist interpretation, then we have no way
visible motion it has developed two ways of of distinguishing it from any other theory that
representing it by elementary processes; moving makes the same claim. A Freudian and a Jungian
particles and waves. No other way of gaining a psychologist can both account for the same facts
pictorial description exists, so we have to apply in their own way. Freud may in fact be right, but
these even in the region of atomic processes where we have no way of knowing!
classical physics breaks down’ [5]. Newton’s theory of gravitation on the other
So a representation of the world in thought is hand not only accounts for phenomena we are
made possible by logic. However, we do not know familiar with (e.g. tides), but also made predictions
whether these (logical) conclusions necessarily that might or might not have turned out to be
represent any actual state of the ‘real’ world. Thus the case. For example the orbit of Uranus could
logic can only describe reality. For example not be accounted for in terms of the net effect
a picture represents reality using certain kinds of the gravitational forces due to the then known
of symbols. Similarly mathematics. Similarly planets. The slight discrepancy could be removed
language. Scientists are trying to describe the if another as yet unknown planet existed to provide
world. They try to describe it using some scientific just the right gravitational force. Thus the mass
or mathematical language. However, there is a and orbit of this hypothetical planet were predicted
problem about whether what their theories and and the planet we now call Neptune was first
laws say about the world corresponds to the way observed in 1846, exactly where Newton’s theory
the world ‘really’ is. predicted it should be. Had Neptune not been
Wittgenstein has turned Descartes’ famous there, then Newton’s theory could not have been
dictum ‘I think; therefore I am’ on its head. the whole story.
For Wittgenstein, you only know what you think Although being testable is a necessary
because of language and language is a public thing condition for a theory to be scientific, it does
that has to be acquired. not seem in practice to be a sufficient condition.
Failed predictions do not automatically lead to
Induction avoided; Karl Popper the downfall of a theory. In the 1670s it was
(1902–1994) observed that the eclipsing by the planet Jupiter
There are many apparently successful theories of its moons seemed to happen at inconsistent
about aspects of our world and the way we live. times. If their motion was governed by Newton’s
For example Isaac Newton’s theory of gravity, theory of gravitation then it could not have had the
Karl Marx’s theory of history and Sigmund simple inverse square law form which fitted all the
Freud’s theory of the unconscious all have their evidence up until then.
committed adherents. All are based (at least in The discrepancy seemed to depend on
part) on observation. So are they all scientific whether the Earth was on the same side or the
theories? Karl Popper in his book the Theory of opposite side of the Sun to Jupiter. However,
Knowledge [6] argued that the answer was ‘no’. Newton’s theory would not need modifying or
To be scientific a theory must be falsifiable— rejecting if an alternative idea proved to be correct.
i.e. capable of making a prediction that could be This was the radical proposal that the inconsistent
tested. For Popper, no amount of data can confirm times were a consequence of light only having a
a hypothesis, but one counter-instance can refute finite speed. The maximum time difference was
it. Towards the end of the Weimar Republic a due to the extra time that light needed to traverse a
pamphlet entitled 100 Authors against Einstein distance equal to the diameter of the Earth’s orbit.
was published. Einstein responded in a very In the late 1600s this distance was believed to be
Popperian way—‘if I were wrong, then one would 2.77 × 1011 m. The maximum discrepancy in
have been enough’. the timing of the eclipses was 22 min (=1320 s),
Both Marx’s and Freud’s theories provide an implying a light speed of 2.77 × 1011 /1.32 ×
explanation of everything about social change or 103 = 2.10 × 108 m s−1 (somewhat lower than
the unconscious respectively. For Popper, this today’s accepted value of 3 × 108 m s−1 , due to


