THE NO CRITERIA OBJECTION TO THE ARGUMENT FROM RELIGIOUS EXPERIENCE

Kai-man Kwan

        Critics of RE or TE have always emphasized the point that we have no (clear or precise) criteria to find out which TEs, if they are delusive, are really so.  I call these criteria of veridicality (sometimes abbreviated as CV).  Firstly, we note that the allegation of 'no criteria' can mean several things:

1) no criteria at all!

2) no criteria which are non-circular!

3) no criteria which are like those of SE!

        I'll only deal with the second and third claims because as it stands the first claim is false.  It is just not true that the God-experient will accept every experience of God as veridical.  Even the mystics, e.g. St Teresa, would doubt whether their mystical experiences come from the devil.

        The reason why lack of criteria is regarded as damaging is again not uniform.  In relation to my argument, the no criteria objection can conceivably cut in several ways:

1) It can be meant to show that the Type PCT or the like should not be applied to TE. 

2) It can be meant to show that either TE is unreliable or unjustifiable.  This can be due to different reasons:

a) Criteria of veridicality are intrinsically related to the reliability or justifiability of a type of experience.

b) Lack of criteria of veridicality shows an invidious disanalogy with SE.  (This would amount to a variant of the Disanalogy Objection.  I'll discuss this in the next chapter.)

       

Criteria of Veridicality and Applicability of Type PCT

        The most common criticism of Swinburne is the claim that the PC is only applicable when we have rational criteria to distinguish the veridical experiences from the unveridical ones.  For example, Michael Martin alleges that the PC should not be applied

"unless one has a right to assume that perceptual conditions hold under which the entity at issue is likely to appear to an observer if the entity is present.  This right may be justified on inductive grounds, by one's background theory or in other ways" (1986, pp.85-6). 

Consider ordinary perception first.  Let T be "there is a table in front of me" and C1 be "a certain perceptual condition" and A1 be "a table appears to S".  That the PC is applicable in such a case is because we know:

"(1)            If T and C1, then A1.

(5)              If -T and C1, then probably -A1."

So given -A1 and C1, then -T.  Also given A1 and C1, then probably T.  That is why "a table appears to S" is good evidence for "there is a table" and why "there is no table appearing to S" is good evidence for "there is no table before S".  However in the case of alleged perception of God, the conditions under which God will appear to S, if he exists, are unknown.  In this case "it surely seems illegitimate to suppose that an appearance of God is grounds for supposing God exists" (1986, p.85). 

        So though Martin doesn't "wish to deny that (PC) operates in ordinary life and science", he thinks "there are more limitations on its use than Swinburne imagines, and they need to be more tightly drawn" (1990, p.186).  His aim is to restrict the PC so as to reject REs while avoiding scepticism.  The basic restrictions he proposes seem to be twofold.  Firstly a necessary condition for the PC to be applicable is that we have knowledge of the perceptual conditions of veridicality of a kind of perception:

"For example, in order to be able to justify my spontaneous perceptual belief that there is a brown table in front of me, it would seem to be necessary in principle to be able to argue thus: Spontaneous beliefs of a certain sort occurring under certain conditions are usually true, and my belief that there is a brown table in front of me is of this sort and occurs under these conditions.  Consequently, my belief is probably true" (1990, p.157; italics mine). 

Secondly, probably another necessary condition is that there are tests for veridical or unveridical perceptions which are external and non-circular.  This requirement is not explicitly stated but it can be inferred from Martin's comment on 'the test of scriptural compatibility' which some suggest can be used as a criterion of veridical TE.  Martin forthrightly rejects this as a criterion.  It is because, he believes, it "already presumes that the Bible is the revealed word of God and therefore that the Christian God exists, it cannot be used to support an argument from religious experience for the existence of God" (1990, p.160).  If he rejects the test of scriptural compatibility for the above reason, then he has no reason to accept other criteria which are similarly circular.

        Now Martin is right that were we to subscribe to his conditions, the independent evidential force of REs would be greatly endangered.  Yet it is not clear that Martin's sanguine claim that "this would not result in skepticism about ordinary objects" (1990, p.185) is then justifiable.  The reason is that all criteria used for sifting away the unveridical perceptions depend upon sensory experiences and assume anyway the existence of the external world.  Take again Martin's example of seeing a table.  Consider premise (1): If T and C1, then A1.  Presumably C1 will include the following conditions:

a) the environment is not prone to generate erroneous perceptions, e.g. the lighting is normal, etc. and

b) the subject S is attentive and his whole perceptual apparatus (including eyes, optic nerves, brain, etc.) is functioning normally.

        But (a) and (b) alone do not seem to be adequate.  Even if they obtain, it is still possible that the demon or evil superscientist can tamper with his perceptual apparatus in such a way that S does not have the relevant experience.  So a condition (c) has to be added for (1) to be correct:

c) no evil demon or superscientist is tampering with his perceptual apparatus or manipulating his environment and the like (so as to block relevant experiences).

