Updated: 21 January 2000


 Science - observation and principles


OK - but cooking too involves observation and principles


Bacon Novum Organum


“Idols and false notions … are now in possession of the human understanding, and have taken deep root therein, … truth can hardly find entrance …”




“The conclusions of human reason as ordinarily applied in matter of nature, I call for the sake of distinctions Anticipations of Nature (as a thing rash or premature). That reason which is elicited from facts by a just and methodical process I call Interpretation of Nature.



“One method of delivery alone remains to us; which is simply this: we must lead men to the particulars themselves, and their series and order; while men on their side must force themselves for awhile to lay their notions by and begin to familiarise themselves with facts”




“… the winning of assent, indeed, anticipations are far more powerful than interpretations; because being collected from a few instances, and those for the most part of familiar occurrence, they straightway touch the understanding and fill the imagination; whereas interpretations on the other hand, being gathered here and there from various and widely dispersed facts, cannot suddenly strike the understanding; and therefore they must needs, in respect of the opinions of the time, seem harsh and out of tune;…”



Here a white swan, there a white swan …


Ergo : All swans are white




      “a scientist should make a large number of careful observations, and then obtain predictions and generalizations by the process of inductive inference from these observations”


Gillies Philosophy of Science in the Twentieth Century (Blackwell 1993)


both –

a method of discovery

a method of justification




      further observation – testing




Not the entire story


Brownian motion


Inference to the best explanation - posit hypothesis or theory  on the grounds that it provides best explanation


Typical of modern science


      we are puzzled about what we observe, scientist tells a story written in a new vocabulary


      argues for the story on the basis of its explanatory power


      realist vs instrumentalist construal of the story


Inductivism  -  a role to play ? a starting point ?



HUMEAN EMPIRICISM -      experience as source of all knowledge and of all meaning





      “All ideas are derived from impressions”


expose nonsense

            rule out verbiage


some words have meaning by standing for a kind of experience - “red”


all other words have meaning in virtue of being definable in terms of such words


end of metaphysics

trouble for physics  ? “absolute space”


but see Galen Strawson for contrary




      we know the contents of our own mind


what else?


      will today’s bread nourish?


      uniformity of nature ?




induction vs deduction

      induction by simple enumeration

      inference to the best explanation


Hume - not even sure its bread


logical thesis : bemusing condition

psychological thesis : can’t help it


contrast with Bacon







Vienna 1920 - 1936


Inspired by Hume and Einstein and the “scientific revolutions”


Neurath Schlick Carnap Feigl Reichenbach (Godel, Quine, Wittgenstein, Popper, Ayer, Nagel)



Principle of signifigance

      Hume - problem of words like “and”


      Verifiability Principle - a sentence a   has meaning if and only if it is verifiable


meaning given by verifiability conditions


      Analytic - OK


      Mathematics - definitions & consequences




the new physics

Einstein - STR 1905


1.     “A simultaneous with B”

      not verifiable

      inspired by Hume ???


“Einstein told the physicists (and philosophers): you must first say what you mean by simultaneity, and this you  can only do by showing how the statement ‘two events are simultaneous’ is verified. But in so doing you have then also established the meaning fully and without remainder. What is true of the simultaneity concept holds good of every other; every statement has a meaning only insofar as it can be verified; it only signifies what is verified and absolutely nothing beyond this. Were someone to maintain that it contains more, he would have to be able to say what this more is, and for this he must again say what in the world would be different if he was wrong; but he can say nothing of the kind, for by previous assumption all observable differences have already been utilized in the verification.





2.     Kant - a priori arguments to determine structure of space and time


philosophy to emulate science


Reichenbach The Rise of Scientific Philosophy


      Newton - a bad guy



Contrast with Bacon


      positivists were not “inductivists”



context of discovery

context of justification


2. justification of induction


Positivists worried a lot


“vindication” of induction

“blind faith”

“problem to be solved”



      Construction of Science


“If we suppose that I at once take note of every observation – and now start out from thence to construct science, I would have before me genuine ‘protocol propositions’, standing temporally at the outset of knowledge. From them the remaining propositions of science would gradually be evolved through the process which we call ‘induction’, and which consists simply in the fact that, stimulated or incited by the protocol propositions, I tentatively set up general propositions (‘hypotheses’), from which these first propositions, along with innumerable others, logically follow. Now if these others say the same as later observation-statements, obtained under quite specific circumstances that have to be exactly  stated in advance, then the hypotheses continue to rank as confirmed so long as observation-statements do not crop up which are in contradiction to the propositions derived from the hypotheses, and hence to the hypotheses themselves. So long as this does not happen, we believe ourselves to have guessed correctly at a law of nature. Induction is therefore nothing else but a methodically guided guessing, a psychological, biological process whose execution has certainly nothing to do with ‘logic’.”



algorithmic conception of scientific method


mathematical logic - algorithm checking proofs


inductive logic - algorithm for checking support provided to a given hypothesis by a given body of evidence


slow progress


Nagel in 1953


      “… we do not possess at present a generally accepted schema for weighing the evidence for any arbitrarily given hypothesis so that the logical worth of alternative conclusions relative to the evidence available for each can be compared.”





