Cup of Nirvana Philosophical and Contemplative Explorations

Getting Sober about Survival (Part 3 of 3)

In my previous blog (2/11/14), I argued that (i) empirical survival arguments depend on the survival hypothesis having predictive consequences and (ii) the survival hypothesis has no predictive consequences unless it’s supplemented with various auxiliary hypotheses.  A substantial part of the discussion was devoted to laying out eight different assumptions required by survival arguments based on the data of mediumship.  If we were to look at survival arguments from cases of the reincarnation type or near-death experiences we would find a similar need to adopt other kinds of auxiliary hypotheses, but the focus on mediumship at least illustrates the auxiliary hypothesis requirement.  This will suffice for the line of argument in the present blog, which is that the auxiliary hypothesis requirement generates a problem for empirical survival arguments. 

The problem of auxiliary hypotheses arises from the fact that the auxiliary statements needed by survival arguments are not independently testable.  If this is correct, then we don’t actually know whether postmortem survival as such would lead us to expect any empirical phenomena, much less what the general or specific observational features of the world should be.  If the justification of the claim that there is empirical evidence favoring the hypothesis of survival depends on the survival hypothesis having predictive consequences (of even a general sort), then empirical survival arguments face a formidable difficulty.  It would appear that we are, at best, at the mercy of conjecture, which of course serves competing hypotheses equally well.

The situation is parallel to arguments from complex adaptation in organisms to an intelligent designer.  In the first blog of the present series, I looked at Elliott Sober’s critique of intelligent design arguments.  On this view, we cannot determine whether postulating an intelligent designer would lead us to expect the salient features of living organisms.   The reason:  we cannot independently test the range of auxiliary hypotheses that would, in conjunction with the hypothesis of an intelligent designer, lead us to expect the observational data.  In particular, we cannot test hypotheses about the designer’s requisite abilities and goals.  How, then, can we justifiably say anything about what the hypothesis of intelligent design as such would lead us to expect? 

In the present blog, I’ll explore the problem of auxiliary hypotheses for survival arguments, again illustrating this with survival arguments based on the data of mediumship.  What I hope to show, at least in a preliminary way, is that there is a genuine and serious problem here that undermines classical empirical arguments for survival, especially where postmortem survival is treated as a scientific or quasi-scientific hypothesis.

1.   The Independent Testability of Auxiliary Hypotheses

As explained in the previous blog, single hypotheses rarely have significant predictive consequences. The Duhem-Quine thesis in the philosophy of science highlights the point that predictive consequences depend on content provided by sets of statements taken together.  This is true in both scientific theorizing and in a variety of everyday inferences.  If we treat evidence collected at a bank robbery as evidence that Mr. Phinuit robbed the bank, this depends on establishing a connection between the evidence and Mr. Phinuit being the robber. Auxiliary hypotheses are enlisted to make this connection, e.g., hypotheses about Mr. Phinuit’s physical traits (e.g., fingerprint pattern, height and weight, facial features, speech patterns and accent), the make/model of his vehicle, and his whereabouts at the time of the robbery.

The dependence on auxiliary hypotheses introduces a minor wrinkle in the procedure of hypothesis testing.  In testing some hypothesis by its predictive consequences, we are trying to evaluate or assess the merits of the hypothesis by way of its connection to observational evidence.  If hypothesis H makes the observational evidence O unsurprising (i.e., leads us to expect O), then—to that extent—O evidentially supports H.  If O is not what we would expect, then O counts against H to some extent. If some different hypothesis H* renders O less surprising than H, then O favors H* over H.  These relations capture some widespread intuitions about how evidence supports hypotheses.  However, the role of an auxiliary hypothesis a is such that, in its absence, we can’t really say what H would lead us to expect.  It’s H + a that leads us to expect observational evidence O, not H alone. 