        But suppose we have A1 and we know C1 obtains.  We can't infer from (1) that T.  Hence the argument Martin requires hinges on (5) instead: If -T and C1, then probably -A1.     Given (5), we can argue from the fact that A1 and C1 to the conclusion that T, i.e. the existence of the table.  But note here that while (a) and (b) above remain unchanged, (c) above has to be interpreted slightly differently: it has to assert that no demon is tampering with S's perceptual apparatus in order to generate unveridical experiences.  Now we must ask how do we come to know (5)?  Martin mentions that it can be known by induction.  Perhaps it is something like this: "We observe that in many cases when there is a table and the lighting is normal and S is attentive and his whole perceptual apparatus is functioning normally, if he is asked, we hear S telling us that he appears to see a table."  The problems become crystal clear as soon as this answer is spelt out.  How are we supposed to know any of these things without assuming some of our perceptual claims are at least "innocent until proven guilty", not to mention the problem of justifying our interpreting some sounds emitted from a body as a testimony?  Perhaps we should replace "there is a table, etc." by "there seems to be a table, etc."?  This would be more secure but it is only a correlation of our experiences and it does not justify (5).  Of course I have also ignored the requirement of clause (c) to make Martin's case more plausible.  If we take it into consideration, then it is even harder to see how we can establish (5).  Furthermore, Martin's argument for veridicality also needs to assert that C1 actually obtains in a particular case.  Again how on earth are we going to know it?  It seems that we can only check the lighting by our eyes.  We can check our attentiveness by introspection.  We could check our sensory equipment by doing some surgery but it is rarely done in practice.  The problem of vicious circularity still remains: if I need to justify my application of PC to my 'seeing the table' by further premises and inference, by parity don't we also need to justify the premises in the inference?  They seem to belong to the same type of perception which is in need of justification.  This problem is so obvious and familiar that I find it very surprising Martin never tries to bring it out.  When further arguments and illumination are required, he simply asserts that "knowledge is available about when chairs will appear if they exist" (1990, p,185; italics mine).  He also assures us that there are "various ways" in which "the skeptical questions" can be "silenced" (1990, p.186).  These claims are totally mystifying to me and I wish Martin had spelt them out.  Certainly a theory 'about when chairs will appear if they exist' is available.  It is just not clear how this theory can be transformed into knowledge.  Perhaps there are ways that skepticism can be defeated but it does not seem likely given the above necessary conditions of application of the PC.  It looks as if Martin is so eager to defeat REs that he unhesitatingly lays down the above stringent conditions.  But when the skeptical consequences begin to loom large, he simply asserts or hedges.  Anyway Martin owes us an account how he can hold to both the application of PC to ordinary life and the above conditions of applicability.  In the meantime, we are justified in asking the question: "how are (1) or (5) known?" to which Martin does not give any answer. 

        Actually criteria of veridicality do not only include ascertaining of perceptual conditions.  In SE we also have checks like consensus, employment of scientific equipment, etc.  So perhaps the objection can be put in a broader way: the PC or PCT is applicable only if we have criteria of veridicality.  For example Rowe repudiates Swinburne's PC and formulates this one instead:

"When subjects have an experience which they take to be of x, and we know how to discover positive reasons for thinking their experiences delusive, if such reasons do exist, then it is rational to conclude that they really do experience x unless we have some positive reasons to think their experiences are delusive" (p.91). 

Now we have to distinguish two requirements:

a) that some CV are specifiable.

b) that whether some specified CV are satisfied is ascertainable in particular cases.

        Requirement (a) seems to be legitimate and it is needed to substantiate the conceptual distinction between the veridical and the unveridical.  However, I think this can be met by TE.  Analogue to Martin's (1) can be specified:

If God exists and He decides to reveal Himself, and S's spiritual sensitivity is functioning normally, and S's spiritual environment is not prone to block a TE, then it would seem to S that God exists.

Similarly, analogue to (5) can be specified.  Of course, it would be objected that this is very vague.  But the same goes for SE as well: there we can also ask what exactly is normal lighting and normal functioning of perceptual apparatus.  Moreover, I think the above specification is not too vague to provide the conceptual distinction needed (I do not claim that it can have any other use): it does give us some idea when a TE would be veridical.  It also seems unreasonable to demand the criteria to be very precise  and fully determinate in every situation.  (It is dubious whether the tests of SE can satisfy this demand.)  There is no obvious reason to claim that for a type of experience to be roughly reliable, we must have such precise tests.  In many other situations, we can make objective distinctions without having precise criteria: legal judgment, choice of large scale scientific theories, judgments of others' personality and motive, etc.  Perhaps lingering behind all these is the ghost of verificationism? 

        So if requirement (a) is not damaging, what about requirement (b) then?  We can have two types of CV: some confirming (CVp- 'p' for 'positive') and some disconfirming (CVn- 'n' for 'negative').  (Talk of proof here would seem unrealistic.)  Suppose the following can be formulated:

*) If CVp(SE) obtains, then a token SE is probably veridical.

**) If CVn(SE) obtains, then a token SE is probably unveridical.

        For example, CVp may include consensus and CVn may include failure to be detected by a scientific equipment.  So perhaps Rowe's claim is that PCT, to be applicable, requires these conditions:

1) we know something like (*) or (**), and

2) In particular cases, we know whether (*) or (**) obtains. 

Now clearly my reply to Martin can be generalised to show that Rowe's type of objection would face enormous problems.  It  seems impossible to satisfy the above conditions (1) and (2) if we don't have basic trust in some SEs.  To do the latter would be unjustified because we cannot yet apply the PC or PCT before satisfying conditions (1) and (2).  Rowe's version of PC would probably land us on the sceptical bog again.  The conclusion is that if we insist that we must have noncircular CV before we can apply PCT, then it would be damaging to both SE and TE.  In fact, CV(SE) are circular in two respects: a) it is derived from and justified by SE itself and b) in applying them, we need SE to do that.  If it is the case, should we not be open to the possibility that CV(TE) can also be derived, justified and applied in similar ways?  The reply of Losin is apt:

"Rowe has simply assumed that reasons drawn from experiences of God cannot themselves be 'reasons for thinking that particular experiences of God are delusive', that experiences of God cannot themselves provide a (fallible and provisional) means for the critique of other such experiences.  I see no reason to think that this assumption is true, and good reason to think that, when suitably amended and applied to sensory experience, it is false.  Nor do I see the slightest reason why we cannot use knowledge or beliefs about God not gleaned from experience of God to identify and dismiss particular experiences of God as non-veridical" (p.69).

