A Brief and Incomplete

History of the
Philosophy of Science

Based largely on John Losee (1993) A Historical Introduction to

the Philosophy of Science. Oxford: Oxford UP.
 Plato (427 - 347 BCE)

Platos epistemology denigrated scientific

knowledge (knowledge of natural and
material regularities)such knowledge
was not of the true reality, but merely of
shadows in the cave
Most important for Plato was knowledge
of the Forms, the abstract entities which
define the moral and metaphysical
structure of the universe
Knowledge of the Forms was to be
gained not via observation and
inference, but through pure reason and
philosophical discourse
 Aristotle (384-322 BCE)

Inductive-Deductive Model:
General Principles

(1) Induction Deduction (2)

Observed Phenomena

From observations one proceeds by inductive inference (1) to

General Principles which explain the observations in virtue of
the fact that those same observations can be deduced (2)
from the principles
 Aristotles I-D Model
Induction: enumeration, direct intuition
Deduction: categorical logic
Aristotle required that the General Principles be
at least as evident as the observations
ultimately, they should be self-evident or
necessary truths
The motivation here is to avoid arriving at claims which
describe only accidental regularities
Rather GPs should be self-evident necessary truths
reflecting the essences of objects and relations in nature
This is related to the issue of the nature of laws
The problem is that it is hard to see how we can get to
necessary truths via induction
This can be seen as an outcropping of the problem of
induction
 Aristotles Four Causes
Material Cause: substance which undergoes a process
Formal Cause: general conditions required for, and pattern
or form of, process
Efficient Cause: immediate conditions which precipitate
the process or bring the object into being
Final Cause: the purpose or end for which the process
occursx occurs in order that this sort of explanation is
called teleological
To fully explain a phenomena, each of its four causes must be
explained
Whereas now we focus primarily on a combination of the first
three and often try to eliminate teleological explanation,
Aristotle saw the final cause/teleological explanation as most
important to understanding the nature of things
this raises the issue of the nature of laws and causation
 Pythagoreanism
Pythagoreans revered numbers and mathematical
relations to the point of mysticism
The real is the mathematical patterns and harmonies
discoverable in nature
Describe the mathematical structure of a phenomena and
you have knowledge of its essence
This contrasts with Aristotelianism in that it focuses on the
formal cause to the exclusion of the others, it especially
neglects final causes
Our current mathematical physics is, indeed, quite
Pythagorean
Pythagoreanism resembles Platonism in that it gives pride
of place to abstract entities (numbers) and their relations,
but it does not have much to say about moral Forms
 PythagoreanismProblems
Given their knowledge of the Pythagorean Theorem, and
the fact that they conceived of all numbers as ratios,
certain quantities were thought to be mysterious and
incommensurable (immeasurable or unable to be compared
with known quantities, what we would call the irrational
numbers), e.g.,

Of course, any good mysticism has to have mysteries

A further problem is that the same phenomena can be
described by various different, but observationally
equivalent, mathematical models
This is the issue of underdetermination of theory by evidence

This puts in question the ideal that mathematical

description gets at the true nature of things
 Pythagoreanism and Saving
Appearances
The tension between Pythagoreanism and the
possibility of observationally equivalent
mathematical descriptions was especially acute in
astronomy
It was well knowne.g., by Ptolemy
(100-178)that motion of the planets
could be equally well accounted for by
various mathematical models
The question then becomes whether
to view mathematical description as
revealing underlying nature or as
merely providing a convenient description of the
observable phenomena (saving the appearances)
This leads to the issue of the observation/theory
distinction and realism/anti-realism
 Three Models of Planetary Motion
Epicycle/Deferent Moving Heliocentric
P Eccentric Circles
epicycle eccentric
c
P P

a b E
c

E E S

deferent

Planet P revolves around Planet P revolves around Planet P and Earth, E, both
point c, while c revolves point c, while c revolves revolve around Sun, S
around Earth, E around Earth, E
When E passes P
When P passes through a When c and P are on retrograde motion of P
and b, P will appear to opposite sides of E occurs
move backwards against retrograde motion of P
the night sky seen from E occurs
(retrograde motion of P)