The importance of testing auxiliary hypotheses arises here.  If we had no way to test a, we would not be able genuinely to test H itself by means of predictive consequences.  This is most apparent in cases where the observation we would expect given H + a doesn’t pan out. What do we conclude?  Does the failed prediction count against H, a, or both?  Which statement(s) should carry the burden of epistemic culpability?  Suppose I adopt the hypothesis that Mr. Phinuit, a Frenchman, robbed the Bank of America in New York City.  It would be natural to expect that if the robber spoke during the robbery, witnesses would report that the robber spoke with a French accent.  Suppose, though, that the witnesses all report that the robber spoke with a thick Bronx accent.  Does this count against the hypothesis that Mr. Phinuit robbed the bank?  It’s not clear because our expectation that the robber would speak with a French accent is based not merely on the hypothesis that Mr. Phinuit robbed the bank, but on the additional assumption that, being French, Mr. Phinuit would speak with a distinguishable French accent.  But in this case, the failed prediction (i.e., the robber would speak with a French accent) might indicate that Mr. Phinuit is not the robber or that Mr. Phinuit, though a Frenchman, does not always speak with a French accent.  In other words, the auxiliary hypothesis may be what needs to rejected or modified in some way, not the hypothesis concerning who actually robbed the bank.

If we had no way to test the auxiliary hypothesis about Mr. Phinuit’s accent, it would be difficult to decide what the hypothesis that Mr. Phinuit robbed the bank would lead us to expect with respect to the accent the robber reportedly used.  How do we know whether it’s surprising or not that the robber spoke with a Bronx accent, given the supposition that Mr. Phinuit is the robber?  However, now suppose that we had a way to test the auxiliary hypothesis about Mr. Phinuit’s accent.  Perhaps further investigation turns up evidence that Mr. Phinuit, though he normally speaks with a French accent, has the ability to speak convincingly with a Bronx accent.  Suppose that video documentation is uncovered that shows Mr. Phinuit in an acting gig two years earlier in which he played a New Yorker and displayed an impressive Bronx accent.  Here we acquire evidence that the auxiliary hypothesis is false, or at any rate in need to modification.   Our ability to test the auxiliary hypothesis concerning Mr. Phinuit’s accent enables us to determine that the hypothesis that Mr. Phinuit robbed the bank is at least consistent with evidence that otherwise seems quite surprising.  Furthermore, while we might have simply modified the assumption about Mr. Phinuit’s accent, the ability to do so on the basis of independent evidence helps avoid ad hoc adjustments to a theory to retrofit data that are otherwise not to be expected. 

2. Survivalist Auxiliary Hypotheses

Now let’s return to the survival hypothesis.  

As I argued in the previous blog, a simple hypothesis of survival—positing the postmortem survival of the self or individual consciousness—is not robust enough in content to lead us to expect any of the data associated with mediumship.  We must adopt various assumptions about the knowledge, powers, and intentions or purposes that some persons would have if they were to survive death.  We must also make some minimal assumptions about the process of discarnate communication, for example, assumptions that account for “communicators” providing incorrect information on matters we would otherwise expect them to know. The specific auxiliary hypotheses I sketched were as follows:

[A1] There are some living persons P such that, if P were to survive death, P would be consciousness in a discarnate state, where “discarnate state” refers to a state of existence without a physical body. 

[A2] There are some living persons P such that, if P were to survive death, P would retain many of the detailed and highly specific memories of their ante-mortem existence.

[A3] There are some living persons P such that, if P were to survive death, P would possess knowledge of events taking place in our world after their death or the states of mind of living persons.

[A4] There are some living persons P such that, if P were to survive death, P would possess the desire and intention to communicate with the living.

[A5] There are some living persons P such that, if P were to survive death, P would possess the ability to communicate with the living. 

[A6] There are some living persons P such that, if P were to survive death, P would exhibit efficacious psychic functioning in the form of extra-sensory perception and psychokinesis.

[A7]  There are some living persons P* (where P* is a subset of P) such that, if P* were to survive death, P* would retain some of their significant general and particular skills and personality features. 

[A8]  Either C1, C2, or C3, where

C1: There are some living persons P such that if P were to survive death and communicate with the living at postmortem time t1 . . . tn, P’s cognitive and causal powers would become attenuated during t1 . . . tn.

C2: There are some living persons M such that if M were to receive information from some discarnate person Pi at time t1 . . . tn, the information would be subject to a cognitive process in which filtering and interpretation by the medium’s own mind lowers the accuracy and reliability of the content of the communications.