        Now let us deal with the suggestion that there is an intrinsic connection between CV and justifiability of experience.

Type-veridicality, Token-veridicality and Type-reliability

        In order to clarify the issues, I propose to make the distinction between type-veridicality and token-veridicality.  Type-veridicality is the verisimilitude of the basic ontology of the type of experience.  So theistic experience is type-veridical if there is a being whose nature is quite like that of God, as believed by the major theistic traditions.  In contrast, we can consider the token-veridicality of a particular theistic experience, e.g. Isaiah's vision in the temple.  (It should be noted that token-veridicality of an experience is in general relative to the level of description of that experience.)  We can further define 'type-reliability' of a type of experience as the probability of a token of that type to be veridical.  Clearly this is a continuous variable which takes a value from 0 to 1. When we say loosely that a type of experience is type-reliable, it may mean only that the probability is larger than 0.5.  We should note that the question of type-veridicality is in principle separable from the question of type-reliability.  One case of token-veridicality is sufficient to establish the type-veridicality.  So the type-veridicality of a type is compatible with gross unreliability of the type, i.e., the token-unveridicality of the majority of the token experiences of that type.  However, even in this case the type of experience is not necessarily absolutely unreliable and evidentially irrelevant.  This possibility will be explored in chapter 11.

        Let us now raise the question of the epistemological relevance of the CV.  Consider this argument:

A) If there are no CV to distinguish the veridical tokens from the unveridical ones, then we can't know which tokens are veridical.

B) If we can't know which tokens are veridical, then we can't know whether the type is veridical.

C) Hence if there are no CV, then we can't know the type-veridicality thesis.

        The general idea is that for a type of experience to have evidential force, tokens of it have to have evidential force.  However, if we allow that the tokens can be unveridical, then only the veridical tokens can have evidential force.  But if we can't distinguish the veridical tokens from the unveridical, then we have no access to the evidential base.  Perhaps this kind of reasoning is behind the No Criteria Objection.  Of course if the objection is interpreted in this way, then one reply is that we do have CV for TE which are established in analogous ways to the case in SE.  Now I would further suggest that we do not need to decide the token-veridicality in each case before the type-veridicality thesis can be judged to be plausible.  To make a case for the latter, we only need to show that it is unreasonable to believe that token-unveridicality is universal.  Consider this argument:

D) If any token of a type is token-veridical, then a type of experience is type-veridical.

E) If it is reasonable to believe that a type of experience is not type-veridical, then it is also reasonable to believe that no token of that type is token-veridical.

F) It is not reasonable to believe that no token of that type is token-veridical, i.e., all token experiences are totally delusory.

G) Hence it is not reasonable to believe that the type of experience is not type-veridical.

        I suggest this argument can work for TE.  The crucial premise, of course, is (F).  It seems to me and some others, e.g. Gutting, that (F) has strong intuitive appeal.  Though it may be hard to produce an explicit argument for (F), it may nonetheless be  compelling after we have surveyed all the relevant evidence.  It is not necessarily dependent on demonstration of token-veridicality of a particular TE.  When we ponder the numerous TEs, their enormous effects sometimes, the honesty of the witnesses, the depth dimension of life and so on, it seems hard to believe that all of them are delusory.  Moreover, we may also encounter the life story of a person who has dramatic experiences of God throughout his life.  We also find that the person is honest, sane, wise and intelligent.  We also find his story corroborated by many others' stories throughout history in many countries.  Isn't it rash to say that all of them are entirely and chronically deluded?  Ordinary people may also find it hard to produce an explicit and water-tight argument for his belief that the earth is round rather than flat.  Yet we won't deny that their intuitive judgment, which is based on many empirical clues, is rational.  Can't we also claim that some ordinary believers in God, which have access to the relevant experiential evidence for God, can be rational in judging that (F) is true? 

        Nevertheless, it would be nice if we have philosophical arguments to back up the intuitive judgment.  (F), when applied to TE, is roughly equivalent to "it is not unreasonable to believe that at least one token TE is veridical."  This is exactly the conclusion I am trying to defend via the CTA.  I have defended a Criterion of Ultima Facie Justification and suggested that it can be used to judge when a token TE is veridical.  Of course, we can then point to a particular TE and argue for it.  However, I would prefer another strategy: let us point to a collection of TEs and suggest that it is reasonable to believe that at least one of them is veridical.  I would choose this subset of TE: experiences of the presence of God.  Firstly, the claim of this TE is more modest and it is not liable to be in serious conflict with many TEs or non-TEs.  Secondly, it is quite widely shared cross-culturally and across the epochs.  Many spontaneous experiences of non-believers  are also similar to this sense of presence of God (Hay).  Thirdly, these TEs occur spontaneously in very diverse situations to many different kinds of people which are not in abnormal psychological or physiological conditions.  This would make the naturalistic explanation of all these quite difficult.  So this sub-type of TE seems least problematic and it is easier to argue that it is undefeated.  My conclusion is that (B) is false: it is possible that while we cannot pinpoint one particular token as veridical, we can still argue that some token or other can be reasonably judged to be veridical and hence the experience is type-veridical.  There is no intrinsic connection between having precise CV and the justifiability of a type of experience.

