In my prior blog (1/28/14), I surveyed one of Elliott Sober’s objections to organismic design arguments—arguments that purport to show that physical features of organisms, features like complex adaptation, provide evidence for the existence of an intelligent designer. Sober argues that the hypothesis of intelligent design is an untestable hypothesis because it depends on untestable auxiliary hypotheses, that is, assumptions in addition to the hypothesis of an intelligent designer. Without these additional hypotheses the hypothesis of intelligent design would make no actual predictions. Sober contends that this prevents us from justifiably believing the comparative likelihood claim:
Pr(Observational evidence / Intelligent design ) > Pr(Observational evidence / Darwinian evolution),
that is, the probability of the relevant observational evidence given intelligent design is greater than the probability the same evidence given Darwinian evolution.
The range of possible auxiliary hypotheses that bear on the likelihood expressed on the left side of “>” produce varying results, from unity to zero. If we can’t independently test the auxiliary hypotheses, then we’re pretty much in the dark about what the likelihood of the design hypothesis is in relation to any competitor, including Darwinian evolution. According to a Likelihoodist account of the evidential favoring relation, evidence e favors hypothesis h1 over h2 just if Pr(e / h1) > Pr(e / h2). Hence, it follows from Sober’s analysis that we’re not justified in believing that the relevant evidence favors intelligent design over Darwinian evolution.
From Sober to Survival
My interest in Sober’s critique is its implications for assessing empirical arguments for survival based on the data of psychical research, for example, data drawn from the phenomenon of near-death experiences, living persons (mediums) who claim to receive and convey communications from the dead, and living persons who claim to remember past lives. During the past two years, I’ve drawn attention to how traditional empirical arguments for survival from such phenomena depend on various auxiliary hypotheses, for in the absence of such assumptions the survival hypothesis has no explanatory power. While empirical survivalists begrudgingly acknowledge their dependence on auxiliary hypotheses, at least if pressed on the point, the implications of this for the critical assessment of the empirical survival arguments has been ignored and therefore unexplored. On my view, satisfying the auxiliary hypothesis requirement generates various problems for empirical arguments for survival—what I’m designating “the problem of auxiliary hypotheses” (hereafter, PAH). In fact, I think PAH generates the most formidable challenge to empirical arguments for personal survival.
Succinctly stated, my view is that the survival hypothesis makes no significant predictions unless it’s supplemented with various auxiliary hypotheses, including a range of statements about what consciousness would be like if persons should survive death. This is problematic because most, if not all, of the required assumptions are not independently testable. More broadly speaking, they don’t carry the appropriate kind of “epistemic credentials,” especially if survival is supposed to be a scientific or quasi-scientific hypothesis. As Sober argued with reference to the intelligent design argument, I argue with reference to empirical arguments for survival: we’re actually not in a position to assess whether the survival hypothesis has a higher likelihood than rival hypotheses vis-à-vis the body of relevant data. This in turn undercuts the modest claim that the evidence favors the survival hypothesis over the competitors.
In the present blog I’m going to sketch the first phase of this argument, which involves showing that empirical arguments for survival depend on a range of auxiliary hypotheses. E.R. Dodds drew attention to this import fact in his 1934 article “Why I Do Not Believe in Survival.” As Dodds said, “the spiritualist [i.e., survival] hypothesis is hydraheaded. It is in fact not one hypothesis at all, but a series of hypotheses” (Dodds 1934: 170). Unfortunately, most of the survival literature since Dodds’s time has systematically suppressed this fact, together with its significant implications for the assessment of the empirical case for survival. Although I’m going to look specifically at survival arguments from the data of mediumship, my main argument can easily be extended to empirical arguments for survival from other kinds of ostensibly paranormal phenomena. In the next installment in the series, I’ll show how this dependence creates significant problems for empirical arguments for survival.
1. Why Likelihoods Matter
Sober’s critique of design arguments construes such arguments as Likelihood arguments. I suspect some empirical survivalists will question whether this is the optimal framework in which to formulate the empirical case for survival. So let me explain why likelihoods are essential to empirical arguments for survival.