R Marshall
the poor value for the diameter of the Earth’s orbit any significant difference to the development of
used at the time). It was over 50 years before human intelligence. Aristotle clearly was not
this value for the speed of light was corroborated. stupid. He said many things we consider to be
In the meantime however, Newton’s theory was perfectly reasonable. So why for example were his
not rejected (maybe because Newton himself was ideas about falling bodies so obviously wrong for
happy to accept the finite speed of light hypothesis today’s scientists? This thought was the starting
rather than abandon his theory of gravity!). point for Thomas Kuhn’s thinking that culminated
After its discovery, the orbit of Mercury could in his celebrated work the Structure of Scientific
not quite be accounted for by Newton’s theory, Revolutions [8]. Kuhn argued that to understand
but in this case no explanation in terms of a new Aristotle, we should look at the universe through
planet, or an odd shape for the Sun, could account Aristotle’s eyes, in terms of the beliefs etc accepted
for the discrepancy. No planet could be found, and in his time. Then what he said is at least rational.
the Sun’s shape was too symmetrical. However, These shared assumptions, what counts as science,
Einstein’s general theory of relativity (which is and how to do it, Kuhn called a paradigm. Sticking
essentially a theory of gravity) did account for with the accepted paradigm is what most scientists
Mercury’s orbit and for this reason, amongst do most of the time. Kuhn calls this normal
others, it came to replace Newton’s theory of science.
gravity. Eventually if too much evidence is accu-
Notice that in Popper’s view of the develop- mulated that just does not fit into a prevailing
ment of scientific theories, the problem of induc- paradigm it will be rejected in favour of a new
tion is irrelevant. He does not rely on arguing from paradigm. So it is not just one counter-instance
the particular to the general. In Popper’s view the- that brings down a previously successful theory.
ories get better and better by an incremental pro- The period of searching around for a better the-
cess of dealing successfully with predictions that, ory (i.e. one that accounts for everything the failing
had they not been confirmed, might have led to theory can explain as well as the things it does not
their rejection. But Popper’s view also seems con- explain) Kuhn calls a period of revolutionary sci-
sistent with the view that all statements (no matter ence. The most prominent examples are Einstein’s
how wacky) are potentially true until falsified. But theories of relativity replacing Newton’s theory of
then some theories do start out at the time as very gravity and laws of motion, or quantum theory su-
odd ideas indeed! Could light possibly travel at perseding the so-called classical physics. The key
200 million metres per second? thing is that the new paradigm is incommensurable
Since 1986 string theory has become the with the old. e.g. although both Newton and Ein-
dominant fundamental theory of elementary stein talk about mass, they understand quite dif-
particle and fundamental forces if measured in ferent things by the concept. For Newton, mass
terms of the numbers of physicists working on and energy are quite distinct. For Einstein, mass
it and the scientific papers they produce. 10500 and energy are essentially alike. The debate over
is the number of possible string theory solutions, string theory vis à vis quantum electrodynamics
but zero is the number of testable solutions it has looks like just such an episode of revolutionary sci-
come up with so far. A recent editorial in the ence concerning the way we think about elemen-
scientific journal Nature stated ‘String theory is far tary particles and fundamental forces. A paradigm
from experimental verification . . . the prime goal change may be on the horizon, if string theory can
must be to turn this project ultimately into testable be developed to the point of being testable.
science’ [7]. So the change from old to the new paradigm
is a discontinuous ‘jump’. Theories do not start
Paradigms rule; Thomas Kuhn small and grow to get better and better for ever.
In terms of ideas we have come a long way from Are there logical limits to scientific
Aristotle and the Greeks to where we are now. knowledge?
But it is a blink of an eye in terms of human We have seen that Wittgenstein explored the role
development. The time span since the Greek of language (used to express logical propositions)
civilization is far too short for evolution to make in what we can learn about the world. Gödel came


How do we know that we know what we know?
to realize in 1931 that there are limits to what logic A theory of everything is impossible
itself can reveal about the world.
Some see the goal of physics as the establishment
His incompleteness theorem is perhaps the of a theory of everything (ToE), from which all the
most significant result of 20th century logic. It rest of physics can in principle be derived. If you
states: in any self-consistent logical system there assume that a ToE is possible and that it is based on
are statements that can be made which can be a set of fundamental laws (conservation of energy
neither proved nor disproved within the framework etc) then Gödel’s incompleteness theorem implies
of the system. that there are new truths to be discovered that will
Thus no logical or mathematical theory can need the discovery of additional fundamental laws
include methods of reasoning strong enough for for their explanation, so the original theory cannot
the proof of its own consistency. Consistent means explain everything.
that you must not be able to prove results that
are false, e.g. proving 1 = 2 using the rules of
arithmetic. ‘We have to know, so we will know’ [9]
So any formal system of axioms (i.e. assump- Can we ever be absolutely certain about something
tions and rules of procedure) is incapable of prov- in the physical world? For example, do atoms
ing some truths, and no formal system can prove exist? In effect scientists believe in the existence
all truths. For example: given the rules of arith- of atoms—it is part of their shared paradigm.
metic, there are arithmetic statements that you can- A belief is stronger than pure faith. Sebastian
not prove or disprove simply using the rules of Faulks put the distinction so well in his novel
arithmetic. One example is the Goldbach con- Charlotte Gray ‘. . . belief (is a) logical conviction,
jecture that every even number is the sum of two while faith . . . (admits) doubt . . .’ [10]. Or to
prime numbers. Try it! Every example you chose paraphrase slightly the literary critic I A Richards:
works, but you cannot prove using the rules of ‘We (accept what) scientists (say) because they can
arithmetic that it is true for all prime numbers, no substantiate their remarks, not because they are
matter how large. eloquent and forcible in their enunciation’.
Another example of Gödel’s theorem is this We have evolved partly by a process of asking
linguistic conundrum. Consider the following: questions and finding answers. Scientific answers
Every man is either a knight or a knave. have a unique hallmark—they do not represent the
Knights always tell the truth but knaves never absolute truth of the matter. They are provisional
do. beliefs subject to continuous critical revision. For
now they represent the most reliable conclusion
Someone (you do not know if he is a knight or
that can be drawn on the basis of the available
a knave) declares ‘You will never believe that I am
evidence. Thus science is more a way of thinking
a knight’.
than a collection of knowledge. Its real success is
Is the speaker a knight or a knave?
not so much in providing answers, but rather in the
It is impossible to answer this question using way these answers can be found or ‘how science
the information given. works’.

(i) You cannot believe that he is a knight Received 28 November 2006

because he would then always tell the truth,
and what he says must be a lie.
(ii) If you believe him to be a knave and thus References
telling a lie, he is making a statement that [1] Ridley B 2001 On Science (Oxford: Routledge)
you do hold true! [2] Weisskopf V 1985 Forty Years After: Thoughts of
(iii) You thus stay undecided, which means that a Nuclear Witness (San Francisco, CA:
the speaker did tell the truth and was a knight [3] Carrière J-C 2006 Please Mr Einstein (London:
after all—but if he is a knight . . . (back Harvill Secker)
to (i)). [4] Crystal D 1999 Ezra Pound Guardian (25th Oct.)


R Marshall
[5] Born M 1962 Atomic Physics 7th edn (Gasgow: [3] Flew A 1975 Thinking about Thinking (London:
Blackie) (revised with R J Blin-Stoyle) Fontana/Collins)
[6] Popper K 1959 The Logic of Scientific Discovery [4] Macrone M 1998 A Little Knowledge (London:
(London: Hutchinson) (originally published in Ebury)
German in 1934) [5] Casti J L 1990 Paradigms Lost (London: Abacus)
[7] 2006 Editorial Nature 443 482 [6] Casti J L 2001 Paradigms Regained (London:
[8] Kuhn T 1962 The Structure of Scientific Abacus)
Revolutions (Chicago, IL: University of [7] Ridley B 2001 On Science (Oxford: Routledge)
Chicago Press) [8] Brockman J (ed) 2006 What We Believe But
[9] Inscription on the mathematical physicist David Cannot Prove Pocket Books (New York:
Hilbert’s gravestone Simon & Schuster)
[10] Faulks S 1999 Charlotte Gray (London: Vintage)
p 420
Rick Marshall spent two thirds of his
Further reading full-time working career in universities
and one third in schools. He now
[1] Koestler A 1964 The Act of Creation (London: contributes to the University of Keele
Hutchinson) Physics Enhancement Course and the
[2] Paulos J A 2000 I Think Therefore I Laugh 2008 revision of the Advancing Physics
(London: Penguin) A-level course.