        My strategy has a parallel in SE.  Some argue for the veridicality of SE by the Paradigm Case Argument.  The idea is that when we decide the veridicality of a type of experience, we should not focus on the most problematic tokens, e.g. the optical illusions.  Instead we should concentrate on the least problematic tokens, i.e. the paradigm cases.  However, it seems to me this move is best construed as pointing to a subset of SE which is regarded as paradigmatic and then arguing that at least one token of this is veridical.  It is not so wise to stake the argument on a particular SE.  Moore has once given this example of his certain beliefs: there were windows behind the curtains in the hall where he was lecturing.  It turned out to be false.  What is mistaken here is not exactly a SE but I hope the idea is clear.  If we point to a particular table in front of us, isn't it genuinely possible that later on a much more convincing sequence of SEs would disconfirm that?  For example, later experiences of similarly 'high quality' may suggest to one that the table was only dreamt of.  However, this sort of possibility wouldn't disprove the claim that at least one of these 'high quality' SEs is veridical. 

Criteria of Veridicality of Sense Experience

        Let us explore further the question how we can determine and justify some CV in the first place.  Consider these two statements:

1) We possess some criteria of veridicality.

2) We can identify tokens of veridical experiences.

       The dilemma comes out clearly when we ask which is logically prior, (1) or (2).  Neither option seems palatable.  If we don't know how to identify tokens of veridical experiences, it is hard to see how we can arrive at some CV.  But if we don't have some CV and our experiences are fallible, how can we identify tokens of veridical experience?

        Suppose (1) is logically prior.  Obviously we are faced with the problem how we arrive at those CV.  Nobody would suggest that they are revealed by God.  Perhaps it can be suggested that the relevant criteria are conceptually derived from the ontology of physical objects in space-time.  Gale seems to adopt a similar line.  Now let us take Gale's understanding of the nature of physical objects as given and explore his position.  According to him, from the nature of the physical objects, we can derive these results: if a SE is veridical, then we can predict future SEs of the subject and others, and also the subject's sensory faculty is in good working order and he is in the proper spatio-temporal position to perceive the apparent object of the SE and so on.  (He gives 11 tests altogether.)  Suppose these criteria are then formulated:

CVp

If the prediction is borne out and the man is in the proper position and so on, then the SE is probably veridical.

CVn

If the prediction is not borne out or the man is not in the proper position or ..., then the SE is probably unveridical.

Now the same problem afflicts him: we cannot ascertain whether the CV are satisfied apart from some basic trust in SE.  Moreover, CVn seems to be less problematic since they follow from the above theory about physical objects.  It is otherwise for CVp.  If we are just given the theory about the physical objects and not their existence, then the satisfaction can be explained by the sceptics as well.  The theory need not be about some actually existing things: it can be in the mind of the evil scientist instead.  Perhaps the evil scientist is also a good philosopher: he first lays down some ontological descriptions of the nature of 'physical objects' and then derives the principles which govern the 'veridicality' or 'unveridicality' of SEs.  These principles are then used as the most basic axioms of his program which governs the inputs to a brain-in-a-vat.  (Of course he also needs to draft some contingent causal laws which conform to the axioms.)  In this way all the experiences of the brain-in-a-vat would indeed conform to Gale's theory of SE.  Perhaps the brain-in-a-vat also reasons as Gale does and concludes that all his experiences must be real?!  So Gale's tests are not an adequate reply to scepticism.  However, his discussion certainly shows further the coherence of SE.  It is because the tests need not be passed by our experiences and the fact that they are frequently passed is impressive.  The result is that if the basic trust in our SE is justified, then unveridical SEs are indeed much easier to weed out.  This would increase the intra-coherence of SE.

        Now I have to raise the question: how is Gale's understanding of physical objects derived from the first place?  Perhaps we can rely on the grasp of some a priori necessary truths.  But they do not seem to be sufficient to sustain such a detailed theory like Gale's.  For example, Gale assumes that physical objects are stable and universally accessible to persons.  I don't think it is an a priori truth.  (Conceivably, our SEs could be different from what they are.)  So our theory is partly determined by the contingent characteristics of the SEs we do have.  Without prior trust in these experiences, can we trust the theory derived from them?

        So all these questions push us to take seriously the case that (2) is prior to (1).  We have a priori principles which justify our basic prima facie trust in our SEs.  The veridical tokens are then established by the ground level sifting.  Only after that can we investigate the perceptual conditions of veridical SE and the nature of the physical objects.  Of course, such knowledge about physical objects and perception would help us to lay down our criteria of veridicality.  The process is ongoing and further knowledge may help us to modify, revise or add to our previous CV.       If it is the correct picture, then we shouldn't expect the same CV to apply to different sorts of experience.  Take interpersonal experience as an example.  How is the consensus test to be applied here?  Chesterton has an example roughly like this.  Suppose we are told that a young lady calls her fiancee by a very intimate nickname.  Are we going to test it by summoning fifteen psychologists to observe their interaction and conversation?  Similarly, when A tells us that B told him a very traumatic experience, we would not insist that B has to tell the same experience before all of us!  It is because by the very nature of the case, these experiences, if veridical, are not likely to be publicly corroborated.  So it seems to me, the following principle should be kept in mind when we decide the CV for a type of experience:

The CV for a type of experience should be appropriate to the nature of the alleged object of experience and the subject of experience.

This principle is operating all along even in SE.  We tend to forget that there are many kinds of physical objects and perceivers as well.  The way to test the existence of wind would be very different from that of a table, not to mention atoms, EM fields, etc.  We won't apply the test of touch to a far away flying object.  We won't insist that a table has to be seen by a blind man.  All these illustrate the fact that our actual tests of a particular physical object are actually context-dependent, i.e. depending on the whole perceptual context which is constituted by the subject, the object and the environment.  Any change of this context would affect our decisions as to which tests are relevant.  It is a false picture to suggest that we have fixed, uniform tests even in SE.