These models also account for variations in speed and distance relative to E. Further epicycles, eccentrics,
deferents, and equants can be added for greater precision. Of course, none of these is correct
 Saving Appearances
Especially in astronomy, a tradition evolved of
not claiming reality for the mathematical
modelsthe task for the astronomer is not to
hypothesize about the unobservable nature of
things, but to provide convenient and
observationally adequate models
This is a form of anti-realism
Relates to the observation/theory distinction

This is very similar to naive positivism,

operationalism, and current constructive
empiricism
Ptolemy was inconsistent on this issue, usually
stressing Pythagorean realism, but sometimes
weakening his claims to saving the appearances
(though he never considered heliocentrism
plausible)
 Saving Appearances
and Oneself
The famous heliocentrists, Copernicus (1473-
1543), Galileo (1564-1642), and Kepler (1571-
1630), each had Pythagorean commitments
each held to the reality of his model, and
each was strongly motivated, not just by data and Copernicus
observation, but also by strong mathematical
aesthetics, a desire to find certain kinds of harmonies
in nature
This raises the issues of scientific revolutions and the
rationality of theory change
Yet each of the heliocentrists was advised to
present his work as a mere saving of appearances
to avoid persecution from the Church. Galileo did
not take great pains to hide his commitment to the Galileo
reality of heliocentrism. As a result, Galileo was
brought before the Inquisition and forced to
recant, spending the his last eight years under
house arrest.
Again, the issues of rationality, revolutions, and social
forces

Kepler
 Atomism
A further metaphysical/explanatory picture that can be
contrasted to Pythagoreanism and Aristotelianism is
Atomism
Proponents included Leucippus (490-430 BCE) and
Democritus (460-360 BCE)
The general approach was to explain observed qualitative
and quantitative changes by reference to quantitative
changes at a more elementary level of organization
This, too, neglects the Final Cause, and is thoroughly
materialistic, thus it is antithetical both to Aristotelianism
and to Platonism/Pythagoreanism
One difficulty is to avoid simply attributing to the atoms the
very property to be explained at the macro level. Doing so
would create a circular explanation, which is to say, no
explanation at all (e.g., day-old coffee is bitter because it
has acquired large numbers of bitter atoms)
This raises the issue of the nature and quality of scientific
explanation as well as the observation/theory distinction
Development of the I-D model
in the Middle Ages

The Middle Ages saw a number of

modifications to and developments of
Aristotles basic I-D model
 Robert Grosseteste (c. 1168-1253)

Attempted to systematize choice among

competing theories
Use Modus Tollens to eliminate all but one possible
hypothesis
I.e., deduce a consequence, C, from a hypothesis, H,
show that not-C, conclude not-H
Modus Tollens:
If H, then C
not-C
not-H
Problem is, this cannot be done;
it is not possible to eliminate all
but one hypothesis
underdetermination
 Roger Bacon (1214-92)

Three Prerogatives of Experimental Science:

1. From general principles deduce claims about new
phenomena, and put these to experimental test
(Aristotle required only that the original
phenomena be deduced)
2. Actively and systematically experiment in order to
increase data and knowledge of phenomena
3. Use this knowledge to develop new techniques for
gathering data and testing hypotheses, as well as
for developing practical tools and new crafts; look
to old craft traditions as a source of data and
technical knowledge
These constitute advances over Aristotles simple
inductive-deductive method, because it stresses
systematic gathering of data, the extension of implications
and tests to new phenomena, and a bridging of the gap
between intellectual knowledge and craft knowledge
 Scotus and Ockham on Induction

Further forms of induction articulated

Duns Scotus Method of Agreement
(1265-1308) Case Circumstances Effect Conclusion
1 ABCD e
2 ACE e
A can be the cause of e
3 ABEF e
4 ACD e
William of Method of Difference
Ockham Case Circumstances Effect Conclusion
(1280-1349)
1 ABC e
A can be the cause of e
2 BC -

