Sunday, 3 August 2014

Does Falsifiability Identify Science?

There have been some interesting discussions of falsifiability and science recently: Sean Carroll's essay Falsifiability is a good starting point:
In complicated situations, fortune-cookie-sized mottos like "theories should be falsifiable" are no substitute for careful thinking about how science works. 
I tend to agree. Alan Sokal wrote a series of articles at Scientia on the definition of science:
The bottom line is that science is not merely a bag of clever tricks that turn out to be useful in investigating some arcane questions about the inanimate and biological worlds. Rather, the natural sciences are nothing more or less than one particular application — albeit an unusually successful one — of a more general rationalist worldview, centered on the modest insistence that empirical claims must be substantiated by empirical evidence.
Professor of Astrophysics Coel Hellier wrote:
It is the model that needs to be falsifiable, not every statement deriving from a model. Thus falsification remains important in science, but it is wrong to reject an idea such as the multiverse owing to an over-simplistic application of falsifiability.
And Massimo Pigliucci wrote in answer to the hypothetical proposition 'There is a specifiable “scientific method” that possesses some definable core or essential steps, used by all genuine sciences':
No. Philosophers of science have looked for just such an algorithm (e.g., the famous “hypothetico-deductive” method [1], or Popper’s falsifiability [2]) and have come up short. The post-Kuhn [3] consensus is that there is no such method, and that science helps itself to a loosely defined toolbox of methods, heuristics and intuitions. 
So how valid is Popper's criterion?

In the mid twentieth century when Karl Popper was writing, science was identified with the empirical method – observations, and inductive inference from them. He noticed how this process could be abused by theories he came to think were not science; he wanted ‘to distinguish between science and pseudo-science’ (while ‘pseudo-science’ is different from ‘non-science’, since Popper employs it I shall use it here synonymously with ‘non-science’). Popper thought that the followers of Marx, Freud and Adler could fit any conceivable evidence into their worldview. Furthermore he agreed with David Hume that inductive inference – that the past is a guide to the future - could not be logically justified, and so induction does not guarantee knowledge.

To differentiate knowledge-delivering science, then, from non-science, Popper suggests the criterion of falsifiability. By positing a principle that might hold universally (a generalisation) we can establish a deductive argument that allows us to refute it, should observation disconfirm it.  The argument is in the form modus tollens, denying the consequent:
If Theory then Prediction
Not P
Therefore, Not T
Predictions met would corroborate the theory, but in deductive terms to draw the conclusion that T is true from repeated Ps would be the fallacy of affirming the consequent. So we can never prove a theory true by this construction; just prove it false. Popper says:
...the criterion of the scientific status of a theory is its falsifiability, or refutability, or testability...
So I think it's fair to say that he thinks falsifiability is both necessary and sufficient for scientific status. It does not matter how conjectures are formed, whether by induction or simple invention; testing them will eliminate the false conjectures. Indeed Popper also thinks that conjectures arise from our expectations prior to observation, a reversal of the traditional view.

I’m tempted to look for a ‘science’ that is not falsifiable to refute Popper, but, although his thoughts were prompted by Einstein overturning Newton’s theories, there is something normative about his proposal; he is saying that this is what science should be, and, if it’s not, it should not be called science. So if I offered a science that was not falsifiable, Popper would say it should not be called 'science'.

Thomas Kuhn. in his analysis in The Structure of Scientific Revolutions, defines ‘normal science’  as ‘research firmly based upon one or more past scientific achievements that some particular scientific community acknowledges for a time as supplying the foundation for its further practice’. Kuhn is describing the mundane day-to-day science he thinks more accurately represents the bulk of scientific work; in his view Popper’s ‘critical attitude’ is restricted to those rare periods in science when revolutionary changes occur which alter the way that the scientific community look at the world; he even thinks this is more philosophy than science. In his response to Kuhn, Popper agrees that ‘normal’ scientists exist, but insists they are ‘badly taught’, lacking the critical attitude. He identifies the critical attitude with the ‘scientific attitude’ and its converse, the dogmatic attitude, with the pseudo-scientific. So he thinks they are not doing science at all during the ‘normal science’ phase, and it should not be called 'science'.