        Let us look at some concrete example.  We seem to see a bending stick when part of a straight stick is immersed in water.  We would count that as an illusion while taking the perception of the straight stick as veridical.  This distinction is taught to us since our early days and most of us would take it for granted.  But can we give reasons for this move?  Let us not first take into consideration the science of optics.  If the optical account of the illusion is necessary for the above distinction, then we are denying the right of most of our ancestors to make this distinction.  This seems most implausible.  If we just consider the more common sense response, then probably most of us would reason as follows: "If we take the stick out of the water, the stick again looks straight.  Furthermore, if we grope for the stick partly immersed in water, it still feels straight.  Therefore it must remain straight although it seems to be bent."  However, this reasoning can be doubted.  For the first reason to be conclusive, it has to make two more  assumptions: firstly, the visual perceptions are veridical; secondly, the shape of the stick has remained constant throughout the process.  If we are prepared to grant that the stick is bent when it is partly immersed in water and it is straight when all of it is out of the water, then we can insist that both perceptions of a straight stick and a bent one are veridical.  As for the tactual experience of straightness, why can't we insist that vision is more reliable than tactual experience and take the tactual experience as illusory instead of the visual one?  To settle the question in favour of the common way of interpretation we seem to need two rational principles:

Other things being equal, we should choose an interpretation of experience which fits with the simpler ontology.

Other things being equal, we should choose an interpretation of experience which would render more of our perceptual experiences veridical.

        The second principle in itself is not sufficient: both interpretations would need to render some experience illusory and it does not decidedly favour the common sense interpretation.  However, the common sense interpretation posits a sort of stick which would not suddenly change shape when immersed in water and this ontology is simpler.  Someone may object that a stick is just not the type of object that would easily change shape.  This piece of knowledge, it might be argued, is the foundation of our common sense interpretation of experience rather than my alleged rational principles.  This reply won't do.  The above statement about the stick seems obviously true only because it is repeatedly confirmed in our experiences.  But given that our experiences can be illusory, can we decide what is confirmed by our conflicting experiences unless we have a way to distinguish which experiences should be counted as veridical and which illusory?   The answer seems to be no.  If so, then the above principles seem to be basic principles of experiential interpretation in accordance with which we posit the veridical-illusory distinction.  The first one is effectively a form of the Principle of Simplicity.  The second is a variant of my Rule of Ground Level Sifting which is a corollary of the PCT.  In other words, adoption of this principle already commits us to a basic prima facie trust in our experiences.  If so, then my CTA can nicely account for the way we make the veridical-illusory distinction. 

        Similar arguments are possible for other CV.  For example the consensus test is effectively the employment of the Principle of Consensus which is a corollary of the CTA.  As for the test by scientific equipment, it seems to be a case of my feedback sifting.   I agree that consideration of ontology comes in when we make the veridical-illusory distinction but this distinction is still prior to the knowledge of the nature of the physical world.  To gain such knowledge we need to justify it by veridical experiences and hence before we can know which claims can be regarded as knowledge we need to decide which experiences are veridical.  My idea is that our knowledge of the nature of the physical world is justificatorily dependent on the application of PCT to SE.  Of course when it has been built up, it can be used to do feedback sifting and to formulate more precise CV.  In SE, the ontology of the physical objects is believed to be correctly described by our scientific theories and that is why science can play a crucial role in confirming (or disconfirming) our common sense way of making the veridical-illusory distinction.  I have earlier raised the question why, in the case of the apparently bending stick, the visual experience is taken to be illusory rather than the tactual experience of it.  One powerful reason for this is provided by the theory of geometrical optics because it gives a more detailed explanation of the visual illusion by positing the process of refraction of light rays by water.  On the other hand, if we take the interpretation that the stick is bent when partly immersed in water, we do not have explanations why the bending occurs and how the alleged tactual illusion of straightness occurs.  So another principle seems to be involved:

If two interpretations of experience both render some experiences illusory, the interpretation which gives better explanations of the alleged illusions is to be preferred.

This seems to be another example of feedback sifting by inference to the best explanation. 

        My conclusion is that a plausible account of the 'whence' of our criteria of SE actually fits quite well with my CTA.[1]  This process of deriving the CV can be summarized:

1) Basic prima facie trust in our experiences.

2) By explanatory ascent and data sifting, we decide which types of token are more reliable and which unreliable.  At this level, some preliminary CV would emerge.  The CVp are basically derived by observing which tokens exhibit higher degree of coherence with the majority of other token experiences of oneself and others.  The CVn are derived by observing which type of tokens conflict with the more established ones.  (CVn are actually the Second Order Critical Principles we have talked about.)  Investigation of perceptual conditions would help both.

3) It is possible that by further explanatory ascent, we can have a more detailed theory about the object of experience.  This theory is then used to confirm, modify or revise our preliminary CV in such a way to gain overall coherence.  More precise CV may then result.

       

        So far I have argued that the lack of precise ascertainable criteria does not tend to make a type of experience unjustifiable or unreliable.  Nor can it render the PCT inapplicable.  So why do the critics think it is so damaging?  I suggest in the end such objections may only betray the bias that all experiences should be checkable by criteria similar to those used in sensory perception.  So it is the disanalogy with SE that is seen to be damaging.  This claim will be refuted in the next chapter.

        Some would suggest that although sensory perception can't be judged by non-circular criteria, its claim can be sufficiently justified by its coherence with other ordinary perceptual beliefs or "one's background theory" instead.  For example, David Conway thinks that what we call circularity is not a problem: "Indeed, this reciprocity is crucial; any one test checks on the veridicality of the others and the others check on it.  The over-all criterion is the coherence of the results of the individual tests.  Ultimately an experience which coheres with our other experiences is accepted as veridical; one that does not cohere is not" (p.159).  This raises the question whether a purely coherentist justification is fully adequate.

