Towards an Objective Theory of Rationality

7. Objective Criteria and MSRP

This completes my development from the general demands of an objectivist epistemology rational criteria for explanatory theory and sentence appraisal. These criteria may now be summarily presented.

The conditions for a statement to be evidence for a theory are:

Criterion 1:
the statement is true;
(requirement of dependence on epistemic value)
Criterion 2:
the statement is logically entailed by the theory in conjunction with accepted auxiliary hypotheses, but not by the auxiliary hypotheses alone;
(requirement of dependence on logical deducibility)
Criterion 3:
the statement is not a statement of the relationship between the theory and a subjective state or states;
(requirement of independence from psychological states)
Criterion 4:
the epistemic status of the evidence-statement is determinable independently of the theory;
(requirement of epistemic independence from theory)
Criterion 5:
the evidence-statement is independent of the data used to construct and modify the theory;
(requirement of independence from theory development)
Criterion 6:
the statement is interpreted in terms of accepted physical and observational theories;
(dependence on accepted presuppositions)
Criterion 7:
the statement is not the result of distorting subjective influences, such as prior beliefs and experiences, expectations, needs, values and personality.
(requirement of independence from bias)

Criterion 1, 2 and 3 state the minimal conditions for a statement to be regarded as evidence for a theory. Statements satisfying Criterion 4 and 5 are regarded as strong evidence for a theory. Criterion 6 and 7 apply specifically to observation statements about the external world, with Criterion 7 extending the scope to include introspective psychological reports.

Book cover: For and Against Method by Imre Lakatos and Paul Feyerabend

As can be seen, I have tried to isolate and clarify both the various dependence and independence requirements necessary for an adequate objectivist epistemology. Some of these are common to other epistemologies. The dependence requirements specified in Criterion 1 and 2 are seen in all sophisticated hypothetico-deductive systems. Less sophisticated systems assume, wrongly, that scientific theories entail predictions in isolation. The independence requirements stipulated in Criterion 3 and 4 are usually taken for granted, while that in Criterion 7 is only recognized by epistemologies that incorporate a more sophisticated theory of observation than the camera lens analogy.

Lakatos had repeatedly emphasised that there is no theory-neutral observation base and that observational theories are revisable.[27] This is a fundamental tenet of his MSRP and is reflected in my Criterion 7. However, Lakatos had little cause to specify this condition separately and in the form that I have stated it here, for he had focused his attention almost entirely on the history of physics. The historical record shows physicists deciding the worth of rival theories, in large part, by their ability to predict novel phenomena within physical systems. Such phenomena were characterised by being open to scrutiny by the scientific community and interested public. Any theory that revealed phenomena that were only 'observable' to the protagonists of a particular theory, and not to the rivals and disinterested parties, was not taken seriously and was quickly forgotten.[28] The situation is more ambiguous in the case of the social sciences, such as psychology and sociology. In these disciplines, the primary evidence for rival theories is the psychological responses of human beings and these may be easily distorted by inadequate testing procedures, as I had explained in §6.3 above.

The 'requirement of independence from theory development', Criterion 5, is reflected in Lakatos' and Zahar's stipulation that for a series of theories (or, in my terminology, successive versions of the same theory) to be rationally justifiable, it must successfully predict novel facts or novelly derive known facts. This historical view of confirmation forming the basis of Lakatos' MSRP specifies that we cannot judge the epistemic relationship between a theory and evidence-statements in isolation, but must know the prior history of the theory. In other words, for an evidence-statement to support a theory, it must not have been used by its proponents in the construction and revision of the theory. The above considerations of the general demands of an objectivist theory of knowledge appear to vindicate Lakatos and Zahar's historical approach to theories of rationality.

The epistemology formulated here, as with Lakatos' parallel, is universally fallibilist about empirical knowledge. There is no immediately known observational knowledge and no indubitable synthetic a priori principles. Even realism is not on absolutely solid foundations, for it may turn out that those phenomenological experiences that we label 'sense-experiences' have been misdescribed, or even that they never existed. It may seem that this fallibilist epistemology can never be applied because we can never get to use it. Observation statements, to be acceptable in testing a theory, must be interpreted in terms of accepted physical and observational theories. However, such physical and observational theories require evidence in the form of observation statements to be deemed adequate. So, it may seem that the evidence-theory-evidence circle can never be broken long enough for us to enter into. I think there are two ways to answer this quandary; the philosophical and the psychobiological.

Dealing with the philosophical first, it seems epistemically prudent simply to jump into the circle at any point. Just as we tentatively accept the existence of those experiences that we label 'sense-experiences' in order to enter the debate between idealists and realists, we also tentatively accept some relatively inarticulate and undeveloped observational, psychological and physical theories to start the enterprise of theory improvement. We may then improve a theory by using our tentative theories as a support structure in order to make and test novel predictions. We only gain knowledge by initially pitching our tent somewhere and then investigating where we can cut a trail into the unknown using our epistemological exploration tools. In this way, we systematically improve our knowledge of the world, testing one domain at a time.

With this philosophical response to our predicament, we can either accept the invitation to set up a base camp or not. We come to the second approach to this puzzle by realising that we have no choice. Our biological evolution ensures that we do enter the evidence-theory-evidence circle at a predetermined point. It appears that we do not act, according to this evolutionary view of theory improvement, by a conscious volition. In §5 above, I outlined how research in neurophysiology and psychology show that our brains are hard-wired to recognise a world of external physical objects from a very early age. Our perceptual experiences and beliefs are shaped to see and accept mind-independent objects behaving in law-like ways. Realism most likely evolved as a privileged hypothesis during our evolutionary development because it offered survival advantage.[29] In a predatory environment, our ancestors were more likely to pass on their genes if they had the cognitive ability to quickly recognise a running tiger.