1.     Observational vs theoretical terms

            O - meaning unproblematic

            T - meaning via O


2.         axiomatic system (Newton)


      i.     pure T- postulates (eg  F=ma)


      ii.     mixed  - correspondence rules

            x has mass t if and only x placed          in     S gives reading of t


3.     discovery vs justification


4.     confirmation - successful derived predictions support


5.      explanation by deduction


6.     conventionalism, coherence,   instrumentalism


7.     philosophy - analysis of sci. language









      realism vs instrumentalism





            what about the principle itself ?


            Circle self-destructed







      accepts Hume’s logical argument


      rejects Hume’s pyschological thesis


induction by simple enumeration

inference to the best explanation


1     Sa & Wa, Sb & Wb, ...

2     \    All Ss are W.


      Invalid argument

      Premises provide no support for conclusions


3     Sc & not Wc

4     \    It is false that all Ss are W.


      Deductively valid

Method : bold conjectures and refutation.     




1.     no justification for induction

2.     part of our nature to use induction



1.     no justification for induction

2.     not used (at least by “great” scientists such as Newton)



Deductive Arguments

      Premise(s) entail the conclusion.


Good Inductive Arguments

      Premise(s) support but do not entail the conclusion.


For Popper only deduction is legitimate:

“only the falsity of a theory can be inferred from empirical evidence and this inference is purely a deductive one”  (C&R)




“without waiting for premises we jump to conclusions.  These may have to be discarded later, should observation show that they are wrong” (C&R)



Premises do not even raise the probability of the conclusion.  Probability of all empirical universal laws is zero!




      Lots of content

      The more the risk, the greater the content


      All swans are white or pink or black or green.


      All swans are white.





      No induction from the data.


      Inspiration, hunch, creativity.

      See Medawar’s The Limits of Science




      scientist wants to refute theory


      derives prediction, conducts test


If refuted, start again. If not refuted, it is corroborated.


Corroboration is not inductive support


Passing a test provides no positive evidence.



Popper’s aim -


Criterion of demarcation

      A theory is scientific if and only if it is falsifiable.


Einstein vs Adler, Freud, Marx


1919 Eddington’s test of General Relativity


Contrast with Vienna Circle


      - nothing to do with meaning


      - to explicate cognitive meaning via verification is to use induction.



Realism - aim of science is truth


            truth - correspondence theory




Rationalist - critical approach


            search for refutations

            learn from mistakes

            all we learn is that something is false


Really rational ?


Finite number of theories ?


Verisimilitude -


      T2 more ver. than T1


contains more truth but not more falsity


contains same truth but less falsity


Theory is a set of sentences






Thesis of Verisimilitude TV

      pessimistic induction

      Animal Farm Move - progress



What does verisimilitude really mean?


Miller: all false theories have same degree of verisimilitude.


Good project for graduate school.




How could we show that our theories are increasing in verisimilitude?



1. Induction by simple enumeration ?


      only with a little help from an omniscient being


2. Relative success rates ?


3. Inference to the best explanation



      Galileo - mountains on the moon

      Newton-Smith’s moles

      Einstein - atoms produce Brownian motion


Best explanation of the fact that theories improve is predictive power?


Theories have greater verisimilitude.


“...it would be a highly improbable coincidence if a theory like Einstein’s could correctly predict very precise measurements not predicted by its predecessors unless there is ‘some truth’ in it.  ... it is probable that the theory has a higher degree of verismilitude than those of its competitors which led to predictions that were less successful,...”(Schlipp Vol.)


 ... “there may be a ‘whiff’ of inductivism here.”



If induction allowed here, why not elsewhere?


If not here Popper fails to fit rational model


And if elsewhere what is unique in Popper?


Role of experiments in philosophy.


Quine and meaning

Nietsche and truth



Other Problems:


1.     Why assume a falsified theory remains falsified?


2.     Why rely on a theory if you have no positive evidence for it?


3.     Probability judgments are not scientific.


4.     “All theories are born falsified.”

      Never abandon except for a better theory.


      some dogmatism allowed if scientist declares in advance conditions for rejection.


5.     Status of observation reports.

      Popper: matter of convention not evidence.


6.     No tests in isolation.