In January 2013 Jime Sayaka interviewed me on the topic of postmortem survival for his now defunct blog Subversive Thinking. In what turned out to be a lengthy interview (and preview of arguments in my forthcoming book), I outlined in considerable detail my critique of empirical arguments for survival, as well as explained why common survivalist defenses of these arguments lack cogency. Below I repost my responses to questions #5 through #8. Question #5 concerns Chris Carter’s contention that the survival hypothesis is the most natural inference from the relevant data. Question #6 concerns the strength of counter-explanations of the data in terms of psychic functioning among living persons. Question #7 concerns Chris Carter’s “silver bullet” objection to appeals to living-agent psi to explain the relevant data. Question #8 addresses the alleged ad hoc nature of appeals to living-agent psi as a counter-explanation. Note that in my responses I rely on symbolism used in confirmation theory to provide a formal account of various logical relations between evidence and hypotheses.
5 – Sayaka: Survivalists like Chris Carter and others suggest that survival of consciousness is the most natural, obvious and straightforward inference from the empirical data from mediumship, near-death experiences and reincarnation type cases. What do you think of this argument?
Sudduth: I’m not inclined to dispute the claim. I think the claim is entirely compatible with my central thesis and the premises of my central argument. Many theists say that the existence of God is the most natural, obvious, and straightforward inference from the fine-tuning of the universe. And it is . . . to them. I’m quite sure that for Carter and many other survivalists the survival inference is natural, obvious, and straightforward. However, as in the case of proposed theistic explanations of the existence and regularities of the universe, the obviousness of the inference lies in the (often unspoken and unconscious) adoption of a whole array of background assumptions. As a philosopher, I’m interested in identifying these assumptions and assessing their role in the inference to survival, and this is in the interest of ultimately evaluating the cogency and strength of survival arguments. That the survival inference is natural, obvious, and straightforward to lots of people is a psychological truth that really isn’t relevant to the kind of question that is central in the empirical survival debate.
6 – Sayaka: You have argued that the super-PSI explanation of the data is adequate, if not most adequate, than the survival hypothesis. Can you explain briefly the super-PSI hypothesis and why is it so good as an alternative explanation for the data?
Sudduth: I don’t believe I’ve argued that the super-psi explanation is adequate, much less most adequate or good. In fact, I’ve explicitly stated in a few publications now that we should dispense with talk of “super-psi” altogether and simply utilize the language of “living-agent psi,” with the further caveat that such a hypothesis may appear in more or less robust forms depending on the range of auxiliary hypotheses added to it. My view is that appeals to robust living-agent psi hypotheses are no less adequate or no less plausible than the survival hypothesis, at least when these hypotheses are compared in their robust forms and we’re considering a maximal data set, not just narrow strands of data. It’s quite another matter to say that either explanation is adequate, much less good.
I suppose I should say something here about strategies for critiquing arguments, as there seems to be confusion in some of the literature as to what it takes to defeat survival arguments. If the argument for survival depends on the premise that the survival hypothesis is the best explanation of the data, to defeat the argument I only need to show that the survivalist is not justified in asserting the premise. One way to accomplish this is to show that the premise is false, to show that survival is not the best explanation of the data. Of course, to do this it’s not necessary to show that there is some rival hypothesis that better explains the data. It would suffice to show that there is some rival hypothesis that is at least as good as the survival hypothesis in leading us to expect the data. However, another way to show that the survivalist is not justified in claiming superior explanatory power on behalf of the survival hypothesis is simply to show that the survivalist is not justified in supposing that this premise is true, which is different from showing that the premise is false. There are defeaters that constitute overriding reasons for supposing that a statement is false (rebutting defeaters) and there are defeaters that remove or otherwise neutralize reasons for supposing that a statement is true (undercutting defeaters). This distinction is frequently lost sight of in the debate.