With the expectation of how objects move and behave hard-wired in our brains, current research also shows that we do not give up these expectations lightly in the face of seeming counterexamples. Studies in neurophysiology and psychology suggest that our propensity to revise our beliefs is tempered by a measure of epistemic conservatism. Situations where we are met with perceptual ambiguity are excellent illustrations of this kind of conservation of belief in the face of apparent refutations. For example, consider the illusions I raised in §5 above. When we encounter the trapezoidal room in the Ames illusion, we don't give up our belief that rooms are symmetrical. Similarly, when we see the inside of the convex face mask in the rotating mask illusion, we don't abandon our belief that noses point outwards on human faces.

Book cover: Worldviews: An Introduction to the History and Philosophy of Science by Richard DeWitt

Another example of perceptual incongruence is the McGurk effect,[30] where we hear the sound 'Fah' in synchronism with a person's lips appearing to make that sound, even though the actual sound emanating from the person is 'Bah'. One last example I'll draw your attention to is the vanishing ball illusion[31] in which a ball appears to leave a person's hand even though it was not actually thrown into the air. In these last two examples, we do not give up on the belief that sounds invariably match the speaker's lip movements and that balls continue to persist in space and time. What is instructive about these illusions is not that we consciously refuse to update our beliefs when faced with these types of counterexamples. It's that our brain conspires to hide the counterinstances from our conscious awareness.

No doubt, this epistemic conservatism also proved evolutionarily advantageous. Abandoning a well-established construct of the external world after a seeming counterexample—to be left with no model of reality at all—would have cursed our ancestors to the next predator that happened to cross their path. There is a corollary here to naïve versions of Popper's falsificationism. Abandoning an established scientific theory at the first sign of an anomalous observation would similarly leave us without a model with which to view the world.

Of course, we do change our beliefs about the trapezoidal room, the hollow mask, the imaginary ball thrown in the air, and so on, when we are shown how the illusion works. That our illusory perceptions persist even when we know how the illusion works is testament to the conservative nature of our perceptual and cognitive apparatus. To bring this discussion back to Lakatos' theory of rationality, the philosophical and evolutionary psychological considerations discussed here demonstrate how the epistemically conservative approach elucidated by his methodology of scientific research programmes  is well-grounded not only in theory, but also in practice.

The theory of rationality I have advocated in this essay also partakes in the dialectic between observation and theory. Our knowledge of how we know improves with the improvement in our observational, psychological and physical theories. And improvements in our theory of rationality will, likewise, require modifications in our appraisal of observational, psychological and physical theories. The general acceptance of double-blind procedures in recent years was motivated by the new developments in the psychology of perception. And this improvement in our methodology of theory appraisal has led to further improvements in our theory of perception. A universal fallibilism should cause us no concern, for it cannot undermine an objective search for truth.

Nonetheless, the theory of rationality that I have developed here is far from complete. The definitions set out in §4.1 were necessarily brief and require much more precision and expansion. Secondly, I had discussed briefly how to apply the criteria for strong evidence to the appraisal of competing theories in §4.3, but much more work needs to be done in developing a method of application in real historical situations. For example, I have not mentioned the problem of anomalies for theory appraisal, and Lakatos had deliberately ignored the problem completely. However, it does seem that the severity and number of anomalies for a theory is important for theory appraisal.

Furthermore, the situation is complicated by the fact that many of our theories are not strict deductive systems, as Lakatos had already pointed out. A complete theory of theory appraisal must incorporate some measure to gauge the strength of corroborated novel evidence, and this measure will have to be an adaption of Lakatos' measure of the severity of a test [Lakatos 1978b: 175]. So, on this account, the measure of the relative strength of corroborated novel evidence e for a theory T relative to a rival theory T', and sharing auxiliary theories A, will be the probability of the evidence e given T and A minus the probability of the evidence e given T' and A, or p(e,T&A) − p(e,T'&A).


  1. [27] See, for example, Lakatos [1978a: 14–16, 23, 45f].
  2. [28] For some amusing examples, see Martin Gardner [1957: ch. 10].
  3. [29] See, for example, Kveraga et al [2007: 147].
  4. [30] For a visual demonstration, see BBC [2010].
  5. [31] For a visual demonstration, see Sunderland Echo [2008].

Copyright © 2016

You will be interested in
Book cover: Free: Why Science Hasn't Disproved Free Will by Alfred R. Mele
The God Argument: The Case against Religion and for Humanism by A. C. Grayling
Book cover: Theories of Ethics by P. Foot
Book cover: Enlightenment Now: The Case for Reason, Science, Humanism, and Progress by Steven Pinker
Book cover: Freethinkers: A History of American Secularism by Susan Jacoby
Book cover: A Historical Introduction to the Philosophy of Science by John Losee

Share This

  • twitter
  • facebook
  • linkedin
  • googleplus
  • gmail
  • delicious
  • reddit
  • digg
  • newsvine
  • posterous
  • friendfeed
  • googlebookmarks
  • yahoobookmarks
  • yahoobuzz
  • orkut
  • stumbleupon
  • diigo
  • mixx
  • technorati
  • netvibes
  • myspace
  • slashdot
  • blogger
  • tumblr
  • email
Short URL: