Bayesian Probability and the Resurrection: A Reply to Brian Blais

Last weekend, Premier Christian Radio’s Unbelievable? show aired a conversation between myself and atheist author Jonathan MS Pearce. The topic was “Do Extraordinary Claims Require Extraordinary Evidence?” (about which I recently published an article critiquing Pearce’s writing). It was a cordial exchange, though unfortunately due to the time constraints and the breadth of the topic we were not able to cover the subject in anything like sufficient depth to do it justice (though we did succeed in highlighting where our key differences lie). This week, a two part series of essays was published by atheist physicist Dr. Brian Blais, a professor of Science and Technology at Bryant University, and a research professor at the Institute for Brain and Neural Systems, Brown University (I have had a public conversation with Dr. Blais previously; see here). These essays challenged some of my key contentions from the debate, particularly my employment of Bayesian epistemology. In this essay, I reply to Dr. Blais’ remarks.

Blais opens,

In the Unbelievable podcast episode Do extraordinary claims require extraordinary evidence? Jonathan McLatchie vs Jonathan Pearce I was struck by several claims and points, especially made by Jonathan McLatchie. I went back and listened to two other debates with him, one with Matt Dillahunty and one with myself and found some of the points Pearce made against McLatchie were things that both I and Matt had brought up — mostly about McLatchie’s apparent over-fondness for stories over real evidence.

I find it curious that, in an essay that alleges that I do not understand Bayesian epistemology, Blais refers to my debate with atheist TV personality Matt Dillahunty. Dillahunty’s lack of understanding of epistemology and the nature of evidence is second-to-none, as I reveal in this interview, and in more detail in a series of discussions with my colleague Dr. Timothy McGrew, a professional epistemologist (part 1; part 2; part 3; part 4). Given Blais’ apparent interest in the proper application of Bayesian probability theory, why does Blais not say a word about Dillahunty’s incredibly poor understanding of it, which was made plain at numerous points during his debate with me?

Spontaneous Proton Decay

In the debate with Pearce, I borrowed an analogy from Timothy McGrew, drawn from the physical sciences. According to one proposed theory in nuclear physics, spontaneous proton decay occurs, though it is such a rare event that no instances have ever been observed. In order to test this theory, scientists set up sensitive detectors in underground water tanks and leave them there for decades in order to determine whether spontaneous proton decay in fact occurs. According to a frequentist metric, this would be judged to be an extraordinary claim. However, in the case of spontaneous proton decay, there may be theoretical considerations that inform our background knowledge and therefore increase the event’s prior probability. The hypothesis of spontaneous proton decay, despite its having never been observed, is not a wild guess, but rather has a theoretical underpinning. Thus, the rarity or unprecedented nature of an event should not be the only, or even the primary means by which we assess the prior probability of said event. Likewise with the resurrection, the idea that Jesus rose from the dead, even before consideration of the direct evidence, is not some wild guess without any relevant background considerations. Rather, it is made plausible by other background considerations, as enumerated in my previous essay.

In reply to this point, Blais comments,

The comparison is laughable to me, given the differences in the “theoretical underpinning” of each, but let’s see how the comparison plays out. The Standard Model predicts that proton decay will not happen — that it violates a conservation law — however many models that build off of the Standard Model do include proton decay. Given the state of affairs, I’m not sure if scientists put a large prior on proton decay anyway, but let’s grant that — say, we are reasonably confident that proton decay will occur. What theoretical underpinning would we have to give us this confidence?