These two methods are often called the (first two) of Mills Methods,
after J.S. Mill (1806-73), whose arguments in favor of inductivism were
widely influential.
 Ockhams Razor
William of OckhamOckhams Razor
A demand for simplicity, stated in various forms:
Assume nature takes the simplest path available
assume the minimum number of (types of) objects necessary to the
theory/explanation
do not unnecessarily complicate theory
eliminate superfluous concepts.
Ockham would not like the first formulation because it makes a
metaphysical claim about nature (that it pursues the simplest
path), and, Ockham would say, we cannot know how God has
designed natureGod could complicate nature unnecessarily if he
so chose
Instead, Ockham cast his injunction so as to apply to our theories
rather than to nature itselfkeep the theories as simple as
possible
while in principle God could complicate things, we should not pretend
to knowledge of Gods design
we should make our theories as simple and tractable as possible,
given the evidence
Necessary Truth of First Principles
Note that Ockhams caution in stating his razor, as well as the cautious form of the
conclusions in the two inductive methods above, point out a growing recognition of
the fallibility of inductive inference, as well as a reconsideration of Aristotles
requirement that the First (General) Principles be self-evident. For some thinkers
self-evidence may still have been a goal, but many began to recognize (in theory if
not in practice) that one had to be more cautious about the strength of ones
inductively generated conclusions.
Duns Scotus believed that sense experience allowed us to recognize necessary
truths, but that such truths were true in virtue of the meanings of the terms, and it
was understanding of these meanings, not sense experience, which justifies our
belief in them (see Herschel); such truths are necessary, as their denials are self-
contradictory
An early version of analyticity

Aristotle, Scotus, and others had assumed that certain first principles of the special
sciences could be known to be necessary, hence what counted as self-contradictory
extended beyond just what could be reduced to a logical contradiction
Nicolas of Autrecourt (c. 1300-1350+) had a much stricter notion of necessary truth,
restricting it to claims whose denials are logical contradictions
Much like David Hume (1711-1776) four centuries later, Nicolas concluded that we can have
no certain knowledge of causal relations (Hume also draws stronger conclusions)
Much unlike Hume, Nicolas used his critique to encourage faith in a Christian God
The issue of laws and causation
Galileo and Francis Bacon
Galileo (1564-1642) stresses the role of
abstraction and idealization in the inductive stage
Takes to heart and implements R. Bacons
injunction to test hypotheses against new
phenomena
Is a master of qualitative observation and
experimental design
And, of course, one of the first the use the
telescope to make astronomical observations
Francis Bacon (1561-1626) tries to develop a more
systematic and more careful version of Aristotles
I-D model
Eliminate all prejudices and assumptions
Gather a huge amount of data and generalize
cautiously
Build up to ultimate generalities through a hierarchy
of intermediate steps (restricted generalities)
Science should be an organized community endeavor
Science should have practical results, eventually man
should regain his dominion over nature
 Ren Descartes (1596-1650)

Rejects Aristotelian I-D Model, and inverts F. Bacons ascent

to generality
Rather than inductively building from
observations to successively more general and
more fundamental truths (as F. Bacon),
Descartes proposed to start with the most
general and most certain truths and derive more
specific knowledge and observations from those
Clarity and distinctness a guide to a priori
knowledge of concepts, their implications, and
their application
E.g., Descartes thought he could derive, a priori,
laws of physical matter from metaphysical truths
about extension and motion of bodies, and general
truths about the relation between mind and body
from truths about the different substances
Observation has a role in determining under
which circumstances regularities occur, but
observation cannot support general laws
 Isaac Newton (1642-1727)