C3: There are some living persons P such that if P were to survive death and communicate with the living, certain modes of communication would produce more accurate and reliable information than others.

I refer readers to “Getting Sober about Survival II” for a more detailed discussion of these auxiliary hypotheses as requirements for arguments for survival from the data of mediumship.

3.  The Testability of Survivalist Auxiliary Hypotheses 

Some philosophers might wish to argue that we have evidence against some of the above auxiliary hypotheses, even if we don’t have evidence against survival as such.  For example, one might object to [A1] on the grounds that consciousness is dependent on a functioning brain. Or we might suppose that, even if consciousness were to persist after death in a discarnate state, it would be substantially discontinuous with our ante-mortem consciousness.  We might not remember much of our ante-mortem life.  Our purposes might be different.  We might not have any epistemic access to postmortem events taking place on earth, much less causally interact with the world or living persons.  Hence, even if the self were to survive death, it would not be capable of a rich conscious life, at least not in the absence of a body or some appropriate physical substratum.

However, the force of the above objections depends on the assumption that survivors would not have bodies of any sort.  And that’s just not clear.  And if communicators in mediumship are who they say they are, many of the deceased are at least of the opinion (rightly or wrongly) that they have bodies of some sort.  The problem, as I see it, is not that we have evidence that these auxiliary hypotheses are false.  The problem is that we have no way to determine that they are true, that is, no way to justifiably determine this independent of the hypothesis of survival and the data that are being adduced as evidence of survival.  

But let’s look more closely at the problem of the testability of survivalist auxiliary hypotheses.

I stated the auxiliary hypotheses above in the subjunctive mood, as subjunctive conditionals, specifically as conditional statements that state what would happen (or probably happen) if such-and-such were true, where the antecedent is entertained as a hypothetical situation, not a contrary to fact condition (a so-called “counterfactual”). 

To understand the difficulty with the independent testability of the survivalist auxiliary hypotheses, consider first the justification we have for many kinds of similar subjunctive conditionals.  I can say with relative ease what would happen to a glass jar if I dropped it from my second-story window onto a concrete driveway.  After all, there is independent evidence that jars of “this sort” break when impacting surfaces “like this” after being dropped from a distance “like this.”  The hypothetical situation closely resembles others that have actually taken place.  But even in the absence of this, if I understand the properties of the glass jar and the concrete, I can deduce that the jar will shatter on impact against a concrete surface if the glass jar impacts the ground with a certain velocity.  Our justifiably believing what would happen to the jar under the specified conditions is based on background knowledge, including various empirically testable claims about the properties of the objects in question.

But let’s take an example that’s a bit closer to home.  Survivalist auxiliary hypotheses involve claims about what human persons would know, desire, intend, and do in some hypothetical situation, namely in a postmortem state.  We’re often justified in believing what living persons would (probably) do under certain “hypothetical” circumstances.  We may believe that if John saw a person drop a $20 bill, then he would (probably) take it, or if Mary visited Oxford during the summer, then she would (probably) tell me about the Bodleian library.  To the extent that statements like these are justified it’s because we already know about the character or behavioral patterns of John and Mary, and we also know enough about the hypothetical situation to relate it in the appropriate way to the character or behavioral patterns of Mary and John. Maybe we have independent evidence that Mary likes libraries and tells friends about the details of her trips. Perhaps I’ve watched John pick up money people drop and pocket it himself.  So we can extrapolate what to expect of them under hypothetical situations, especially if they closely resemble actual, past circumstances.  Otherwise stated, what I independently know (or can test) about Mary and John, and what I independently know (or can test) about features of various hypothetical situations, gives me good independent reason to believe something about what persons would or would not do in a range of hypothetical situations. 

But are we in a similar position with respect to subjunctive conditionals that state what some (indefinite) living persons would remember, know, desire, intend, and efficaciously execute if they were to survive death?  Here it would seem that we don’t have access to the right stock of background information against which our conjectures could be empirically tested.  We know a lot about living persons, and we can extrapolate much about what they would probably know, desire, intend, and be able to efficaciously execute in fairly diverse range of hypothetical situations, but this is because we’re assuming that they are living persons in situations that, while hypothetical, are known to bear enough resemblance to actual life situations for us to have the appropriate background knowledge or procedures for empirical testability.

To see the difficulty here with greater clarity, consider our epistemic situation vis-à-vis an exotic hypothesis designed to explain the mysterious disappearance of massive amounts of sugar from a sugar factory in Santa Rosa, Texas.  We could postulate the covert operation of invisible time travelers from earth’s distant future to explain this datum.  Of course, the datum would be unsurprising given this hypothesis only if we embedded it in a certain story supplied by various auxiliary hypotheses: (i) humans have powerful cravings for sugar, (ii) there is an abundance of sugar in the world today, (iii) sugar will become increasingly scarce in our distant evolutionary future, (iv) environmental conditions in our distant evolutionary future will make the mining of natural resources for fructose impractical, and (v) new advancements in technology will make it possible for future humans to travel to earlier time periods in human history when sugar was produced in abundance and transport limited amounts of sugar back to the future. 

The time traveler theory is implausible in part because some of the auxiliary hypotheses are unwarranted and untestable.  One of these concerns the physical (and some would say “logical”) possibility of time travel. More subtly, though, the time traveler theory makes a crucial unstated assumption, namely that the human species will retain its current degree of craving for sugar into our distant evolutionary future.  However, the kinds of desires and intentions humans form in the distant future will be based on their actual needs and interests at that distant stage in their biological evolution. This is neither observable nor subject to extrapolation from anything we presently observe.  Our biological and psychological needs are shaped, in the long term, by many unpredictable environmental and technological changes. This cannot be predicted with any accuracy over millions of years into the future.  Our evolutionary descendants are just as likely to have developed a powerful aversion to sugar, for example, if it leads to health problems that threaten the survival of the species.

The time traveler theory illustrates how difficult it is to know or justifiably believe what human persons would know, desire, intend, or have the capacity to efficaciously execute in highly exotic hypothetical situations.  The reason for this is that we cannot assume that conventional aspects of ordinary life at present would obtain in these exotic situations.  We not only lack the relevant kinds of background knowledge, but testing procedures elude us.  In a similar way, empirical survival arguments require that we adopt assumptions about what living persons would know, desire, intend, and have the capacity to efficaciously execute in a highly exotic hypothetical situation: the persistence of consciousness after the death of the body.  That this consciousness would be personal, retain much of the knowledge, desires, and intentions that characterized its ante-mortem identity and phase of existence, be endowed with extremely potent powers of psychic functioning for efficaciously communicating with the living, and yet also be incredibly inept at either remembering basic facts concerning its ante-mortem existence or communicating such information to living persons—all of these assumptions are little more than untestable conjectures at this stage. 

4.  C.J. Ducasse’s “Plane-Crash Survivor” Argument

Even the better literature on survival has often been blind to the covert dependence on unsupported and untestable auxiliary hypotheses.  And this has given empirical survival arguments a deceptive appearance of cogency.  I’ll illustrate this with one of my favorite philosophical explorations of empirical survival arguments, C.J. Ducasse’s classic work A Critical Examination of the Belief in a Life after Death (1961).  

Unlike many other writers on survival, Ducasse was deeply conscious of and gave attention to the evidential criteria at work in assessments of the data allegedly suggestive of survival.  Hence, Ducasse asked, “What would prove, or make positively probable, that survival is a fact?” (1961: 199). Ducasse attempted to answer this question by exploring evidential criteria we would sensibly use to determine that someone had survived a plane crash.

Let us suppose that a friend of ours, John Doe, was a passenger on the transatlantic plane which some months ago the newspapers reported crashed shortly after leaving Shannon without having radioed that it was in trouble.  Since no survivors were reported to have been found, we would naturally assume that John Doe had died with the rest. (1961: 200) 

Ducasse went on to propose three situations in which we would acquire evidence that would convince us that John Doe had survived the crash. 

(1) We encounter a man on the street we recognize as John, he recognizes us, he has John Doe’s voice and mannerisms, and he is conversant about things that John Doe would have known, including information of a highly personal matter familiar to each of us.  

(2) Instead of physically encountering a man on the streets who resembles John Doe, we receive a phone call from a man who sounds like John Doe, and who freely exhibits the kind of first-personal perspective knowledge that would be characteristic of John Doe including private matters familiar to each of us. 

(3) We receive a phone call from someone who informs us that John Doe survived the crash and he wants us to know about his survival, but for some reason John Doe cannot come to the phone.  We’re told that John Doe is in need of money and wants us to deposit money into his bank account.  To acquire assurance that John Doe is indeed alive, we request through the intermediary, information of a sort freely disclosed in scenarios (1) and (2).  The intermediary provides us with the names of John Doe’s friends, various personal matters with which John Doe would be familiar, and we discern in the intermediary’s responses some of the peculiar features of John Doe’s thoughts and phraseology. 

Ducasse argues that in cases (1) and (2), we would take ourselves to have sufficient evidence to believe that John Doe had survived death.  He further argues that in case (3) we would be convinced of John Doe’s survival if we had robust evidence, that is, if we had no conclusive proof that John Doe had not survived death.  What I’m calling “robust evidence” captures Ducasse’s claim that the evidence would need to be abundant, sufficiently detailed, and of diverse kinds.  Ducasse essentially argues that we can imagine cases like (3) arising where we would be confident on the basis of the evidence that John Doe had survived death. 

Ducasse connects the discussion to survival in this manner.  He contends that the evidence for survival from mediumship duplicates the essential features of the evidence we could have that would convince us that John Doe survived the plane crash. 

Ducasse wrote:

This parallelism between the two situations [mediumship and plane-crash scenario (3)] entails that if reason rather than religious or materialistic faith is to decide, then our answer to the question whether the evidence we have does or does not establish survival (or at least a positive probability of it) must, in the manner of survival of death, be based on the very same considerations as in the matter of survival after a plane crash. That is, our answer will have to be based similarly on the quantity of evidence we get over the mediumistic “telephone;” on the quality of that evidence; and on the diversity of kinds of it we get. (1961: 203)

Ducasse goes on to argue that “the balance of the evidence” favors personal survival, and by this he appears to mean favors survival over various competitors (cf. 1961: 199). 

While Ducasse’s analogy is an interesting one, I think it commits a crucial mistake.  It’s true that in case (3) we would depend on the evidential factors Ducasse cites. However, the evidential force of these factors depends on a fourth factor not acknowledged by Ducasse: dependence on independently testable and/or supported auxiliary hypotheses.  And this is where the evidence for survival from mediumship is significantly different than the evidence we might have for a person having survived a plane crash.

If we return to Ducasse’s scenario (3), it must be acknowledged that we would evaluate the ostensible evidence in the light of various additional but independently plausible assumptions.  The list of specific assumptions would vary depending on various details of the scenario, but here are some illustrations of the kinds of assumptions that would plausibly be operative. 

a.  If a person survives a plane crash, we are positing the survival of their body.  Given that bodies have spatial location, the hypothesis that John Doe survived the plane crash yields the prediction that John Doe is located somewhere on earth.  Hence, we would be justified in supposing that if John Doe survived the plane crash, then his body would be spatially located somewhere on earth.

b. If we receive a phone call from an intermediary originating from Windsor, Connecticut, and the intermediary is allegedly relaying information to us from John who is present, it follows that John Doe is in Windsor, Connecticut.  This is an auxiliary hypothesis that can be independently tested.  Hence, we would be justified in supposing that if John Doe survived the plane crash and the intermediary is telling us the truth, then a physical body of a particular sort (matching the description of Joe Doe’s body, though perhaps missing a limb or two) would be presently located in Windsor, Connecticut.  This can be independently tested.  If John is there, others can in principle observe a body matching his there, etc.

c.  Based on our background knowledge, we already know that some people survive plane crashes, and we also know that the majority of plane-crash survivors have contacted, directly or indirectly, family members or friends, to let them know they are alive. Hence, this background knowledge provides an empirical basis for supposing that if John Doe survived the plane cash, then he would probably seek to communicate this information to family and friends.

d. Depending on the location and specifics of the crash, as well as background knowledge about other plane-crash survivors, we could be justified in supposing that if John Doe survived the plane crash, then we would probably receive communications from him during a relatively specific period of time.