Whether or not spontaneous proton decay has a high prior is a moot point, since the point of the analogy is that considerations other than frequency can and do bear on the prior probability of an event occurring. Indeed, other considerations can increase the prior to a level that is higher than would be the case if we were only going by frequency, even if it does not raise the prior to a level greater than fifty percent. Incidentally, it is not necessary for us to defend that the hypothesis of Jesus’ resurrection has a high prior (if by that we mean greater than 0.5). Rather, one need only show that other considerations render the prior not incredibly low (as might be the case if we were assessing the prior only in terms of the frequency with which resurrections have been observed to occur). Blais notes five examples of times that science has predicted rare and unprecedented events or effects, including:

  • the motion of the Earth through the Ether (e.g. Michaelson-Morely experiment) 
  • the bending of star light by a massive object (e.g. Einsteins GR prediction) 
  • civilizations on Mars (e.g. Lowell’s canals) 
  • the existence of the Higgs Boson (1967 work by many scientists) 
  • proton decay (predicted by many Grand Unified theories)

Blais notes that “several of these were not confirmed, yet at the time there were good reasons to think they were true. Whenever science has predicted a rare, unprecedented event there have been common patterns.” These include (1) consistency between the novel hypothesis and other confirmed predictions of other hypotheses; (2) specific limits  on the new phenomena that are predicted by the hypothesis; and (3) confirmed predictions.

Blais then draws a comparison between the theories around proton decay and the resurrection hypothesis and asserts,

  • “one is consistent with every other confirmed prediction from the other theories — the other isn’t.”
  • “one predicts specific limits on the novel phenomena, including time frames and what specific steps we’d need to confirm it — the other doesn’t.”
  • “one has a pile of other novel specific, quantitative predictions that have since been confirmed — the other doesn’t.”

There is nothing inconsistent between natural law and the hypothesis that God has worked a miracle. Atheist philosopher John L. Mackie rightly defined a miracle this way [1]:

What we want to do is to contrast the order of nature with a possible divine or supernatural intervention. The laws of nature, we must say, describe the ways in which the world — including, of course, human beings — works when left to itself, when not interfered with. A miracle occurs when the world is not left to itself, when something distinct from the natural order as a whole intrudes into it.

Thus, special divine action does not contradict the laws of nature since the laws of nature do not describe a miraculous occurrence — that is, when nature is not left to itself. There is nothing in theoretical physics, therefore, that contradicts the hypothesis that God miraculously raised Jesus from the dead, since on this hypothesis the regularities of nature were disturbed by an intervention from outside the system.

Blais finishes this section of his remarks with four questions addressed to me. The first is, “Can you make a prediction of when miraculous healing, or any other miracles, will happen — or is it always post-hoc?” To this question, I would point out that (contrary to popular notions) it is not necessary for a hypothesis to be able to make high probability predictions in order for it to be well evidentially supported. Rather, it is only necessary that the pertinent data be rendered more probable given the hypothesis than it would be on its falsehood. My colleague Timothy McGrew has often given the following analogy: Suppose you are walking in a forest and stumble upon a shack that, upon initial inspection, appears to be uninhabited. Nonetheless, you decide to investigate. As you open the door, you notice a table, upon which there is a tumbler containing Earl Grey tea, which is still steeping. Now, on the hypothesis that the shack is inhabited, does it predict with high probability the presence of the steeping Earl Grey tea on the table? Hardly! Nonetheless, this observation is very strong evidence that the shack is inhabited, since on that supposition the presence of the tea (even though improbable) is far, far more probable than it would be on the falsehood of that hypothesis. What is important, then, is the likelihood ratio of the probabilities. A corollary of this is that a hypothesis does not need to render an observation probable in order for that observation to count as confirmatory evidence in its favor. In regards to the various considerations that bear on the prior probability of Jesus’ resurrection (e.g. the Trilemma argument and Messianic prophecy), it is also not necessary that they render Jesus’ resurrection highly probable. The point is only to show that these factors render it at least plausible.

Second, Blais asks, “Can you make any description of a measurement that could be done, no matter how impractical, for directly confirming any of the predictions from the God-theory?” As stated above, it is not necessary for theism to make high probability predictions in order for it to be confirmed by the evidence. Even if theism does not predict the information content in DNA, or the fine-tuning of the physical laws and constants, it certainly renders those observations much more probable than they otherwise would be. This raises the probability of theism being true.

Third, Blais asks, “You mentioned a case of a healing of Irene McDonald with multiple sclerosis, but I can’t find the medical details (i.e. the actual doctor’s reports, the detailed timeline, etc…). Same with the case of Barbara Schnyder. Can you provide those?” Having access to the official medical records would certainly strengthen the case for miraculous healing in the case of Irene McDonald and Barbara Schnyder. However, it is a fallacy to say that, because we could in principle have better evidence, we therefore do not have good or sufficient evidence. One can have compelling evidence even when a case could be rendered stronger with even more evidence. In the case of Barbara Schnyder, we have testimonial evidence from her physicians [2], as well as from Schnyder herself. Independent eyewitness testimony to the ‘before’ and ‘after’ condition ought to be sufficient evidence that the healing (whatever the cause) took place. The only thing that remains is what best explains the healing — a miracle or a natural remission. Jonathan Pearce wants to explain these as a natural remission (since MS often presents itself as a relapsing and remitting condition). However, I do not find this to be plausible, since MS remissions are typically over the course of days, weeks, or months. In the case of Barbara Schnyder and Irene MacDonald, the healing has endured (at the time of reporting) for more than thirty-five and twenty-five years respectively. Moreover, I do wonder whether Jonathan Pearce is aware of anyone with MS as severe as Barbara or Irene who spontaneously had a remission to perfect, enduring health. Though MS does fluctuate, it does not, to the best of my knowledge, disappear so completely and permanently once it has become so progressive. The naturalistic hypothesis also of course has difficulty explaining, in the case of Barbara Schnyder, her testimony that she heard a male voice speak from behind her, “My child, get up and walk!”, right as the healing took place.

Fourth, Blais asks, 

How did you determine that these people were healed by prayer and also by God? You can’t answer “because they prayed and she was healed after” (post-hoc fallacy) and you can’t answer “because they prayed to the Christian God”, because some other God may have answered, or some other thing. You have to, in your explanation, be able to distinguish from the prayers that haven’t worked, the cases where healing happened without prayer or with prayer to another God. Without this, you can’t make a reasonable claim that God-did-it.

If parallel examples exist with such complete and enduring healing from MS without the involvement of prayer, that would significantly undermine these examples as evidence for special divine action. If Blais knows of any such examples, then I invite him to offer them. Furthermore, it is legitimate to highlight the co-incidence between prayer and miraculous healing. When miraculous healing and prayer coincide, that does provide evidence that tends to confirm the efficacy of prayer. Indeed, if it is the case (as many atheists allege) that unanswered prayer is evidence against the existence of God, then it necessarily follows that answered prayer is evidence that confirms the existence of God. How strong that evidence is in confirming theism will of course depend upon the probability of the same result assuming only natural operations. In the case of the spontaneous healings of Barbara Schnyder and Irene McDonald, these seem to me to be considerably unlikely if we assume only natural operations. It thus contributes significant evidence that tends to confirm the theistic hypothesis. Jonathan Pearce rightly pointed out in the debate that there plenty more examples of where someone has asked God for healing and yet has not been healed. It is appropriate to take those cases as constituting some evidence against the existence of God. However, as I noted in the debate, there is an epistemic asymmetry — these instances constitute, piece for piece, significantly less evidence disconfirming theism than their counterparts do in confirming it. Christianity even predicts that some peoples’ prayers will go unanswered, since sometimes what we ask for is not what is best for us; or often God wishes to teach and shape us in the midst of our suffering. Sometimes even sin can hinder our prayers (Prov 28:9; 1 Pet 3:7). Thus, while Christianity does not particularly struggle to account for prayers going unanswered, naturalism does seriously struggle to account for the wide variety of apparently miraculous answers to prayer as documented by Keener and others. Though Blais suggests that “some other God may have answered, or some other thing,” I would argue that the prior probability in the case of the Christian God answering prayers is considerably higher than that for other deities, given the independent evidence we have for the truth of Christianity (e.g. the evidence bearing on the resurrection). It is also unclear why another deity, or “some other thing” would have motivation to answer prayers directed to the God of the Bible, whereas it is not at all unclear why the God of the Bible would have motivation to do so. Thus, while these instances do not prove that the God of the Bible was responsible, they do offer evidence that tends to confirm that conclusion.