Conversely, though, there are disciplines which we don’t include in the sciences which are falsifiable; astrology is a discipline that both sides have used to illustrate their views. Popper cites it as a pseudo-science built on a ‘mass of empirical evidence’ and Kuhn agrees it is not a science, but notes that it has made many predictions, which have simply proved to be false. Popper should say that astrology is a science, because it has been refuted, proving its falsifiability; astrology’s (generally agreed) non-science status suggests there is something more to being a pseudo-science than mere non-falsifiability. However, astrologers offer reasons for predictive failure; see Astrology on the Attack in this article, which shows that astrologers can also use the same language to explain anomalies as scientists, but which echoes the language Popper found so objectionable in the followers of Marx, Freud and Adler, so perhaps Popper would continue to deny astrology's falsifiability.

To consider what this ‘something more’ could be, I will discuss a number of problems I see with Popper’s criterion: a) it does not escape the assumption of uniformity; b) it does not recognise different levels of predictive power; and, c) it undervalues supporting evidence.

a) Assumption of uniformity

Hume notes that we have a habit or custom to assume that things will tend to stay the same. Peter Lipton draws a distinction between this sort of inductivist and the opposite sort, to illustrate Hume's issue with induction:
To illustrate the problem, suppose our fundamental principle of inductive inference is ‘More of the Same’. We believe that strong inductive arguments are those whose conclusions predict the continuation of a pattern described in the premises. Applying this principle of conservative induction, we would infer that the sun will rise tomorrow, since it has always risen in the past; and we would judge worthless the argument that the sun will not rise tomorrow since it has always risen in the past. One can, however, come up with a factitious principle to underwrite the latter argument. According to the principle of revolutionary induction, ‘It’s Time for a Change’, and this sanctions the dark inference. Hume’s argument is that we have no way to show that conservative induction, the principle he claims we actually use for our inferences, will do any better than intuitively wild principles like the principles of revolutionary induction. Of course conservative induction has had the more impressive track record. Most of the inferences from true premises that it has sanctioned have also had true conclusions. Revolutionary induction, by contrast, has been conspicuous in failure, or would have been, had anyone relied on it. The question of justification, however, does not ask which method of inference has been successful; it asks which one will be successful. 
The point is, we have no more justification for what will be successful, per Hume, if we are conservative or so called revolutionary inductivists.

But imagine that the world actually has conformed (and will conform) to the revolutionary inductivist principle ('It's Time for a Change'). For the deductive argument we must first propose a generalisation. What would a generalisation look like if we lived in Lipton’s ‘revolutionary inductivist’ world, in which nature is not uniform? All blackbirds are any colour? What goes up will sometimes come down and sometimes not? A non-conservative-inductivist world renders the conjectures we make all-inclusive, so it would be impossible to eliminate potential outcomes by falsification – there would be no events that fall outside the conjectures. Maybe Popper would say the conjectures should exclude the uniform – All blackbirds are any colour but their current one? What goes up will do whatever it did not do last time? I’m not sure why we should exclude the possibility of things staying the same in a revolutionary inductivist world so this counter doesn’t convince. Falsification in such a world would mean uncovering an instance where the status quo is maintained. Positing a generalisation describing constant change is paradoxical, and casts doubt on just how far Popper has escaped the problem of induction with his deductive argument.

b) Variations in predictive power

Candidate theories could be divided into the following categories:

1) The theory cannot make predictions.
2) The theory can make predictions, but they are not met.
3) The theory can make predictions, but some are met and some aren’t.
4) The theory can make predictions and all are met.

As mentioned previously, Popper’s criterion is a yes or no demarcation, so for him any candidate in (1) is pseudo-science and any in (2), (3) and (4) is science. This means that a theory with no corroborating evidence but with predictive power is as scientific as one whose every prediction (so far) provides corroborating evidence. To be fair, Popper does accept there is a difference; he says ‘some theories are more testable...than others; they take, as it were, greater risks’, but his criterion does not make any allowance for this granularity within theories, and indeed, disciplines.

Candidates for scientific status range from the so-called ‘hard’, or established sciences, like physics, chemistry and biology, to ‘soft’ sciences like psychology and sociology, and then to those outside the scientific fold currently: astrology, homeopathy and creationism, for example. The hard sciences are accepted as scientific, but even within these disciplines theories can arise that cannot make predictions. Famously Popper said that ‘Darwinism is not a testable scientific theory’. He later retracted this, but it highlights the difficulty in drawing a sharp distinction; it was not at all clear what predictions Natural Selection could make. The immense number of variables which come to bear in the real world makes prediction difficult. To predict that speciation will occur is plainly not specific enough, but to predict particular speciations in mammals, for example, would require an almost omniscient knowledge of environmental changes and thousands of years within which to experiment.