Coherentist Idea of Justification

        This position denies that there can be a foundation for our epistemic system.  Every belief depends on at least some support from some other beliefs one holds in order to be justified.  Usually it is stated that a belief is justified by its coherence with the believer's entire belief system.  The precise characterization of this support varies from author to author.

        Coherentism is at odds with the PCT since it alleges that justification of any belief has to be based on the relation of this belief with one's entire belief set.  However, coherentism itself has grave problems.   For example, there is the 'fantasy argument': beliefs arising out of sheer fantasy or wishful thinking can be as coherent as you like but it seems counterintuitive to claim that any belief of this corpus is justified by its coherence with the whole corpus.  On the other hand, in other respects, coherentism seems too strong:

"Couldn't there be little islands of belief unsupported by the rest of the mainland corpus?  I may now be justified in believing that I have an itch in my left thigh even though this belief neither supports nor is supported by any other belief of mine" (Alvin Goldman, p.100). 

In other words, coherentism does not take many of our spontaneous beliefs or epistemic seemings seriously enough.  But there is a place for coherence to play in justification.  Suppose we have some justified beliefs already.  Presumably they have some likelihood to be true.  It surely is a virtue for a new belief to cohere with these.  Hence if a belief is coherent with some set of prima facie justified beliefs, it is less likely to be false and the coherence counts towards its justification.  Indeed the CTA needs to introduce coherence considerations in data sifting and explanatory ascent.  However it has to be emphasized that coherence considerations only become important after we have applied the PCT.  The key problem is that if all the beliefs of one's belief set has no independent evidential force, it's doubtful that mere coherence with these beliefs could generate justification.  Consider Goldman's comment:

"coherence is not the highest standard of justificational rightness but a derivable, subsidiary, standard ...  Coherence enters the picture only because coherence considerations are generally helpful in promoting true belief.  On the assumption that prior beliefs are largely true, new prospective beliefs should be examined for consistency with prior beliefs" (Ibid; italics mine). 

This would explain the attractive force of coherentism.  In considering coherentism we are prone to consider coherence with our beliefs which we do regard as prima facie justified.

        I need to add a word here.  If coherentism is true, my argument for TE certainly needs to be modified.  However, I think the task to justify TE would even be easier because in that case we only need to show that the veridicality of that TE is coherent with the theist's doxastic system.  Anyway, in this thesis I will also argue that TE is coherent with most of our other experiences.  That is why I also need to spell out clearly my understanding of 'coherence'.

Kinds of Coherence

        I now try to outline several kinds of coherence I will be talking about:

A) Propositional Coherence:

        This is the kind of coherence that can be achieved between several propositions, or between a proposition and a theory, or between theories.  I propose to distinguish five levels of coherence:

1) Incompatibility: the fact that the several propositions (or theories) being considered cannot be simultaneously true.  This is a decisive kind of incoherence which should be avoided.  The following four levels all imply compatibility.

2) Incongruity: two propositions can be compatible but they somehow do not fit together.  For example,

(P1) Scrooge is a miser.

(P2) Scrooge gives one hundred pounds to a beggar.

        (P1) and (P2) are by no means incompatible but we all feel that they do not seem to belong together.  Given (P1), (P2) is surprising, unexpected, and calling for explanation.  We can also use the language of probability here: the conditional probability of (P2) given (P1) is quite low.  However, I think we can intuitively judge two propositions to be incongruous without basing it on probability considerations.  Incongruity is not decisive because it is always possible to add a third proposition to remove it.  In the above case, (P3) can do the job:

(P3) Scrooge thought the beggar was a disguised billionaire who would reward those who were kind to him.

        However, if such salvaging propositions, though entirely ungrounded, are spun out just to save incongruity, then the resulting system would suffer from being ad hoc.

3) Disparateness: if two propositions are quite unrelated and they do not affect the probability of each other, then they are disparate.  For example, (P4) seems to be just disparate with (P1):

(P4) Scrooge is five feet tall.

4) Congruence: if two propositions seem to go together, fit nicely with one another, etc., then I say they are congruent.  When probability assessment is applicable, two propositions are congruent if they mutually tend to raise their probabilities.  (P5) is more or less congruent with (P1):

(P5) Scrooge turned away his friends who came to ask for financial help.

        Analogy seems to be a kind of congruence.  When scientists discover that theories in different fields have analogous structures, e.g. in mechanics and circuit theory, the confidence in those theories is somewhat boosted.  At least, they can simplify their conceptual scheme by subsuming both theories under the same abstract mathematical structure.

5) Consilience: this can be regarded as only a high degree of congruence.  However, I would like to emphasize the role of theory in this category.  Consider:

(P6) Jane was dead and John was reported to be just beside the corpse.

(P7) John's hand was holding a knife which was stained with blood.

(P8) John was crazy for Jane but Jane repeatedly rejected him.

        Now these propositions seem to point to the same conclusion:

(P9) John has murdered Jane.

        In light of this conclusion they also reinforce the credibility of one another.  I call this kind of coherence consilience.  The individual propositions can be likened to the threads of a rope or the legs of a table.  Just as the threads are consilient after being woven into a single rope and as the legs enable one another to stand by being connected to a common surface, it seems that propositions are consilient by being incorporated into a simple theory or hypothesis.  In the above case, (P6) to (P8) may possess some initial credibility.  (P9) is the simplest hypothesis which explains (P6) and (P7) while (P8) provides the motivation of the murder and, in a sense, explains (P9).  It is in virtue of these coherence relationships that (P9) is vindicated by an inference to the best explanation.  But it also seems to be the case that the availability of (P9) shows the consilience of (P6) to (P8) and in this way reinforces the credibility of each.  It is because each of (P6) to (P8) is defeasible and can be subjected to doubt.  Now because of the murder theory, they are somehow linked together.  To satisfactorily defeat each, we need to defeat the rest (or at least account for them).  This is a more difficult job than it would be if each of (P6) to (P8) is disparate relative to each.