Advances an I-D model and an Axiomatic Method

Method of Analysis and Synthesis; a form
of the I-D model, but also
A three stage, non-inductive procedure
1. Formulate an axiom system of definitions
and relations (e.g., his laws of motion)
2. Specify a procedure for correlating
theorems deduced from the axioms with
observable phenomena (i.e., interpret the
axiom system; e.g., generalizations
concerning planetary motion)
3. Attempt to confirm the observational
implications of the system (e.g., specific
predictions concerning planetary motion)
Newton recognized a form of imaginative,
abstractive, and idealizing induction in
formulating the axiom system in stage 1
 David Hume (1711-76)
Broad empiricist attack on metaphysics and causal
knowledge
Criterion of meaningfulness
A thought is genuinely meaningful only if it can be
traced back its constituent sensory impressions
(similar to 20th c. positivism)
Knowledge divided into
Relations of ideas
Their denials are logically contradictory, so they
are necessary truths, known with certainty
Subject matter restricted to logic, math, geometry
no question of existence or causation is a relation of
ideas (see analyticity)
Matters of Fact
Neither a MOF nor its denial is logically contradictory,
so each is possible
Based in knowledge of cause and effect, which is not
certain In fact,
 David Hume (1711-76)
Induction, Causation, Matters of Fact
Knowledge of MOF based in knowledge of cause and
effect
Knowledge of cause and effect based on experience
All knowledge based on experience presupposes the
fundamental inductive principle that the future will be
like the past
But this principle is not a relation of ideas, so it must be
a MOF
But then the fundamental principle of induction cannot
be justifiedany attempt to do so would require
presupposing that very principle
This is Humes version of the problem of
induction
 David Hume (1711-76)
Hume concludes that our judgments concerning matters of fact
(including causal relations) are not rationally grounded at all
Rather there are (stronger or weaker) habits of expectation which
evolve in us as a result of (i) our natural propensities and
(ii) observation of constant conjunctions of events (fire then heat,
fire then heat)
This skeptical solution is a form of psychological naturalism
description of what we do, how we cannot avoid it
He is rejecting inductive justification (though not inductive
practice), as well as intuition of necessary truth
Indeed, we cannot know the hidden springs and principles
underlying the world we observe, all we ever know of is the
constant conjunctions of kinds of eventslaws and causation
Except, perhaps, for Nicolas of Autrecourt, Hume is the first weve
looked at to completely reject the ideal of somehow arriving at
secure generalizations or certain First Principles of some sort
Hume, unlike Nicolas, used these skeptical results to argue
against metaphysics and religion
 Immanuel Kant (1724-1804)
Kant responds to Hume by distinguishing the form
from the content of knowledge
Form is not given in experience; rather the raw
data of experience is structured in various ways
by the rational human mind
Space and Time are the Forms of Sensory Intuition,
all perception is structured by these forms
Perceptions so structured are further organized and
synthesized according to 12 Categories of
Understanding (e.g., Unity, Substance, Causality, Contingency, etc.)
Judgments are made and organized via the Regulative Principles of
Reason
Since the structures/forms Kant posits are inherent in the human
rational mind, they are open to philosophical investigation, they
can be known via pure reason
Since any knowledge (esp., empirical/scientific) presupposes the
Forms, Categories, and Principles, we can (contra Hume) have
knowledge of the general features of any possible scientific theory,
including fundamental and general truths about causation, matter,
motion, etc.
 Immanuel Kant
Transcendental Idealism
from a point of view which attempts to transcend our forms of
cognition, we recognize how much our mind and its structure
contributes to our knowledge of the world
Empirical Realism
from a point of view which does not attempt the impossible
transcendence, the structures imparted by the Forms, Categories,
and Principles are fully real, and necessary truths regarding such
things as causal relations and matter can be known
Some problems with this ingenious and seductive picture:
What justifies saying this is what any theory or cognition must
presuppose? How can we be sure weve correctly identified the most
basic forms, categories, and principles? Must they be unique? Kant
thought he had identified unique basic forms, but some of what he
took as basic to science has since been changed and rejected by
science
How can one coherently speak of the transcendental unreality of forms
and categories, while maintaining that transcendence is impossible,
and that the transcendental (noumenal) world is unknowable?
 J.S. Mill (1806-73)

Laws, General Principles, and theoretical

claims are justified by inferences from
experience which satisfy inductive schemas
or formsthis is known as Inductivism
Much like the I-D model, but little import
allowed to the D side, and Mill had very
specific inductive methods in mind
Mill a bit unrealistic about how well
inductive schemas can justify theoretical claims
Deduction from hypotheses of successful prediction a
requirement, but not a justifying factor unless all other
possible hypotheses are eliminated
Again, justification of theoretical claims is gained only by
conformity of the data to inductive schemas supporting the
theoretical claims
 Mills Methods
Method of Agreement* Method of Difference*
Method of Concomitant Variation
Case Circumstances Effect Conclusion
1 An BC an b (occurrence of a is varying proportionally
2 A0 BC a0 b to occurrence of A)
So either A causes a, or A and a have a
3 A1/n BC a1/n b common third cause

Method of Residues
Case Circumstances Effect Conclusion
1 ABC abc
2 B b A is the cause of a
3 C c

*See Scotus and Ockham on Induction

 Hypothetico-Deductivism
William Whewell (1794-1866) and
W.S. Jevons (1832-82) rejected Inductivism
Rather than justified on the basis of
inductions, a hypothesis is justified when it
Is consistent with other established
hypotheses, and
The consequences deducible from the
hypothesis agree with observations Whewell

With its strong emphasis on predictive test,

this is in the spirit of Aristotle, R. Bacon,
Galileo, and Newton, but the view gives
more priority to predictive test than these
others (by giving much less importance to
induction)

Jevons
Herschel on Discovery and Justification
John Herschel (1792-1871)
Herschel distinguished the issue of how a
theory was arrived at (context of discovery)
from the issue of its acceptability or
justification (context of justification)
He claimed that context of discovery is
strictly irrelevant to context of justification
Discovery
Use of inductive schemas
Abstraction and imaginative hypothesis
F. Bacon-like hierarchy of generalizations
Justification
Success of deduced predictions (thus a form of hypothetico-
deductivism), especially
Extension of predictions to extreme cases
Deduction of unexpected predictions and their successful testing
Use of crucial experiments to decide between competing
theories/hypotheses
 Problem(s) of Induction
Two Issues:
The Descriptive Issue: we arrive at beliefs regarding unobserved
matters of fact (future particulars, eternal generalizations)How do
we do that?
The Normative Issue: do we arrive at such beliefs the way in which
we ought to arrive at them? I.e., are we justified in our practice? If not
is there any practice we could adopt which is justified?
Problems Regarding the Normative Issue:
Uncertainty/Underdetermination : Inductive inference is
underdetermined, hence not truth-preserving, hence some amount of
uncertainty is involved, even when starting from certain premises
Lack of Rational Ground: The basic principle(s) of induction are not
logical truths, nor self-evident, so what justifies those principles? As
Hume points out, we cannot appeal to experience, because any such
appeal presupposes the very inductive principles in questionthis is
the problem
A useful online discussion: The Problem of Induction
 Underdetermination
The Underdetermination of Theory by Evidence
Given any amount of observational evidence, there will be more
than one (indeed infinitely many) theories compatible with that
evidence
A unique theory is never dictated by the evidence, not even if we
had all possible evidence
This raises the question of how and if we can rationally decide
between theories
WRONG and often WRONGLY ATTRIBUTED TO W.V. Quine version:
a theory can be preserved in the face of any contrary evidence
(what Quine says, in case youre interested, is that a hypothesis
or statement can be preserved as long as others are given up, but
this is a CHANGE IN THEORY, some hypotheses are preserved,
others not)
This correct understanding of Quine implies that
There are no crucial experiments to rule out a hypothesis
There are (near) crucial experiments to rule out whole theories
 Observation/Theory
Distinction
Intuitively, there seems to be a distinction between that which we
can observethe observational; and that which we cannot
observethe theoretical
Observable: people, stars, trees, rocks, grains of sand, a patch of
red
Theoretical: electrons, quarks, viruses, dark matter, the big bang,
trees, people
The theoretical is posited or inferred to help predict and explain
the observable
Problems abound for this distinction:
Does the distinction concern observable vs. theoretical
Objects?
Words and Sentences involved in scientific claims?
Sense Data vs. Things in the World?
How sharp is the distinction?
Does something projected onto the retina through a microscope or telescope
count as observable?
What about artificially colored images produced on a screen by an electron
microscope or an infrared sensitive telescope?
What about observations made by a prosthetically enhanced human?
Revolutions and Rationality
Given the problems of Induction and
Underdetermination, is theory acceptance
(change) at all rational?
If so, there must be some substantial
constraints on theory acceptance beyond
mere induction and deduction of
observable consequences
What are they? How do we know?