First, as explained in my prior blog, “likelihoods” or the “likelihood of a hypothesis” refers to the probability the hypothesis, h, confers on observational evidence e, or the probability of the evidence given the hypothesis, formally represented as Pr(e / h). This is not the same as the probability of the hypothesis given the evidence (and background knowledge), formally represented as Pr(h / e & k). The former concerns how probable the hypothesis makes the evidence; the latter concerns how probable the evidence (together with background knowledge) makes the hypothesis. These probabilities may vary considerably.
So why should an empirical survivalist be concerned with the likelihood of the survival hypothesis?
First, the most common form of the empirical argument for survival depends on likelihoods. The argument is usually stated as a basic inference to best explanation. The central claim of such arguments is the survival hypothesis is the best explanation of some relevant range of data, where the data are drawn from phenomena such as near-death experiences, alleged past life memories and other features suggestive of reincarnation, and mediumistic communications. Why is the survival hypothesis the best explanation of the data collected from these phenomena? The reason is that it fits with or accounts for the data in a way superior to various competing hypotheses. However, upon examination these kinds of claims ultimately concern how well the survival hypothesis leads us to expect the relevant data. Otherwise stated, the survival hypothesis has predictive power. The data are supposed to be what we would expect if survival were true, and less so, perhaps considerably, given competing hypotheses. Some authors make it clear that the predictive power of the survival hypothesis is essential to its “testability.” (See Almeder 1996a, 1996b: 532–533; Becker 1993: 33; Berger 1987: 203; Gauld 1983: 73–75, 77 110; Hyslop 1919: 51, 330; Roll 2006: 167–170; Schmeidler 1977.)
Embedded in the basic explanatory survival argument, then, is a “likelihood” claim: the probability of the relevant data (D) is greater, perhaps much greater, given the survival hypothesis (S) than given the nearest competing hypothesis (C). More precisely, the arguments are concerned with the comparative likelihood of the survival hypothesis and its competitors. Formally stated: Pr(D / S) > Pr(D / C), or—more strongly—Pr(D / S) >> Pr(D / C). This also shows us that the basic explanatory argument can be restated without using the language of “explanation,” “explanatory power,” “explanatory virtue,” and so forth. The basic explanatory argument can simply be stated as a Likelihood argument. While this would not be sufficient to show that the evidence is strong enough to warrant rational acceptance of the survival hypothesis, it at least allows the survivalist to make the more modest claim that the relevant data evidentially favor or support the survival hypothesis over the nearest competitor(s).
Now there’s a strengthened form of the basic explanatory argument, favored by several philosophers, that adds considerations of prior probability to the inference to survival, largely because they aim for a verdict about the net plausibility of the survival hypothesis. In my recent interview with Jime Sayaka I discussed this Bayesian-style argument. It’s worth noting that likelihoods would still be important here because on the Bayesian view the posterior probability of a hypothesis depends in part on values assigned to likelihoods. However, in what follows I intend to focus only on a Likelihood version of the empirical argument for survival. One of the advantages of doing so is that we can bypass the thorny problem of prior probabilities so frequently introduced to defeat survival arguments. I’ll take up the implications of my argument for Bayesian-style survival arguments in the next blog.
So there’s considerable precedent for seeing likelihoods as essential features of evidence assessment in traditional empirical arguments for survival. This is not to say that empirical survivalists can’t propose new rules of evidence assessment. If they wish to propose criteria of evidence assessment that exclude likelihoods, they should do so. For the moment, though, I’m content to consider the implications of arguments that have actually been proposed.
Now likelihoods link a hypothesis to observational evidence by way of expressing the latter as predictive consequences of the former. It should be clear here that by “predictive power” I mean only the modest claim that a hypothesis leads us to expect the relevant data. The data need not be novel, nor need the data be logically entailed by the hypothesis. Moreover, it’s not essential to the argument about to unfold that the survival hypothesis has great predictive power. The focus is the more modest Likelihood claim that Pr(D / S) > Pr(D / C), or—more strongly—Pr(D / S) >> Pr(D / C). My question is not whether this is true. My interest is in exploring the logical and epistemic requirements for showing that it’s true. Are we adequately situated to reach a verdict here?
In the light of the discussion from the final section of my previous blog, it should be clear that to support the claim that Pr(D / S) > Pr(D / C) it will not do to argue that the likelihood on the right side is very low. If a five-year old boy suddenly exhibits remarkable talent as a percussionist, begins speaking with a British accent (though he was born in Memphis), expresses a desire for drinking large amounts of Jack Daniel’s, provides detailed information about the personal life of John Bonham (the drummer of the classic rock band Led Zeppelin), and claims to remember being John Bonham, this certainly seems very surprising given the hypothesis of living-agent psi. However, this fact does not establish that Pr(D / S) > Pr(D / C). The survival literature is replete with what Sober has aptly called “lazy testing”: declare one’s preferred hypothesis the winner by simply refuting competitors. What this “beat down” tactic does is merely evade the burden shared by the survivalist to show that the probability of the data is greater given the survival hypothesis than the competitor. What the survivalist needs to do is justify the likelihood claim by showing that the survival hypothesis would lead us to expect the data and would do so in a way superior to the competitor(s).
2. The Data of Mediumship as Evidence for Survival
Once we attempt to show that the survival hypothesis leads us to expect the data, it doesn’t take much reflection to see that the survival hypothesis leads us to expect absolutely nothing, unless it’s supplemented with various auxiliary hypotheses. To develop this in a way that’s as concrete as possible, let’s focus on some crucial kinds of data drawn from the phenomena of mental and trance mediumship, where deceased persons appear to communicate through living persons.
(m1) Some living person exhibits robust knowledge of facts concerning the public and private ante-mortem life of some particular and identifiable deceased person, where “robust knowledge” = knowledge that is specific in nature and ranges over many different facts about the ante-mortem life of the deceased.
This datum-type captures an essential feature of the data collected from the better cases of mediumship. The data from mediumship is prima facie suggestive of survival because the medium’s knowledge of the deceased is the sort of knowledge that the deceased would be in a privileged position to have. It must therefore consist of more than very general facts about the deceased, isolated and random bits of information, or information that is publicly accessible. It should be qualitatively strong by being as specific as possible and ranging over both the private and public life of the deceased. It should also be quantitatively strong by consisting of as much information as possible.
(m2) Multiple living persons independently exhibit knowledge of facts concerning the public and private life of some particular and identifiable deceased person, where the information from one source corroborates the information from another source, and the information collectively considered is robust and/or exhibits various structural coherence relations.
There are also cases in which no single medium exhibits robust knowledge of the deceased, but the information provided by multiple mediums is robust when collectively considered. There might be various points of corroboration between independent sources, and the body of information collectively considered may exhibit various kinds of coherence relations, e.g., by different strands of information being connected by inferential or explanatory relations.
(m3) Some living person exhibits knowledge of events or facts related to the private life of friends or family members of some particular and identifiable deceased person, where the events took place or facts obtained after the death of the deceased.
Ostensible “communicators” often provide information about the present goings on in the lives of family members and friends. This datum-type suggests, under a survival interpretation, the ongoing presence and involvement of the deceased in the lives of loved ones. This was a prominent feature in the mediumship of Mrs. Warren Elliott.
(m4) Some living person exhibits behavior, skills, or personality features similar or identical to those exhibited by some particular and identifiable deceased person during their ante-mortem life.
Trance mediumship is often regarded as impressive, not merely because of the accurate information conveyed by the medium, but also by the manner in which the information is conveyed, through convincing personations of the deceased. The medium employs mannerisms and turns of speech characteristic of the deceased, exhibits personality traits of the deceased, and various cognitive and linguistic skills, including the skill of identifying “by name” family or friends of the deceased who are present at a sitting, where these were characteristic of the deceased.
For ease of expression, I’ll use DM for the conjunction of m1, m2, m3, and m4. In accordance with Likelihoodism, we’re interested in comparative likelihoods, so we’re interested in the claim that Pr(DM / S) > Pr(DM / C), where C = some competing hypothesis. It won’t ultimately matter for the kind of criticism I’m going to make, but let’s just assume for the purposes of discussion and in the interest of concreteness that the competing hypothesis is the widely discussed appeal to psychic functioning (extra-sensory perception and psychokinesis) in living-persons, which I’ll appropriately designate ψ. Empirical arguments for survival will then be construed to make the modest claim that the evidence favors the survival hypothesis over the living-agent psi hypothesis. So from a Likelihoodist perspective, these arguments are committed to the claim that Pr(DM / S) > Pr(DM / ψ).