I maintain that empirical survivalists are not justified in claiming that the survival hypothesis is the best explanation of the data. More precisely stated, I maintain that empirical survivalists are not justified in claiming that the survival hypothesis leads us to expect data that are otherwise unlikely, or even less likely given the nearest robust competitors. Now I do believe that there are reasons for supposing that there are nearby explanatory competitors that are at least as adequate at survival, or no less adequate if you will. My position involves a parity thesis, and the argument is a parity argument. And this is one way to show that the survival hypothesis is not the best explanation of the data. However, I also maintain, more modestly, that survivalists have not presented good enough reasons for supposing that the survival hypothesis is the superior explanation of the data. If we’re comparing robust versions of the survival hypothesis and living-agent psi hypothesis, then I don’t think survivalists have effectively argued that the data are more to be expected given robust survival than given the nearest robust competitors, for example something like Stephen Braude’s motivated living-agent psi hypothesis.
But let me give a more technical elaboration here. Let C = the nearest robust competitor, S = robust survival hypothesis, and DMAX = a maximal data set. In that case, I argue:
(1) Survivalists have not presented good enough reasons to believe that Pr(DMAX/S) > Pr(DMAX/C), much less that Pr(DMAX/S) >> Pr(DMAX/C).
[Editorial Comment: Pr(DMAX/S) means “the probability of the maximal data set given the survival hypothesis.” Hence, the whole expression states “the probability of the maximal data set given the survival hypothesis is greater than the probability of the maximal data set given the nearest robust competitor. Since “>>” means much greater, the second expression is a stronger one.]
(2) There are overriding reasons for supposing that Pr(DMAX/S) ≤ Pr(DMAX/C).
[Editorial Comment: this expression says “the probability of the maximal data set given the survival hypothesis is less than or equal to the probability of the maximal data set given the nearest robust competitor.]
To be clear, we are here concerned with a comparative probability of the data given each of the competing hypotheses. This is the posterior probability of the evidence [Pr(e/h)], not to be confused with the posterior probability of the hypothesis [Pr(h/e)]. Following the common practice in confirmation theory I’ll refer to such posterior probabilities as “likelihoods,” and by extension the “likelihood of a hypothesis” will refer to the extent to which a hypothesis renders the evidence or data probable. (The likelihood of a hypothesis is distinct from the probability of a hypothesis, as the latter refers to the extent to which the evidence renders the hypothesis probable). So my view with respect to the living-agent psi hypothesis is that I don’t think survivalists have really shown that the survival hypothesis has a likelihood superior to a sufficiently robust living-agent psi hypothesis, at least not if the data set has sufficiently broad parameters. More strongly stated, my view is that the likelihood of the survival hypothesis is less than or equal to the likelihood of the nearest robust competitor.
It’s important to underscore here that the argument for supposing that Pr(DMAX/S) ≤ Pr(DMAX/C) does not require the stronger claim that Pr(DMAX/C) > Pr(DMAX/S). My position is also compatible with the following survivalist claim: Pr(DMIN/S) > Pr(DMIN/C), where DMIN = a more restricted data set. So I work out my position in a way that is actually sensitive to the evidence-parameters problem. Nonetheless, as I see it, (1) and (2) severally suffice to defeat the empirical argument for survival, at least to the extent to which the empirical argument depends on attributing to the survival hypothesis a superior likelihood over competitors. So this will apply to Bayesian survival arguments that make use of likelihoods for the purposes of showing that the survival hypothesis is more probable than not. It will also apply to Likelihoodist versions of the empirical argument for survival that are more modest in their pretensions, aiming only to show that the evidence (strongly) favors the survival hypothesis over the competitors solely on the grounds that the survival hypothesis has a superior likelihood.
It should be clear that the kind of comparative “adequacy” I’ve been focusing on here concerns “likelihoods” but of course many survivalists regard counter-explanations, such as the robust versions of the living-agent psi hypothesis, as (comparatively) inadequate for reasons other than those that bear on likelihoods. For instance, many survivalists reject robust versions of the living-agent psi hypothesis because of its lack of independent testability and increased complexity. Since I regard these issues as determinants of prior probability (rather than explanatory power), I would parse the frequently encountered survivalist objection as maintaining that robust living-agent psi hypotheses have a lower prior probability than the survival hypothesis. So the survivalist would presumably be claiming that Pr(S/K) > Pr(C/K) because C is more complex than S, fits less well with background knowledge, and we have no independent evidence for C (or some auxiliary contained in C). Of course, on my analysis of priors, I think that Pr(S/K) ≤ Pr(C/K), at least if S and C refer to robust versions of survival and the nearest competitor and the background knowledge is what interlocutors in the debate typically include, e.g., scientific knowledge.
7 – Sayaka: Chris Carter has argued forcefully against the super-PSI hypothesis (or super-ESP, as some calls it). For example, he says “Evidence for the existence of ESP of the required power and range is practically nonexistent. Defenders of the super-ESP hypothesis are hard-pressed to find any such examples – outside of cases of apparent communication from the deceased.” According to Carter, no defender of super-PSI has ever been able to challenge this objection. What do you think of this objection?
Sudduth: It’s the stock in trade of empirical survivalists to reject appeals to super-psi on the grounds that this hypothesis lacks “independent support.” Stephen Braude has challenged this objection for a number of years, and I present an argument against it in a forthcoming paper in The Survival Hypothesis: Essays on Mediumship, ed. Adam Rock (McFarland, 2014). An earlier draft of the paper in question, “Is Survival the Best Explanation of Mediumship?”, is available on my professional website michaelsudduth.com. Let me outline some of the salient points that I think significantly weaken the force of this objection.
First, from a Likelihoodist approach to confirmation theory, whether evidence favors hypothesis h1 over h2 depends solely on whether e is more to be expected given h1 than given h2, technically stated, whether Pr(e/h1) > Pr(e/h2). A student walking down the hall from the Philosophy Department with three philosophy books in his hand favors the hypothesis that the student is a philosophy major over the hypothesis that the student is a biology major because the observational evidence is more likely given the former hypothesis than given the latter hypothesis. Whether there is independent support for either hypothesis is not relevant to deciding which hypothesis the evidence favors, confirms, or supports. Now of course, the Likelihoodist approach doesn’t tell us which hypothesis is likely to be true, and therefore it doesn’t tell us which hypothesis to accept or believe. It only tells us which of two or more hypotheses a body of evidence favors or supports. But the point here is that if I’m a Likelihoodist, I can make sense of the relevant data favoring the super-psi hypothesis over the survival hypothesis, even if super-psi lacks independent support.
Now the apparent shortcoming of my proposed Likelihoodist defense of the super-psi hypothesis is that lack of independent support may nonetheless be salient to our overall assessment of a hypothesis, and if we want to compare the survival hypothesis and its competitors, we might want to inquire about more than their comparative likelihoods. For example, the hypothesis that a very powerful demon intended me to pick the ace of spades has a higher likelihood than the hypothesis that my selection of the card was random, for the former hypothesis makes the selection of the card very probable and the latter makes it very improbable. But the fact that the evidence favors the demon hypothesis here does not make the hypothesis very probable all things considered, and the crucial issue here, if we don’t have good evidence against the existence of such an entity, is quite plausibly that the demon hypothesis lacks independent support. More generally stated, the demon hypothesis has a very low prior probability, and this is due in large part to the fact that it lacks independent support.
Now this point is significant from the vantage point of a possible defense of the empirical argument for survival. Let’s suppose that Pr(DMAX/S&K) = Pr(DMAX/C&K). That is, the predictive power or likelihoods of S and C are equivalent. The survival hypothesis might still have a greater posterior probability than C (maybe even be more probable than not) if its prior probability is greater, especially if the prior probability is much greater. From a Bayesian viewpoint, if Pr(e/h1&k) = Pr(e/h2&k), then Pr(h1/e&k) > Pr(h2/e&k) just if Pr(h1/k) > Pr(h2/k). That is to say, if two hypotheses have equal predictive power (or likelihoods), then the evidence and background knowledge confers a greater probability on h1 than h2 just if h1’s prior probability is greater than h2’s prior probability. So a survivalist might simply argue that, worst case scenario, Pr(DMAX/S&K) = Pr(DMAX/C&K), but since Pr(S/K) >> Pr(C/K), the survival hypothesis has a greater posterior probability, maybe it’s still more probable than not. To put this otherwise, a survivalist might argue that the net effect of deflating the explanatory power of the survival hypothesis on the grounds of co-equal likelihoods is negligible since the prior probability of the survival hypothesis is much greater.
I think this counter-argument would work if we were comparing the priors of “C” and a simple survival hypothesis, but as I’ve already argued, the explanatory candidates must be compared in their robust forms because simple survival has no explanatory power. If the survivalist tries to shift to a simple survival hypothesis to inflate the prior probability of the survival hypothesis, this will deflate the explanatory power of the survival hypothesis. It will follow that Pr(DMAX/C&K) >> Pr(DMAX/S&K). But, unfortunately for the survivalist, if “lack of independent support” drives down the prior probability of the appeal to so-called super-psi, it will also drive down the prior probability of the robust survival hypothesis since it also depends on a broad range of auxiliary hypotheses for which there is no independent support. More generally, if “lack of independent support” is a defect of robust living-agent psi hypotheses, it will also be a defect of the robust living agent psi hypothesis. So there’s no advantage to be had here for the survival hypothesis. As I noted above, on my analysis, Pr(S/K) ≤ Pr(C/K), if “S” refers to a robust survival hypothesis.
Finally, the problem for the survivalist is exacerbated since the auxiliary assumptions required by the survival hypothesis (to have predictive power) includes an auxiliary hypothesis that attributes super-psi to discarnate persons (and possibly also living agents). As Gauld, Braude, Emily Williams Kelly, and I have each argued, the survival hypothesis itself is committed to the existence of ESP of a required power and range examples of which survivalists would be hard pressed to find outside cases of apparent communications from the deceased. As I argued my 2009 paper “Super-Psi and the Survivalist Interpretation of Mediumship,” if the communications attributed to the deceased in paradigmatic cases of mediumship are really from the deceased, they too have extraordinary powers of knowledge acquisition, often requiring that they telepathically or clairvoyantly mine information from multiple sources. It’s only the unwarranted assumption that death increases the potency of psi, or some such other speculative assumption, that allows survivalists to think that they are immune from this objection to super-psi. But of course, they’re merely taking refuge in a further assumption for which there is no independent evidence.
But there’s another part of Carter’s objection of which I’m suspicious, namely the demand to produce examples of ESP of the “required” power and range outside cases of survival.
First, why is there a requirement that psi be super-psi in order to deflate the explanatory superiority of the survival hypothesis? Empirical survivalists routinely assert this, but Braude has shown that the assertion rests on various implausible assumptions. Moreover, I’ve discussed in detail why appeals to living-agent psi challenge the survival hypothesis without requiring an appeal to super-psi. See my the previously mentioned forthcoming “Is Survival the Best Explanation of Mediumship?” and my “A Critical Response to David Lund’s Argument for Postmortem Survival” (Journal of Scientific Exploration, 2013, 27: 277-316).
Second, I don’t know what kind of evidence would count as clear evidence for super-psi but not be capable of being construed as evidence for survival by empirical survivalists. You may recall that back in the 1970s the Philip Group produced ostensible living-agent psychokinetic effects that resembled the phenomena of physical mediumship, complete with messages from a “deceased personality” named Philip. Philip was a fictional person created by the group of experimenters, and his ostensible communications through raps and knocks corresponded to the details of the fictional biography created by members of the group. Yet David Fontana gave the Philip Group phenomena a survivalist interpretation by positing an earthbound spirit intent on fooling the group by masquerading as their fictional character Philip. (See Fontana, Is There An Afterlife? 2005, p. 112). Well, of course. If there were some earthbound spirit with such an intention and the power to carry out his deception, we would expect to find the evidence associated with the Philip Group experiments. By parity of reasoning, the hypothesis that a malicious and powerful demon wanted me to select the ace of spades I drew from the deck of cards renders my draw quite probable, certainly more probable than the alternative hypothesis that my draw was completely random. You see, you can select any datum and adopt a hypothesis that renders the datum very probable or more probable than it would be given competing hypotheses. The difficulty in meeting the survivalist challenge to produce evidence for super-psi outside cases of survival may not be the absence of such evidence, but the survivalist proclivity to see such evidence where it arises as evidence for survival. Since what counts as a case of survival is precisely what’s in dispute by the parties in the debate, the challenge begs the question.