Bayes Factors and Confidence

Blais quotes me as saying, “If you have two pieces of evidence each with a base factor of 10 and then combined they have a cumulative base factor of 100.” He then notes, 

He thinks in terms of independent data pushing the likelihoods up and down so that once you’ve used one data point to update the likelihoods, you can start with that updated likelihood with the next data point — ignoring any possible dependency between one data point and another.

Of course, my statement assumes the pieces of data are independent of one another. I do not, contrary to Blais’ suggestion, “[ignore] any possible dependency between one data point and another.” To draw a connection to the argument for the resurrection, the Cambridge philosopher Arif Ahmed has pointed out that in a stage illusion or trick, given that at least one person has been fooled, the probability that many other people will be fooled as well is significantly increased. [3] Thus, successive pieces of evidence fail to add significant force to the case that a true miracle was observed. However, as Timothy McGrew observes concerning the claimed resurrection of Jesus, “The witnesses are not all confined to one vantage point, as they were in the case of the stage magician. If they agree, it is much more difficult to find a single simple explanation for how they could all have been fooled. Their testimonies are not bare assertions that the event in question happened. They may include details that interlock with details in other testimonies in ways that increase their credibility.”

Is the Rarity of Miracles Evidence Against Them?

Blais quotes me as saying:

The hypothesis that God has used miracles as authenticating signs predicts that they will stand out against the normal course of nature. The fact that miracles do in fact stand out against nature cannot be taken as evidence against miracles because the hypothesis predicts very strongly that’s what you’ll find.

Commenting on this point, Blais remarks,

What McLatchie doesn’t seem to see is that regardless of the source any purported cause of an event which is rare has a low prior, which should be used against it when comparing to other causes with higher priors. Since God is supposedly using miracles as “authenticating signs”, if you are going to interpret any particular event as a miracle you have to overcome the low prior — “authenticating signs” would be rare — support it with stronger evidence than would be needed for a more mundane explanation.

However, this is incorrect. A miracle is only able to serve as an authenticating sign if the relevant event has a low prior probability on the assumption that it is brought about by physical processes. But Christians are not saying that miracles are an unexpected anomaly that violated natural law. Rather, we are saying that God has wrought miracles for the purpose of using them as authenticating signs. This is quite different. The hypothesis that God has intervened in nature to bring about miracles, and that he has done so that they may serve as authenticating signs, cannot be assigned an ultra-low prior probability on the pure basis that it deviates from the normal course of nature (since this is highly predicted on the hypothesis). Other considerations must be brought to bear.

Blais further asks,

I would further ask McLatchie, why should authenticating signs be rare in the first place? Why couldn’t God make a miracle like his name written in the clouds every time the sun comes up, to authenticate his existence and power? Also, would McLatchie have to be making arguments like this if God were a lot more obvious?

Again, that the evidence could have been stronger does not entail that the evidence we do have is insufficient. If God has given us adequate evidence to believe that he exists, then he cannot be reasonably faulted for not having provided even more evidence. God’s existence is indeed abundantly apparent from the created order, especially the strikingly designoid features of living organisms. For more on the problem of divine hiddenness, I refer readers to my essay on this subject.

Jesus’ Death at Passover

Blais writes,

The one that really perplexes me is Jonathan’s perspective on the prophecy that Jesus would be executed on Passover (or around Passover, depending on which Gospel you read).