Now the standard survivalist argument in favor of this favorable likelihood goes roughly like this. If persons survive death, then DM isn’t all that surprising, but given ψ it is, for even if ψ should lead us to expect that living persons would possess some knowledge about the private and public life of deceased persons, say by telepathically mining this information from living friends and family members, ψ would not lead us to expect robust knowledge of the lives deceased persons. ψ might also lead us to expect that a medium might have knowledge about events taking place in the lives of friends and family of the deceased, but the joint occurrence of m1 and m3 seems very surprising given ψ. The ψ hypothesis would also leave unified streams of data from separate mediums very surprising. What we know about ψ doesn’t really lead us to expect this. And finally, while ψ may account for living persons knowing things about other people in non-conventional ways, ψ would not lead us to expect the personation data found in the better cases of mediumship. So, at the very least, S renders DM considerably less surprising than does ψ. So DM favors S over ψ.
While survivalists usually support the favorable likelihood by arguing how improbable DM is given ψ, I want to explore the assertion (rarely supported with argument) that the survival hypothesis would lead us to expect DM. It may seen intuitively obvious to many that this is so, but upon careful reflection this intuitive obviousness rests on the implicit acceptance of various assumptions beyond the simple idea of individual consciousness surviving death.
3. A Simple Survival Hypothesis has No Predictive Consequences
What kinds of considerations are needed to determine whether or not Pr(DM / S) > Pr(DM / ψ)? As already noted, not simply adducing reasons for supposing that Pr(DM / ψ) is low. We need reasons for supposing that, whatever the approximate value assigned to Pr(DM / ψ), the value assigned to Pr(DM / S) is greater. In other words, we need reasons for supposing that DM is more to be expected (less surprising) given S than given ψ. But in that case S must lead us to expect DM. Does it? This is going to depend on the content of S.
On the whole, I don’t find Antony Flew’s criticisms of empirical survival arguments all that compelling, or even interesting, but he was surely correct on this observation.
. . .until the concept “spirit” is made a great deal more specific than it is at present, the spirit account cannot serve as a scientific hypothesis. To use it as such we should have to be able to deduce from it definite and testable consequences. We should need to say that, if it were correct, such and such tests would yield such and such results. We cannot, because with spirits anything goes; nothing is definitely predictable. Or, to put it less misleadingly, the concept of spirit is hopelessly indeterminate. (Flew 1953/1973: 126)
Flew’s point would apply equally to alternative versions of the survival hypothesis that replace “spirit” with “a personal stream of consciousness with its memories of past earthly life” (Hyslop 1919: 53), “the continuation of conscious life” (Ducasse 1961: 11), or the postmortem persistence of a “non-physical subject of conscious states” (Lund 2009: 62, 83). A simple survival hypothesis—which posits the postmortem persistence of the self, the soul, the person, or even one’s individual consciousness—does not lead us to expect DM.
The point can be easily demonstrated. A priori there are various possible “survival scenarios,” each logically consistent with the simple supposition of survival as illustrated above. Here are five such possible scenarios.
S1: Some persons survive death (as discarnate souls), but in the absence of a functioning brain they do not exhibit any mental states or exert causal influence on our world.
S2: Some persons survive death as conscious beings, but have minimal memorial or character continuity with their ante-mortem existence.
S3: Some persons survive death as conscious beings, desire and intend to communicate, but they lack the ability to communicate.
S4: Some persons survive death as conscious beings, possess the ability to communicate, but they lack the desire and/or intention to communicate.
S5: Some persons survive death as conscious beings, but they lack the ability, desire, and intention to communicate.
In each of these survival scenarios, only some persons are postulated to survive death. Naturally, there are variations on these scenarios in which (a) everyone survives death and (b) S1 through S5 are scenarios indexed to different individual survivors so that there would be a distribution of varying powers, desires, intentions, degrees of knowledge and memory, etc. over the range of various survivors. Perhaps one person’s survival scenario is S2, and another person’s survival scenario is S4. There are also many possible survival scenarios that can be constructed from the above five, for example by conjoining S2 and S3 or S2 and S4, but there’s no need to explore these possibilities to appreciate the central point here. The above scenarios are unfavorable to likelihoods for the survival hypothesis, for though they are compatible with the simple survival hypothesis, they deflate the likelihood of the survival hypothesis. If we accept any of these five survival scenarios, the data are not what we would expect. In fact, on some scenarios, the likelihood of the survival hypothesis would be zero; for example, if survivors did not have the ability to communicate with the living, the probability of DM would be zero. If we assumed S2, then the probability of m1 or m2 would be zero (or close to zero) because we wouldn’t expect deceased persons to communicate robust knowledge of their lives if they don’t actually retain such knowledge in the afterlife.