        It seems that well-accredited scientific theories usually show this kind of consilience.  For example, Newtonian mechanics incorporate facts about motion on earth, planetary motions, tides, velocity of sound, and so on into a single theory.  These heretofore unconnected facts then manifest a kind of consilience which renders the whole theory worthy of rational assent.  Another example is the theory of plate-tectonics (cf. McMullin in Leplin).  Aronson's theory of confirmation also emphasizes similar features.  He argues that scientific explanation, confirmation, etc. should best be construed as the process of mapping diverse empirical phenomena onto a common ontology.  It is interesting that consilience can also manifest in a kind of 'bootstrap'.  Many philosophers of science now realize that many scientific testing methods are themselves theory-dependent.  However, they do not then draw the conclusion that relativism is true.  However, they point out that even if the testing methods are theory-dependent or even circular, this does not guarantee that the result of the test will be positive.  So a positive result should still be seen to be confirming (Glymour, Boyd in Leplin).  Our earlier discussions show that CV(SE) actually manifest a sort of bootstrap.  Although those CV already presuppose some trust in SE, it still does not guarantee that subsequent SEs would satisfy those circular CV.  Hence when those CV are satisfied, it can still be seen to be confirming.  Lastly, of course propositions and theory can exhibit consilience by having intricate mutual entailment relationships. 

        It should be noted that the five levels form a continuum and the distinctions made, though helpful for critical discussions, are not hard and fast.

B) Experiential Coherence:

        Not only can propositions cohere or fail to do so; kinds of experience can do the same.  Of course one way to capture experiential coherence is to consider the coherence of the descriptions of those kinds of experience.  Their deliverances (as tokens or as types) can be incompatible or incongruous or merely disparate.  They can also be more or less congruent, for example, when they (as types) are analogous in their structure, characteristics and so on.  The experiential claims can also be consilient.  For example, visual, tactual, auditory and olfactory experiences are phenomenologically disparate but they can be mapped onto a common ontology of physical objects and reinforce one another.  This explains why they are grouped into one kind.  There are also particular aspects of experiential coherence.  Firstly, experiences can cohere by having similar phenomenal features.  Secondly, different kinds of experience can fuse together and form a single unitary experience.  For example, when I have an experience of driving a car, my seeing of it, my hearing of the engine noises, my feeling of the car's motions and so on, can be fused into a single experience.  I would call this kind of phenomenon  experiential fusion.  It would be very inconvenient indeed if human beings do not possess this kind of capacity. 

        We can now spell out the idea of inter-coherence more clearly.  For a type of experience to be coherent  with another, the ontology of their types and the majority of their tokens have to be compatible with one another.  Secondly, if the two have analogous structures and characteristics or phenomenal features, then they are more or less congruent.  The highest kind of experiential coherence seems to be achieved when we have experiential fusion of the two types and consilience of their experiential claims.  For a certain type of experience, its inter-coherence would be high if it is congruent or consilient with the majority of well-established types of experience.

Tests for TE

        In order to give a fuller reply to the No Criteria Objection, let us come to discuss the CV for TE and their justifiability.  First, let us consider Jantzen's suggestion:

"It is characteristic of REs that they have qualities of intensity and great significance- so much so that typically it is felt that the experience challenges and questions the mystic to his very core, that he would be betraying his own integrity and stultifying the possibilities of his own growth if he did not take it seriously.  The reciprocal questioning and deepening of understanding which then occurs contributes to the process in which self-integration and wholeness, including deeper sensitivity to the needs and suffering of others, can develop in a quality of life lived in conscious relationship to a compassionate God.  Seen in this way, it makes sense of the relationship between specific experiences and the experienced quality of life, and indeed makes that life a continual testing of those experiences- and they of it" (p.289; italics mine). 

So can a mystical experience legitimately tip the balance in favour of belief in a personal God? 

"Well, no: not if it is a one-off odd ecstasy without bearing on the whole of life lived.  But what it can legitimately do is be part of a process of deepening understanding of oneself and others, part of which is bound sooner or later to bring about the recognition that self-transformation is in one sense necessary and in another impossible: more resources are needed than our own.  ...  as religious experience deepens into a quality of life lived in integrity, the reciprocity of encounter and response can legitimately tip the balance: it is rational for the mystic to believe in a personal God" (p.290). 

Because

"this reciprocity ... is part of the basic pattern of rationality, in which questioning occurs from within a perspective, but in which the answers obtained by that questioning can in turn modify that perspective, sometimes radically, leading to deepening understanding and thus to a new round of questioning" (p.289).

        Though I disagree with Jantzen's conservative estimate of the evidential force of REs, she still draws us to an important point: a RE or TE does not just occur, at least for some, out of the blue and then disappears without a trace.  It shapes a life and it has a dimension of depth that conveys a sense of reality that, at least for those who experience it, is hard to deny.  Of course, this still does not prove the veridicality of REs.  It is doubtful we can ever prove the veridicality of a single experience.  I am only saying we have no reason to expect the criteria of REs to conform to the pattern of the criteria of sensory experiences.[2]  If we drop this unreasonable expectation and look at the religious case, is it not clear that the tests we can reasonably expect are somewhat like the reciprocity tests described by Jantzen?  Henderson also claims to find similar ideas in Austin Farrer:

"Farrer insists that verifying evidence cannot consist in certain predilineated events which we observe.  For God is unique; he makes an unconditional claim upon our devotion.  To make him prove his existence through the production of happy effects would be to subvert the affirmation; it would be a subordination of God's will to ours rather than of ours to God's. ...  The affirmation of God, if genuine, is a practical affirmation which runs into the activity of integrating one's own will with God's. ... through the affirmation of God we become participants in a life which surpasses also the life in community with ordinary persons. ...  We verify the affirmation of God's existence, therefore, by entering into life-in-God and finding that the life can indeed be lived and lived in all circumstances and that it blesses" (pp.178-79). 