e.  Regarding the communications, we assume a limited and very specific range of media through which John Doe would initiate communication with family or friends: phone, email, letters, or another human person as a messenger.   Hence, we antecedently know that if John Doe survived the plane crash and had the intention and power to communicate his survival to family and friends, then he would do so by means of specified media that fall within very narrow parameters.

f.  We assume that plane crashes are likely to produce varying degrees of trauma in survivors that affect memory and character, so we would expect communications to exhibit varying degrees of inconsistency and incoherence.  If John Doe survived the plane crash and successfully communicated with friends or family, then the content of the communications would be a mixture of detailed accuracy and significant inaccuracy.

Our assessment of the evidence for John Doe’s survival depends on the kinds of auxiliary hypotheses contained in (a) through (f).  These assumptions, though, are empirically testable, and indeed many of them are already independently supported by our background empirical knowledge.  If the auxiliary hypotheses lacked this quality, we could not sensibly take the evidence Ducasse cites as evidence for John Doe’s survival.  It’s only because we have warrant for the relevant auxiliary hypotheses that we can say what kinds of evidence we would expect to find if John Doe survived the plane crash.  Independent of the details of John Doe’s plane crash and the supposition of his survival, we have good reasons to suppose that there are plane-crash survivors, the majority of them have an interest to communicate, many will have the ability to do so, and we can say in advance the kinds of media they would use, etc.

Now when it comes to the hypothesis of postmortem survival, we’re simply not in a sufficiently similar epistemic situation.  That we can have evidence that someone has survived a plane crash depends crucially on what we already know, not just the quality and quantity of information deriving from the plane crash scenario.  We know about plane-crash survivors.  We can formulate predictions here because we have a stock of independently testable assumptions that tell us what we should expect to find in the way of evidence if our hypothesis is true.  None of this obtains in the case of possible postmortem survivors.  So the cogency of Ducasse’s case for postmortem survival from mediumship depends crucially on our ignoring a highly salient difference between plane-crash survivors and postmortem survivors. 

5.  Gertrude Schmeidler’s “Testable” Survival Hypothesis

Interestingly enough, the covert dependence of survival arguments on untestable auxiliary hypotheses is found even in literature that is conscious of the need to develop a survival hypothesis with testable predictions.  In the 1970s, parapsychologist Gertrude Schmeidler emphasized prediction as a crucial aspect of future survival research, and she also proposed a way of formulating a survival hypothesis open to such testability (Schmeilder 1977).  Other parapsychologist such as Bill Roll subsequently appropriated several of Schmeidler’s insights to further develop an allegedly “testable” survival hypothesis.

       Schmeidler wrote:

Suppose we try to test a hypothesis that makes three assumptions: (1) that there is survival of consciousness after bodily death; (2) that there is some continuity of personality, so that soon after death a surviving entity is recognizably similar to what the living person had been; and (3) that communication from the surviving entity is possible through a medium and in other ways. (1977: 5)

Schmeidler thought that we could rely on information collected from persons in their ante-mortem state as a basis for formulating predictions regarding which living persons should be expected to communicate (and which not) in their postmortem state, as well as the conditions under which postmortem communications from such people should be expected (and when not).

First, if prior to death person A has said that he has no intention under any circumstance to communicate with those still living after his death, then this provides the basis for a testable hypothesis: we should not expect any communications from A under any circumstances.  She adds a second point: “This immediately leads to a specific, testable subhypothesis. If mediums attempt to establish communication with the dead, evidence for such communications will be stronger for those who said while alive that they would want to communicate than for those who had said they would not” (1977: 5).  Finally, if prior to death another person B has said that he has an intention to communicate with those still living only under particular conditions, then this leads to another prediction: if person B survives death, then we should expect ostensible communications from B under the specific circumstances and not others.  As Schmeilder says: “The specific prediction would be that attempts to make contact with such persons after their death would shift between success and failure according to whether or not the conditions which had been stated were present” (1977: 5).

Schmeilder’s suggestion is interesting but nonetheless problematic.  One rather clear problem is that Schmeidler’s suggestion conceals rather than subjects to scrutiny highly questionable assumptions on which the efficacy of the suggested testing procedure depends.  Schmeidler’s project can’t get off the ground unless we make some crucial assumptions about what consciousness would be like if it were to survive death.