To illustrate further, it’s not difficult to construct alternative hypotheses with higher likelihoods than any of the above survival hypotheses. Consider the following alternative hypothesis:
(DH) There is some demonic entity, with significant power and detailed knowledge of the lives of formerly living persons, and who wishes to masquerade as deceased persons for the purpose of engaging in deception.
Survivalists are likely to jeer at (DH), though Evangelical Christians find it perfectly sensible. In both cases, prior probabilities are influencing judgments. But we’re not interested in prior probabilities, only likelihoods. And it’s quite evident that the probability of DM is considerably greater given (DH) than given any of the survival hypotheses above. The same conclusion follows if the competitor is the living-agent psi hypothesis (ψ). Perhaps Pr(DM / ψ) is not very high, but it’s not plausible to suppose that Pr(DM / S2) > Pr(DM / ψ), and so forth. This is also consistent with the prior probability of the ψ hypothesis being very low. The point is that the likelihood of ψ is not lower than the likelihood of the survival hypothesis, if the latter is understood in any of the five ways above.
The problem here can be simply stated: a simple survival hypothesis does not discriminate between survival scenarios that would lead us to expect the data and those that don’t, or more radically that confer a probability of zero on the data. There simply are no predictive consequences for a survival hypothesis that might fall into the logical space of any of the scenarios above. Merely postulating “survival of the self” or “survival of consciousness” just doesn’t tell us enough.
4. Minimally Required Auxiliary Hypotheses
The survival hypotheses sketched above confer low or zero probabilities on the data because they are more robust than the simple supposition of survival. This is crucial. Recall that in connection with his critique of design arguments Sober noted the Duhem-Quine thesis that single hypotheses rarely have (deductive or probabilistic) predictive consequences, unless auxiliary statements are introduced. Hence, we can only test hypotheses (against their predictive consequences) by embedding them in sets of statements that jointly have predictive consequences. Hence, predictive derivations depend on incorporating auxiliary hypotheses. The five survival scenarios above incorporate auxiliary hypotheses that result in predictive consequences that do not fit the actual data. But herein we find the recipe for the survivalist who wishes to argue that the survival hypothesis has predictive consequences that fit the data. He needs a robust survival hypothesis with favorable as opposed to unfavorable predictive consequences.
Given the observations in the prior section above, a robust survival hypothesis with favorable predictive consequences must minimally discriminate between survival scenarios that lead us to expect the data and those that do not. This implies that a robust survival hypothesis with favorable predictive consequences vis-à-vis DM must be incompatible with the survival scenarios above. This can provide a starting point for exploring just what an empirical survivalist must assume for the survival hypothesis to lead us to expect the data. With respect to the data of mediumship sketched above, the assumptions are 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.
[A1] affirms that some survivors would be conscious in the absence of a physical body, thereby ruling out survival scenario S1 above. The next two assumptions concern what consciousness would be like for at least some deceased persons. [A2] concerns the degree of self-knowledge the deceased would have, thereby ruling out survival scenario S2. [A3] concerns survivors having persisting, though perhaps intermittent, knowledge of states of affair in the world of living persons, thereby ruling out survival scenario S3, inasmuch as the ability to communicate depends on survivors knowing what is happening in the world of the subjects with whom they communicate. Just as [A1] is not entailed by positing surviving persons or selves, neither [A2] nor [A3] is entailed by positing the persistence of consciousness in a discarnate state. These are independent conditions. Furthermore, [A4] tells us what some deceased persons would want to do, and [A5] tells us that they would be able to efficaciously bring about their purposes. [A4] rules out survival scenario S4, and [A5] rules out survival scenario S3. [A4] and [A5] jointly rule out survival scenario S5.
[A1]–[A5] are auxiliary hypotheses that rule out survival scenarios that prevent a favorable likelihood for the survival hypothesis, and they lead us to expect that there should be evidence of postmortem communications with content suggestive of the identity of the communicator. So these are minimally necessary.