As a further example, climate science might predict stormier summers, all other things being equal, but that prediction could be disrupted by a freak volcanic eruption. Avoiding the consequences of a failed prediction by appealing to such variables has much the same appearance as the behaviour that Popper found so dissatisfactory amongst the followers of Marx, Adler and Freud. But with disciplines that operate at a higher level of complexity than the ‘harder’ sciences, where it is easier to isolate variables, this is a reasonable explanation for failed predictions, and may not point to unscientific behaviour. Carroll says this about modern cosmology:
We can't (as far as we know) observe other parts of the multiverse directly. But their existence has a dramatic effect on how we account for the data in the part of the multiverse we do observe. It's in that sense that the success or failure of the idea is ultimately empirical: its virtue is not that it's a neat idea or fulfills some nebulous principle of reasoning, it's that it helps us account for the data. Even if we will never visit those other universes.
Cutting edge science often flirts with untestability. In short, sometimes it’s not clear if a theory is unfalsifiable in principle or just in practice, while Popper’s discussion of Freud and Adler indicates he is targeting those theories that he thinks are not falsifiable in principle. Perhaps establishing a theory as unfalsifiable in principle cannot even be established.

c) Supporting evidence

By solving the induction problem by removing it, Popper loses a valuable way of distinguishing between theories. Consider this thought experiment: you are going on holiday and when you arrive at the airport you have a choice of two planes to get to your destination: Plane 1 is supplied by Hume Airlines, and Plane 2 by Popper Airways. Planes of Plane 1’s design have successfully completed 1000 flights while none of Plane 2’s design has yet flown. Don’t worry though, says the chief executive of Popper Airways; the theory behind their plane’s design is as falsifiable as Plane 1’s, and for all anyone knows Plane 1 is going to crash on the next flight anyway! Which would we prefer?

For Popper, both planes’ design theories are equally falsifiable, and equally scientific, but I think it’s obvious which plane we would prefer; that we can see a difference between these scenarios suggests there is some value to the accumulation of inductive evidence, and to call this recourse to inductive inference non-scientific is hard to defend. Granted Popper does discuss corroborating evidence, but his bald formulation does not recognise it as scientific.

Darwin set out to show God’s hand in nature but accumulated anomalies eventually resulted in a ‘paradigm’ change for him, revealing unguided natural processes at work rather than God’s hand (Kuhn’s ‘paradigm’ is ‘an example of scientific practice that scientists in a certain tradition tacitly accept and follow’). Kuhn’s more inductive, perhaps less rational and occasionally unscientific, reading of how science operates is perhaps a more accurate description than Popper’s. Darwin’s careful collection and documentation of data over decades looks like Kuhn’s ‘normal science’, and, again, to call this non-scientific (as Popper calls normal science) is too restrictive.


In speciation there is no hard line between ducks and their non-ducky ancestors, but when something walks like a duck, swims like a duck and quacks like a duck, we call it a duck. Ironically this inductive inference we also use when we draw a familial resemblance between theories we call scientific, drawing on many factors, such as predictive power, parsimony, consilience with other disciplines, internal consistency and, yes, supporting data. That there is no hard line between science and non-science reflects how science changes over time and how varied it is. Paul Feyerabend calls science a ‘narrow-minded institution’ (pp. 174–5), but, contra that characterisation, it is instead a wide-ranging enterprise encompassing many different methods and techniques; Pigliucci's 'loosely defined toolbox'.

If knowing when a theory is wrong increases our knowledge, then Popper’s falsifiability is an important element in the progress of science. But science has elements and episodes that could be called pseudo-scientific, and pseudo-sciences have elements and episodes that could be called scientific. The assumption of uniformity, variations in predictive power and different levels of supporting data combine to show that the line between science and non-science is fuzzy and cannot be drawn by falsifiability.


Chimisso, C. (2011) Knowledge, Milton Keynes, The Open University.

Cottingham, J. (ed.) (2008) Western Philosophy: An Anthology, Oxford, Blackwell Pub..

Okasha, S. (2002) Philosophy of Science: A Very Short Introduction, Oxford, Oxford University Press.

Popper, K. (1976) Unended Quest. An Intellectual Autobiography. LaSalle, IL: Open Court.

Feyerabend, P. (1978 [1975]) Against Method, London, Verso


Post a Comment