There is no prophecy that Jesus would be executed on Passover, assuming a narrow definition of prophecy which requires a specific prediction and a corresponding fulfilment. However, it is a striking feature that Jesus’ death coincides with the day of Passover, which commemorates Israel’s deliverance from Egypt and the sparing of the firstborn sons of the Hebrews through the smearing of the blood of a slaughtered lamb on their doorposts, leading the angel of death to “pass over” those homes. In New Testament theology, Christ is the fulfilment of the Passover imagery. Though by no means anything like sufficient on its own to establish the truth of Christianity, that is, in my view, given the theological import and symbolism, evidence that tends to confirm the truth of Christianity (which may be taken to contribute to a broader cumulative argument). It is therefore relevant to our assessment of the prior probability of Jesus’ resurrection. I discuss other such examples in this essay.

Blais appears to have bought into the false idea (though popular among New Testament critics) that the synoptic gospels and John differ on the precise day on which Jesus was crucified. This, however, is demonstrably false, as I will show shortly.

In response to this point about the Passover, Blais remarks,

There are very good story-reasons for having this be the case, the authors of the Gospels knew the Old Testament stories, and could interpret the Jesus story in light of them and write the story accordingly. This is so straightforward an explanation, with very little to counter it — even if you believe the writers to be eye-witnesses. All you’d need is for Jesus to be executed sometime in the vicinity of Passover, a short time of an oral tradition or retelling of the stories by the same people (which get naturally embellished) and you’ve got it. McLatchie’s argument is so uncompelling for anyone who isn’t already convinced that it is remarkable.

There are actually various lines of evidence that support the historicity of Jesus’ death on the day of Passover, though time did not permit me to discuss them in the debate with Pearce. First, this is a detail attested by all four gospels and implied by Paul in 1 Corinthians 5:7. The evidence bearing on the substantial trustworthiness and scrupulousness of the gospel authors is also relevant here. This is not a detail that the gospel authors (or their sources) plausibly misremembered, since so many details in the gospels are connected to Jesus’ death being at the time of Passover. For example, in John 18:28, we read, “Then they led Jesus from the house of Caiaphas to the governor’s headquarters. It was early morning. They themselves did not enter the governor’s headquarters, so that they would not be defiled, but could eat the Passover,” (Jn 18:28). Furthermore, Pilate is said to have declared to the Jews, “But you have a custom that I should release one man for you at the Passover. So do you want me to release to you the King of the Jews?” (John 18:39). All three of the synoptic gospels also explicitly refer to the last supper (the night before Jesus’ crucifixion) as being the Passover seder (which was eaten on the eve of Passover on the fourteenth day of Nisan, the night before the day of Passover). This meal is independently attested by John. Though he does not explicitly call it the Passover meal, it is clearly the same event that is described in the synoptics. This is revealed by two undesigned coincidences (which also tend to confirm this meal as an historical event). Here are the relevant texts from Luke and John: Luke 22:27 – 

  • “For who is the greater, one who reclines at table or one who serves? Is it not the one who reclines at table? But I am among you as the one who serves.” 
  • John 13:4-5 – “[Jesus] laid aside his outer garments, and taking a towel, tied it around his waist. Then he poured water into a basin and began to wash the disciples’ feet…” 

A reader of Luke’s gospel might ask what Jesus is referring to when he says “I am among you as the one who serves.” John’s account, however, illuminates the broader context. Jesus gave the disciples an object lesson in servanthood by washing the disciples’ feet, an event not recorded in Luke’s account of the same occasion. A reader of John’s gospel might ask, why does Jesus wash the disciples’ feet? What prompted this object lesson in servanthood? The answer is given in Luke 22:24 – “A dispute also arose among them, as to which of them was to be regarded as the greatest.” In other words, John reports the object lesson in servanthood, whereas Luke reports the occasion that gave rise to it — namely, the dispute among the disciples over who was the greatest. Thus, Luke explains John, and John explains Luke. These two undesigned coincidences corroborate the historicity of the event and show that the two accounts reflect one and the same event.