5. Additional Auxiliary Hypotheses
However, further assumptions are plausibly required. For example, philosophers and parapsychologists have generally acknowledged that, inasmuch as survivors are discarnate persons, a survivor’s epistemic access to the world would need to be a potent form of extra-sensory perception (e.g., telepathy, clairvoyance) and a survivor’s causal influence over the world would need to be a potent form of psychokinesis. Since discarnate persons are ex hypothesi without physical bodies, their modes of knowing and causal interaction would have to be direct or wholly unmediated by a body or cognitive system associated with a body. So for any discarnate survivor [A3] and [A5] will logically entail a more specific assumption about the cognitive and causal powers of the deceased, namely
[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.
But further assumptions are needed. (m4) refers to those strands of data from some instances of trance mediumship in which “communicators” (via a medium) exhibit behavior, personality traits, or skills characteristic of the deceased. For example, the medium might speak with a particular tone, accent, use particular words or phrases, or physical gestures, where these were characteristic of the deceased. The medium might even speak with words or phrases in a language foreign to the medium but native to the deceased. Communicators also are able not only to provide the names of family and friends but are able to pick them out from among sitters. This implies not merely knowledge that people are present at the sitting (acquired through telepathy or clairvoyance), but arguably the skill of identifying persons present “by name” as former family members and friends.
[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.
I regard survivors in [A7] as a subset of the larger group of communicating survivors because if the data of mediumship extends to non-trance mental mediumship, the data might not include indications that the communicator is continuous with his ante-mortem life in the manner specified in [A7].
But there’s more to consider here. As Hornell Hart explained in his classic Enigma of Survival (1959), when we explore the wider context in which strands of data such as (m1) and (m2) are embedded, we find that there are relevant data of a different sort. For example:
(m5) In “trance mediumship” communicators are often unable to provide basic information about their lives requested by sitters, or give inconsistent and incorrect information about their own lives, or they are otherwise mistaken about matters we would expect them to know (at least given A2 or A3).
(m6) Ostensible “communicators” in trance mediumship often lack various cognitive, linguistic, and other skills that characterized the formerly living person they claim to be.
The total evidence requirement for inductive reasoning implies that we must include all relevant evidence, and so data captured by (m5) and (m6) must be included within the total evidence set. The problem should be apparent. While [A2] leads us to expect postmortem communications to exhibit the kind of knowledge we would use to identify persons in our present experience, (m5) tell us not to expect a consistent display of such knowledge. So some further assumption are needed to bring the survival hypothesis into an optimal fit with the total relevant evidence.
Survivalists have proposed a few different hypotheses at this juncture. Drayton Thomas said that the locus of the problem was in the communicator who, during communication with living persons, experiences diminished causal power and a temporary weakening of cognition, including memory (Hart 1959: 87-88, 106; cf. Braude 2003: 66).
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.
Alternatively, we might suppose that the locus of the problem is not in the communicator but in the medium (Braude 2003: 54-55, 66-67). In mental mediumship we might suppose that information originating from the deceased has been filtered, interpreted, or otherwise altered by the medium’s own mind by the time it teaches her consciousness, especially if the information passes through or is influenced by medium’s unconscious mind. In trance mediumship, the communicators may be “virtual survivors,” a joint product of the medium’s own unconscious construction with information originating from the actual deceased person.
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.
A third possibility concerns the method of communication. Perhaps direct control of the medium’s body is a more reliable method of communication than telepathic interaction, or vice-versa.
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.
So it looks like the survival hypothesis would need to assume at least the disjunction of each of these possibilities, that is, the case for survival would need to assume:
[A8] Either C1, C2, or C3.
I’ll refer to the conjunction of [A1]-[A8] as A*. The conjunction of A* and the simple hypothesis of survival S is a robust survival hypothesis, but—unlike S1, S2, etc—it’s a robust survival hypothesis that is favorable to the likelihood of survival. We can now say that the simple survival hypothesis S + A* prevents the survival hypothesis from having a likelihood of zero vis-à-vis the relevant data. We can also say that this robust survival hypothesis leads us to expect at least the following five very general data:
(i) There will be features of the empirical world suggestive of post-mortem communications originating from some formerly living persons.
(ii) The content of the communications will include specific and detailed information about the ante-mortem life of some particular deceased person.