This blessing is the "blessing of a union of will with the primal will" which is also "the fundamental blessing of finding oneself where one belongs".

        Let me spell out more clearly my suggestion (the rationales for these criteria will then follow):

CVp(TE)

A TE of a person S is more likely[3] to be veridical if:

a) S has more TEs later in life; the probative force is stronger if these confirming TEs are more frequent and more various.

b) the original experience and the subsequent ones exhibit a high degree of coherence, e.g. consilience, with the original experience.  Reciprocity is a kind of consilience: the original experience and the subsequent ones lead to a continuous process of questioning and deeper understanding and further questioning and so on.  The original object of TE seems to have inexhaustible richness that eludes one's expectations.

c) a substantial amount of people also have similar experiences.  (The more the better.)  It is even better that the experience is shared at the same time, e.g. a corporate TE.

d) the experience is coherent, congruent or consilient with other experiences, e.g. moral or existential experiences.

e) the experience leads him on the road to self-integration, e.g. he understands himself better, leads a more fulfilling life.

f) the experience helps him to lead a morally better life, e.g. higher sensitivity to others, more loving, better moral insights, humility.

g) the experience produces a strong sense of reality which is not clearly induced by abnormal conditions or psychological processes.

h) the experience is accompanied by 'happy effects', coincidences, or miracles which are hard to explain naturalistically.

Normally, it would be objected that these tests draw on the theistic traditions' understanding of God.  But I have already shown that this procedure is not illegitimate.  So let me instead explain further why these criteria hold.  They seem to follow more or less straightforwardly from the categorial nature of the object of experience, i.e. God in this case.  If the experience is veridical and God exists, it seems likely that God will reveal to S further, if S is open to that.  Since God can manifest in many ways, it seems more true of God's nature if S's experience is confirmed by different modalities of TE.  So (a) is plausible.  (b) follows from the general probative force of coherence between initially probative experiences.  If God is independent and transcendent and the original experience is not self-generated, then further experiences should reveal unexpected, surprising characteristics of God and the object should appear to be unfathomable by experiences.  (c) follows from the Principle of Consensus.  If God is the source of all veridical experiences, a veridical experience should also be coherent with other veridical experiences.  As I will argue in chapter 10, moral experience and existential experiences should not be dismissed and they can plausibly be argued to come from God as well.  If so, (d) ensues.  If God's approach to man is meant to give his life meaning and wholeness and if God Himself is man's telos, a TE which achieves this is more probably veridical, hence (e).  If God is also the source of community and values, then contact with him is likely to produce moral transformation, hence (f).  If God is the Ultimate Reality and he can more directly act on the soul, then it is expected that genuine experience of God should have a strong sense of reality.  So (g) is not surprising at all.  If God is the Creator, he can also perform providential acts in order to confirm his purpose, hence (h).

        Now one crucial point is that the above are positive CV.  (Note there is only a 'if'- not 'only if'.)  We cannot argue that if either one of the above is lacking, then the original experience is likely to be unveridical.  This is not an ad hoc move.  In general I deny that CVp and CVn have to be symmetrical.  More positively, the asymmetry in this case is due to freedom of God and man.  God may not choose to use providential acts to confirm his purpose, hence (h) can't be used as a CVn.  Similarly, a man can resist further contacts with God and he may even deliberately run away from God after the first encounter.  Hence (a) to (f) can't be used as CVn.  However, as believers also recognize the fallibility of TEs, the lack of (a) to (h) does (and should) lessen his confidence in that particular TE.  The following principle seems true:

If a large number of conditions from (a) to (h) are not satisfied, then although the TE is not thereby regarded as unveridical, it will be treated with some reserve.

CVn(TE)

A TE of S is likely to be unveridical if:

a) it conflicts with the well-established TEs (by the above criteria).

b) it conflicts with other well-established experiences, especially moral experience.

c) it leads to a disintegrated life, e.g. insanity.

d) it conflicts with well-established knowledge.

My overall conclusion is that we can justify and specify some CV for TE.  The circularity is not damaging if the CTA is true and anyway CV for SE is similarly circular.  However, it should be conceded that the CV for TE are much vaguer than those in SE.  The CV for SE, in some situations, can be quite precise and their applicability often produces a bootstrap which further increases its intra-coherence.  Let's discuss whether this disanalogy (and others) is damaging.

---

[1] A similar picture emerges from the discussions of Alston 1991, pp.217ff.

[2] cf. Alston: "it is an unthinking parochialism or chauvinism, or epistemic imperialism ... to suppose the CMP (Christian mystical practice) is properly assessed in terms of the checks and tests appropriate to SP (sensory practice).  Judging CMP outputs on the basis of SP tests is no more appropriate than evaluating introspective, memory, or mathematical beliefs by the same tests.  The objection to CMP I have been considering is guilty of the same kind of chauvinism as Plato's and Descartes's low assessment of SP as lacking the precision, stability, and certainty of mathematics and Hume's low assessment of inductive reasoning as lacking the conclusiveness of deductive reasoning. ... I have been stressing the irreducible plurality of doxastic practices in the tradition of Reid and Wittgenstein" (1991, p.220).  However, I also emphasize that a general scheme like the CTA seems applicable.

[3] Remember the CVp are not meant to prove.  They are formulated on the assumption of the PCT.  It is possible that all the criteria are satisfied and the theistic experience still unveridical.  The same goes for CV(SE).