And of course the previous point highlights the final problem with Carter’s objection. Although it’s not clear what would count as unambiguous evidence for living-agent super-psi (vs. survival), what is clear is that no empirical survivalist has met the challenge to provide independent support for the dozen or so required auxiliary hypotheses required for survival to have predictive efficacy. And this request does not beg the question. It’s simply another instance of the general requirement imposed by Carter himself with respect to the super-psi hypothesis. What Carter and other survivalists who take his position need to do is (i) explicitly acknowledge the content or range of the assumptions required for survival to yield likelihoods (of the evidence) that exceed the likelihoods (of the evidence) given rival hypotheses and (ii) provide independent support for as many of these auxiliary hypotheses as they can. Until this can be done, the empirical case for survival has not been worked out with adequate logical rigor, and it certainly does not deserve to be considered a genuine scientific or even quasi-scientific hypothesis.
8 – Sayaka: In connection with the above objection, survivalists suggest that the super-PSI hypothesis is ad hoc, because of the lack of any independent evidence for super-PSI, besides the putative cases of survival. (It’s like arguing that the reincarnation type cases are best explained by extraterrestials implanting false memories, without having any independent evidence for the existence of aliens, a point pressed by philosopher Robert Almeder in his response to atheist philosopher Steven Hales). Some survivalist consider this to be the most crushing objection against super-PSI. What’s your reply?
Sudduth: Given what I have argued above, if this objection is a crushing objection against the super-psi hypothesis, it’s also a crushing objection to the survival hypothesis, in which case the survivalist is hoisted by his own petard. As I’ve already noted, the simple supposition of survival makes no specific predictions, much less does it predict any of the fine-grained features of the actual data, unless it’s supplemented by auxiliary assumptions of a wide-ranging sort. Hence, the lack of independent support objection is just as applicable to the robust survival hypothesis as it is to the super-psi hypothesis. Even if we assume that there is independent evidence for survival, there would also have to be independent evidence for the range of auxiliary assumptions needed for the survival hypothesis to have predictive power. Almeder has not provided this independent support, nor have other survivalists.
So why aren’t the auxiliary hypotheses employed by the survivalist ad hoc in nature? I noted above that among such auxiliary hypotheses would be the attribution of super-psi to discarnate persons. Well, then, if the super-psi hypothesis is ad hoc, so also is the survival hypothesis since it must rely on super-psi assumptions, or further assumptions whose only purpose for being invoked is that they would lead us to expect discarnate persons to have greatly enhanced cognitive and causal powers. But take another example, this time from Almeder. He argues that if reincarnation is true, then we would expect to find people with past life memories, which Almeder says is confirmed by the fact that people claim to have past life memories. Setting aside that this is not a specific prediction, Almeder makes it clear that what sanctions the prediction here is the psychological criterion of personal identity. So here’s an admission of an auxiliary hypothesis, but clearly more needs to be assumed because we would have to account for a potentially disconfirming datum, to wit, many people appear to have no past life memories.
There are, of course, many auxiliary hypotheses we could introduce here so that the reincarnation hypothesis was consistent with the facts: people remember past lives but claim they don’t, people don’t recall their past lives because they possess them in the form of repressed memories, their last reincarnation was as a non-human and their memories were erased (perhaps memories only pass from human to human incarnations), they will eventually recall their past life at some point in their present life, or people with no past life memories are living their first life. It doesn’t matter which of these we select, or none. The point here is that a reincarnation hypothesis requires that we build into it assumptions that are no less ad hoc than the ones needed by an extra-terrestrial hypothesis. And here it seems to me that living-agent psi hypotheses have a plausible advantage. As Braude has shown, whatever we might say about so-called super-psi, to the extent that survivalists take seriously the evidence for living-agent psi, there is at least independent evidence for “dandy psi,” as exemplified, for example, in the more impressive remote viewing experiments in the Stargate Project. In my paper critiquing David Lund’s argument for survival (referenced above), I argued that ordinary psi, which includes “dandy psi,” is sufficient to pose an explanatory challenge to the survival hypothesis. I’d say this advantage would extend to their comparative prior probabilities, at least to the extent to which independent support is being invoked as a determinant of prior probability. I think it’s plausible to construct a robust living-agent psi hypothesis with reference to dandy psi.