That Jesus died on (or at least around) the day of Passover is also supported by another undesigned coincidence connected to Jesus’ triumphal entry. Consider John 12:1-2,12-13: 

Six days before the Passover, Jesus therefore came to Bethany, where Lazarus was, whom Jesus had raised from the dead. 2 So they gave a dinner for him there. Martha served, and Lazarus was one of those reclining with him at table…12 The next day the large crowd that had come to the feast heard that Jesus was coming to Jerusalem. 13 So they took branches of palm trees and went out to meet him, crying out, “Hosanna! Blessed is he who comes in the name of the Lord, even the King of Israel!”

John here has given us a very specific extraneous detail (which none of the other gospels gives us): Jesus arrived at Bethany six days before the Passover, and the following day rode into Jerusalem on the back of a donkey (which would have been five days before the Passover). Can we confirm John’s accuracy on this? Yes, we can. Turn over to Mark 11:1-11, which parallels the arrival at Bethany (although Mark does not give us the time-stamp that John provides): 

Now when they drew near to Jerusalem, to Bethphage and Bethany, at the Mount of Olives, Jesus sent two of his disciples 2 and said to them, “Go into the village in front of you, and immediately as you enter it you will find a colt tied, on which no one has ever sat. Untie it and bring it…7 And they brought the colt to Jesus and threw their cloaks on it, and he sat on it. 8 And many spread their cloaks on the road, and others spread leafy branches that they had cut from the fields. 9 And those who went before and those who followed were shouting, “Hosanna! Blessed is he who comes in the name of the Lord! 10 Blessed is the coming kingdom of our father David! Hosanna in the highest!” 11 And he entered Jerusalem and went into the temple. And when he had looked around at everything, as it was already late, he went out to Bethany with the twelve. 

Mark does not tell us that Jesus approached Bethany six days before the Passover, nor that it was the following day that Jesus rode into Jerusalem. However, it appears implicit that they fetched the colt early in the morning — since the disciples fetch the colt, there is the triumphal entry and Jesus and the disciples entered the temple and “looked around at everything” (which was presumably a whole day’s activities). If, then, we assume that Jesus entered Jerusalem five days before Passover, then we can begin counting off the days narrated in Mark’s gospel, to see if the narrative synchronizes with that of John. Verses 12-14 narrate the cursing of the fig tree, which according to verse 12 happened “the following day” (i.e. four days before the Passover, assuming John’s chronology to be correct). Jesus then cleansed the temple and according to verse 19 “when evening came they went out of the city.” In verse 20, we read, “As they passed by in the morning, they saw the fig tree withered away to its roots.” We are now therefore at three days before the Passover. In Mark 13, we read of the Olivet discourse on the Mount of Olives. This, we can reasonably assume, took place in the evening, since the Mount of Olives was mid-way between the temple in Jerusalem and Bethany where Jesus and the disciples were staying. This, then, marks the end of three days before the Passover. When we turn over to Mark 14, we read in verse 1, “It was now two days before the Passover.” Mark and John thus calibrate perfectly, yet in a very casual and incidental fashion, thereby corroborating the time-stamp given to us by John.

Jesus’ death on the day of Passover is also confirmed by a reconcilable variation that supports the independence of the synoptics and John on this point. Recall that a reconcilable variation is when there is an apparent tension between two texts that dissolves upon closer inspection or upon learning new information, thereby confirming the independence of these accounts. Consider the following two texts: 

  • Mark 14:12 – “And on the first day of Unleavened bread, when they sacrificed the Passover lamb, his disciples said to him, “Where will you have us go and prepare for you to eat the Passover?” [The Last Supper follows] 
  • John 19:14 – [Describing the scene where Jesus is condemned to be crucified] Now it was the day of Preparation of the Passover. Now, here is the apparent contradiction. In Mark, the Last Supper occurs on the first day of Passover. Jesus’ arrest takes place that same night and his crucifixion on the following day, i.e. the first day of Passover. 