If person A informs us that he has no intention to communicate with the living under any circumstances after his death, predicting that we should not find communications ostensibly originating from this person depends on a strong assumption of continuity of conscious attitudes after death.   But there’s no reason to suppose this, at least not independent of the cases allegedly suggestive of personal survival.   Why not assume that the majority of survivors, having survived death, would be profoundly affected by their death and therefore differently motivated in their postmortem existence? Perhaps ante-mortem attitudes about what I would do if I survived death are very different than the purposes I would actually have if I did survive death.  After all, people change their purposes after relatively less extreme experiences in the course of their ante-mortem existence, sometimes over the course of a week.  Of course, the point here is not that we have good reason to suppose that any survivor would change his or her purposes concerning communicating with the living, only that we simply don’t know what would be the case with survivors.

The same holds with respect to Schmeidler’s claim that we should find more communications ostensibly originating from persons who expressed this interest while alive.  Again, why is this?  The operative assumption of substantial continuity of purposes and interests is highly questionable. It requires independent support.  The interest I express about communicating with the living after my death is an interest I now express as a person situated in a mundane though perhaps very exciting earthly existence.  This feature of my present psychology, contextualized as it is, may or may not persist if I survive death.  I don’t know, nor does anyone else.  It’s only by a subtle projection of our current psychology into the afterlife that we suppose we can know now what it will be like for us then.  But what is the empirical basis for this?

Carefully exploring Schmeilder’s proposal, then, at best forces the problem of auxiliary hypotheses to the surface.  It does not resolve the problem for the empirical survivalist.

6.  Concluding Remarks 

I have devoted three blogs now to discussing the role and implications of auxiliary hypotheses.  Let me briefly retrace the path.

In “Getting Sober about Survival I,” I outlined one of Elliott Sober’s interesting criticisms of empirical arguments for the existence of an intelligent designer based on features of organisms allegedly indicative of intelligent design.  His criticism focused on the untestable nature of necessary auxiliary hypotheses concerning the abilities and purposes of the postulated designer.  I began here since Sober’s criticism of intelligent design arguments seem particularly applicable to empirical arguments for postmortem survival.  They too depend on untestable auxiliary hypotheses, and this has important consequences for the assessment of the force of empirical data allegedly suggestive of survival. 

In “Getting Sober about Survival II,” I argued that empirical arguments for postmortem survival depend on various auxiliary hypotheses, in the absence of which the hypothesis of survival would have no empirical consequences.  The hypothesis of personal survival of death, like the hypothesis of an intelligent designer, is subject to an auxiliary hypothesis requirement.  I illustrated the point specifically in connection with survival arguments from the data of mediumship.  I sketched eight required auxiliary hypotheses concerning the purposes, powers, and knowledge of postmortem survivors, as well as assumptions concerning the nature of the process of ostensible communications from the deceased.

In the present blog—“Getting Sober about Survival III”—I have argued that, unlike paradigmatic empirical hypotheses, the auxiliary hypotheses involved in survival arguments are not at present independently testable, much less actually supported by our stock of empirical knowledge.  This generates what I’ve called the problem of auxiliary hypotheses.  This is a problem for empirical arguments in favor of the survival hypothesis, for in the absence of independently testable auxiliary hypotheses, we do not know what the hypothesis of survival should lead us to expect with respect to features of the empirical world.  There are, as I argued in the second blog of the present series, lots of different survival scenarios. Most of these scenarios, if true, would not lead us to expect any empirical evidence for survival, much less the data from mediumship and other ostensibly paranormal phenomena.  Only auxiliary hypotheses can produce a survival hypothesis that discriminates between survival scenarios with predictive consequences and those with none, and only a considerably robust version of such a hypothesis could possibly lead us to expect the actual data.  However, only independently testable auxiliary hypotheses can reasonably ground the desired assurance that our theory is a genuinely empirical one, rather than an attempt to accommodate or retrofit the data to our preferred metaphysical theories.

Where does this leave the empirical survivalist?