(iii) The content of the communications will include information about postmortem happenings in the life of friends and family members of the deceased.
(iv) The content of the communications will have indications of the beliefs, purposes, and personality traits of the deceased.
(v) The content of the communications will not be fully accurate or consistent.
It’s worth noting, though I’ll not develop the significance of it here, that S + A* does not lead us to expect anything regarding other features of the data, for example, the mode or manner of communications, when or where communications will take place, or which deceased persons will communicate. S + A* does not lead us to expect any general patterns with respect to these features of the relevant data, nor any specific datum within the domain of relevant data. It only leads us to expect some of the general features entailed by the specific data adduced as evidence for survival.
6. Second Phase of Argument: A Preview
In his 1934 paper “Why I Don’t Believe in Survival,” E.R. Dodds noted a number of the assumptions I’ve drawn attention to above. Dodds argued that on account of such auxiliary assumptions the survival hypothesis is a more complex hypothesis than survivalists acknowledge. That’s no doubt true. And it’s especially relevant since many survivalists argue that the survival hypothesis is simpler than appeals to living-agent psi. The shortcoming of all such arguments is that they consider the survival hypothesis in its simple form, and compare it to robust versions of competitors (like the appeal to living-agent psi). More generally stated, the auxiliary hypotheses (required for the survival hypothesis to have minimal predictive power) lower the prior probability of the survival hypothesis. To the extent that prior probability counts in evidence assessment, this will be significant.
However, I’ve been looking at survival likelihoods, and likelihoods are blind to prior probabilities. So I wish to make a different kind of criticism. My criticism is that since we can’t independently test the auxiliary hypotheses required by the survival hypothesis (to yield favorable likelihoods), we’re not in a position to say whether or not Pr(DM / S) > Pr(DM / C) and hence whether Pr(DM / S) > Pr(DM / ψ). After all, if we have no reason independent of S, C, or DM to accept the auxiliary hypotheses, we have no independent reason to prefer them to any number of other auxiliary hypotheses that result in survival likelihoods of zero. Hence, the predictive consequences of survival are inscrutable and so we can’t say that DM is evidence that favors the survival hypothesis over various competitors. I’ll sketch this second phase of the argument in my next blog.
Almeder, Robert. 1996a. “Recent Responses to Survival Research.” Journal of Scientific Exploration 10: 495–517.
Almeder, Robert. 1996b. “Almeder’s Reply to Wheatley and Braude.” Journal of Scientific Exploration 10: 529–533.
Becker, Carl. 1993. Paranormal Experience and Survival of Death. Albany: State University of New York Press.
Berger, Arthur S. 1987. “A Critical Outline of the Prima Facie Evidence for Survival.” In Death and Immortality in the Religions of the World, ed. Badham, Paul, and Badham, Linda. New York: Paragon House, 188–213.
Braude, Stephen. 2003. Immortal Remains. Lanham, MD: Rowman and Littlefield Publishers, Inc.
Dodds, Eric R. 1934. “Why I Do Not Believe in Survival.” Proceedings of the Society of Psychical Research, XLII, part 133, 147–172.
Ducasse, Curt J. 1961. A Critical Examination of the Belief in a Life after Death. Springfield, IL: Charles C. Thomas.
Flew, Antony. 1953/1973. “The Question of Survival.” In Immortality, ed. Terence Penelhum. Belmont, CA: Wadsworth Publishing Company.
Gauld, Alan. 1983. Mediumship and Survival: A Century of Investigations. London: Paladin Books.
Hart, Hornell. 1959. The Enigma of Survival. London: Rider and Co.
Hylsop, James. 1919. Contact with the Other World: the Latest Evidence as to Communication with the Dead. New York: the Century Co.
Lund, David. 2009. Persons, Souls and Death: A Philosophical Investigation of an Afterlife. Jefferson, NC: McFarland.
Roll, William. 2006. “On Apparitions and Mediumship: An Examination of the Evidence that Personal Consciousness Persists after Death.” In The Survival of Human Consciousness: Essays on the Possibility of Life after Death, ed. Lance Storm and Michael Thalbourne. Jefferson, NC: McFarland, 142–173.
Schmeidler, Gertrude. 1977. “Looking Ahead: A Method for Research on Survival.” Theta 5: 2–6.