According to many New Testament critics, John tells us that the crucifixion took place on the day of preparation for Passover — that is, the day before the Passover (thus contradicting the synoptics, which have the crucifixion on the day of Passover). However, John does not say that it was the day of preparation for the Passover. Rather, he says that it was the day of preparation of Passover. Mark 15:42 uses the same term, and also indicates what it means: “And when evening had come, since it was the day of Preparation, that is, the day before the Sabbath.” In other words, “preparation” means preparation for the Sabbath. Now, let’s read John more closely. In John 19:31, we read, “Since it was the day of Preparation, and so that the bodies would not remain on the cross on the Sabbath (for that Sabbath was a high day), the Jews asked Pilate that their legs might be broken and that they might be taken away.” John therefore agrees with Mark that it was the day before the Sabbath. This is what he means by “preparation.”

This still leaves a few loose ends that need addressing. First, what does John mean when he says, “for that Sabbath was a high day?” I would argue that he means that it was a particularly special feast day, not just any Sabbath day, but Sabbath in Passover week. Second, what about John 18:28?: “Then they led Jesus from the house of Caiaphas to the governor’s headquarters. It was early morning. They themselves did not enter the governor’s headquarters, so that they could not be defiled but could eat the Passover.” Does this not contradict the claim that Passover had already taken place? The short answer is ‘no’. Passover is not just one day; rather, it is a week-long festival. Indeed, throughout the gospel of John, the word “Passover” occurs eight other times and always refers to the whole festival, not only to the opening meal. The Passover seder, or supper, that marks the commence of the Passover celebration is in fact not the only ritual meal that is eaten during Passover; in fact, there is another ritual meal, the chagigah (meaning “festival offering”), eaten at mid-day the following day. This is known as the feast of unleavened bread, but Josephus indicates that it was also known as the Passover meal. He writes, “As this happened at the time when the Feast of Unleavened Bread, which we call Passover,…” (Antiquities 14.2.1). If the chief priests had entered Pilate’s dwelling and consequently become ceremonial unclean, the defilement would expire at sundown. All that would be required of them would be to wash and they would be made ceremonially clean for the evening meal. Therefore, the chief priests have to have been concerned about some meal other than the evening meal. That is to say, their concern cannot have had to do with the initial Passover seder meal. Thus, John’s account fits perfectly with Mark’s. The concern of the chief priests cannot have been about the initial Passover seder. The seder was already over, having been eaten the previous evening. Rather, the chief priests were concerned about another meal in Passover, probably the feast of unleavened bread known as the chagigah. Finally, what about John 13:1? “Now before the Feast of the Passover, when Jesus knew that his hour had come to depart out of this world to the Father, having loved his own who were in the world, he loved them to the end.” Does this not contradict the claim that the Last Supper was the Passover meal? In Greek, however, the text does not say that the supper was before the feast: rather, all it says is that before the feast, Jesus loved his disciples to the end. There is thus no great problem here either.

This reconcilable variation suggests that the synoptics and John are independent, and thereby confers evidential support to the view that Jesus’ death coincides with the day of Passover. When all of the above considerations are taken together, one has, in my view, a strong cumulative evidential argument for supposing Jesus’ death to have taken place on the day of Passover. My statement made in the debate with Pearce therefore stands.

Undesigned Coincidences

Blais then offers a brief comment about the argument from undesigned coincidences for gospel reliability (about which I have written quite extensively). He writes,

I think my biggest problem with this line of thinking is that there are major differences between the narratives in the Gospels. For example, the nativity narratives are contradictory — just try to map out the path Jesus’ parents take. The resurrection narratives too do not agree with each other, like the number of people coming and going, whether things were open or closed, or who actually saw Jesus. When apologists have to make arguments (i.e. the text is not clear) to justify seeming discrepancies in large events, why should we pay attention to subtle agreements? Why should we be surprised that there are some agreements, when we know that there are literary and other source dependencies between the anonymous writings of the Gospels? Again, McLatchie’s argument is so uncompelling for anyone who isn’t already convinced that it is remarkable.