I’ve noted several times now that survival literature is plagued by what Sober called “lazy testing.”  Here one attempts to support one’s preferred hypothesis simply by showing that explanatory competitors fail.  That empirical survivalists are driven to this tactic is no coincidence.  It’s the strategic corollary of not being able to show the predictive consequences of one’s own theory.  So empirical survivalists routinely reinvent the wheel of misplaced criticisms of proposed non-survival explanations of the relevant data.  For example, some of the data of mediumship are allegedly surprising or improbable given living-agent psychic functioning, unless of course that hypothesis is stretched into a “super-psi” hypothesis for which there is no independent support.  But, of course, however low the probability of the data given some competing non-survival hypothesis, this does not suffice to show that the data are more probable given survival.  And as I’ve noted repeatedly, if empirical survivalists are epistemically entitled to adopt fantastic and far-reaching conjectures to account for the data, non-survivalists are permitted to do the same.  So-called “super-psi” hypotheses only appear incredible to empirical survivalists who remain unconscious about just how “super” their own survival hypothesis is.

My central claim—the probability of the relevant data (allegedly suggestive of survival) is actually inscrutable given the survival hypothesis, unless the survival hypothesis is stretched into a robust survival hypothesis.  However, since there is no independent support (or testability) for the auxiliary hypotheses that constitute a robust survival hypothesis, the probability of the data given survival remains inscrutable.  The alleged improbability of the data given competing hypotheses is a red herring that distracts from the core issue in the empirical survival debate: we do not know what the evidence for survival should look like, or whether there would be any evidence for survival, even if the survival hypothesis were true.

As explained in “Getting Sober about Survival I,” Likelihoodism is an approach to evidence assessment according to which evidence E favors or supports hypothesis H1 over H2 just if Pr(E / H1) > Pr(E / H2).  It’s not necessary that H1 confer a high probability on the evidence, only that the evidence is more probable (or less surprising) given H1 than H2.  In that case, though, if we don’t know what a hypothesis would lead us to expect in relation to evidence, we certainly won’t know whether the evidence is more probable given one hypothesis than another.  It follows, given Likelihoodism, that we won’t know whether the evidence actually favors the one hypothesis over its competitor. 

Applying this to the survival hypothesis, where S = the survival hypothesis, ψ = the hypothesis of living-agent psychic functioning, and DM = the data from mediumship, my contention is that we don’t know whether

Pr(DM / S) > Pr(DM / ψ) 

More generally, where C = any nearby competing non-survival hypothesis, we don’t know whether

Pr(DM / S ) > Pr (DM / C) 

It may appear that the situation changes once we add survivalist auxiliary hypotheses.  Where SR = a simple survival hypothesis amplified or augmented by the kinds of assumptions outlined above and discussed in the prior blog, we might suppose that there is at least justification for supposing that:

Pr(DM / SR ) > Pr (DM / C) 

Perhaps so, but until the problem of auxiliary hypotheses is addressed, the natural and plausible skeptical rejoinder is that:

Pr(DM / SR ) = Pr(DM / CR), where CR = the nearest competing hypothesis amplified or augmented with its own range of auxiliary hypotheses.   

If empirical survivalists wish to show that the evidence favors survival, they must adopt a robust survival hypothesis and compare it to a robust competitor.  Doing so, of course, only highlights the point that the resultant comparative Likelihoods, to the extent that they can be determined at all, float on an unstable ocean of conflicting conjectures.

Empirical survivalists have largely ignored rather than squarely faced the pivot of their whole project:  what changes, if any, are likely to happen to consciousness if it should survive death?  The inability to locate an empirically grounded answer to this question, for a single person or an indefinite number of them, floats on a larger sea of ignorance—our ignorance about the nature of consciousness itself.  The real question here is not whether we will survive death, but “what is the nature of consciousness itself?”  When the latter question has been answered, the question concerning postmortem survival will probably no longer be asked.

Michael Sudduth



Ducasse, Curt J. 1961. A Critical Examination of the Belief in a Life after Death. Springfield, IL: Charles C. Thomas.

Schmeidler, Gertrude. 1977. “Looking Ahead:  A Method for Research on Survival.”  Theta 5: 2–6.

Comments are closed.

Post Navigation