next slide
Revolutions and Rationality
Kuhn claims that scientific revolutions (periods of significant
change in theory and practice) are highly non-rational affairs,
highly unconstrained
In part, this is because what makes a revolution revolutionary is
that the normally accepted theories and the methodological
practices they ground are in question and changing, hence, it
seems, they cannot constrain their own change in a rational way
But revolutions are lauded as important advances in our thinking
and knowledge, and we think they are good and justified changes
in theorybut how could they be if they are not rationally guided?
Moreover, historians of science often recognize that many social
forces can play a role in revolutions and the (non)acceptance of
theorywhat gives?
Finally, the apparent lack of rational constraint and presence of
social forces raise the demarcation issuehow, if at all, is science
different from other organized bodies of beliefs (religion,
metaphysics, political structures, cultural tradition)?
 Social Forces
How and to what extent do religious, cultural,
political, gender, racial biases and interests affect
scientific theorizing?
Can they be avoided? Ought they to be avoided?
This has implications for the issue of the
rationality of theory acceptance and change
If theory acceptance and change are not
rationally constrained there seems to be plenty of
room for non-rational social forces to be in play
Moreover, the apparent lack of rational constraint
and the role of social forces raises the
demarcation issue
 Demarcation
How, if at all, is science different from other
organized bodies of beliefs (religion,
metaphysics, political structures, cultural
tradition)?
How can a difference be marked out?
Can good science be distinguished from bad
science? Is pseudoscience a third thing, or just
really bad bad science?
Does/should science have a privileged
epistemological standing in relation to these
others?
This all relevant to revolutions and rationality,
and social forces
 Realism/Anti-realism
Given the various epistemological difficulties
(underdetermination, problem of induction,
rationality, social forces), and the lack of a
consensus on these issues, why should we think
that our theories are actually describing reality?
The apparently large gap between observational
and theoretical knowledge inspires worry about
realism
Metaphysical difficulties come into play here as
wellwe do not have good understandings of the
nature of laws and causation, explanation, so
how can we claim that we are discovering the
nature of the universe?
 Laws and Causation
Laws are often thought of as general statements of causally
necessary connection between events, and the statements
of laws themselves are sometimes thought to be necessary
truths
But given the various epistemological problems, especially
the Humean critique, it is unclear whether or not causes
and laws can be or be known to be as described above
If laws do not state a necessary connection and are not
themselves necessary truths, then what, if anything,
distinguishes them from accidentally true generalizations?
Is there really any such thing as a law of nature?
This all connected to issues of realism and explanation
 Explanation
Science is supposed to explain things to
us
But what does it mean to have a scientific
explanation?
Does mere derivability of a description
from more general truths constitute an
explanation?
What sort of explanations can science
provide?
How can we tell good from bad
explanations?
 Analyticity
Statements which are analytic are supposed to be conceptual truthstrue in virtue of
the meanings or concepts involved
Locke: a part of a complex idea is predicated of the whole
Kant: the predicate concept is contained in the subject concept
Carnap: true in virtue of the meanings of the constituent terms
E.g., all bachelors are unmarried

This is contrasted with synthetic statements whose truth (or falsehood) is a matter of
something beyond the meanings or concepts involved (the world, matters of fact)
E.g., all faculty are bachelors

Locke and Kant were the first to make use of this distinction, it played a prominent
role for the logical positivists (as well soon see), Quine repudiated it
Analyticity provides a way (though not the only way) of explaining how at least some
truths are
Knowable a prioriwithout appeal to experiencevia linguistic analysis or merely
understanding the language
General principles, or frameworks, for theorizing
Necessarily true
A matter of linguistic convention
Accepted or rejected on purely pragmatic considerations and thus lack metaphysical import

it all depends on who is making the distinction and to what use they are putting it