The positive argument from undesigned coincidences and the negative argument from discrepancies between the accounts are two different arguments and must be separately evaluated. Even if it be granted that the gospels contradict each other on some points (and I am personally inclined to think that, on a handful of peripheral matters, they do contradict), that does not tell us anything about the positive evidence for the gospels. At most, one would have to conduct a general analysis of the competing evidence for and against the gospels as a whole, but it is highly simplistic to maintain that they just cancel each other out. Both can be taken to be valid forms of evidence. I would also argue that there is an epistemic asymmetry such that the evidence from undesigned coincidences (and other hallmarks of verisimilitude) is stronger evidence confirming the reliability of the documents than their counterparts (such as contradictions) are in disconfirming reliability. In my assessment, no demonstrable examples exist of deliberate distortions of fact in the gospels — all actual discrepancies are plausibly explained by an author making a minor good faith error, which is of very little consequence to our evaluation of these document’s substantial reliability. Furthermore, when we calibrate our expectations by looking at other sources — even those deemed to be generally trustworthy — we find many discrepancies, often much more grave than those we find in the gospel accounts. Timothy and Lydia McGrew document several examples from ancient literature [4]:

Even a passing acquaintance with the documents that form the basis of secular history reveals that the reports of reliable historians, even of eyewitnesses, always displays selection and emphasis and not infrequently contradict each other outright. Yet this fact does not destroy or even significantly undermine their credibility regarding the main events they report. Almost no two authors agree regarding how many troops Xerxes marshaled for his invasion of Greece; but the invasion and its disastrous outcome are not in doubt. Florus’s account of the number of troops at the battle of Pharsalia differs from Caesar’s own account by 150,000 men; but no one doubts that there was such a battle, or that Caesar won it. According to Josephus, the embassy of the Jews to the Emperor Claudius took place in seed time, while Philo places it in harvest time; but that there was such an embassy is uncontroversial. Examples of this kind can be multiplied almost endlessly.

Furthermore, actual discrepancies between accounts also tend to support the independence of the accounts. One might therefore say that actual discrepancies between the gospel accounts have multiple epistemic vectors — being negatively epistemically relevant to the reliability of the accounts while also supporting the independence of the narratives more broadly (and independent accounts that overlap concerning an event constitute evidence of its truth). For those reasons, contradictions between the accounts, even if successfully sustained, constitute evidence that is comparatively weaker in disconfirming the reliability of the testimony than the positive lines of evidence (such as undesigned coincidences) are in confirming it.

That being said, the four examples of contradictions that Blais offers are in fact quite weak. Though it is often alleged that Matthew and Luke contradict each other on whether Jesus’ family travelled from Nazareth to Bethlehem for the census (Luke) or whether they already lived in Bethlehem (Matthew). It is certainly the case that Matthew seems to imply that, by the time of Herod’s order to slaughter the infants in Bethlehem, Jesus and his parents had already been living in Bethlehem for a year or two, since Herod’s instruction, after asking the Magi when they first saw the star, is for his soldiers to kill all of the infants who are two years old and under. There is also no mention of the stable in Matthew, and the Magi are said to have visited Jesus’ family in “the house” (Mt 2:11). It is also apparent in Matthew that, following the death of Herod the Great, Jesus’ family intended to return to the Judea region, presumably Bethlehem (Mt 2:19-22). This again supports that Jesus’ family had settled in Bethlehem. However, this can be readily harmonized if we imagine that Joseph had family in the Bethlehem region (presumably why they had travelled there from Nazareth), and that they had planned to relocate to Bethlehem after the wedding. We know that Mary had family in the Judean hill country (Lk 1:39-45), so plausibly Mary had met Joseph through this connection, even though Joseph’s family lived in Bethlehem rather than Nazareth. Though Luke 2:4-5 is often taken to indicate (implausibly) that the Jews were required to be registered at the hometowns of their ancestors from a millennium earlier, this is not a necessary interpretation. Plausibly, the reason Joseph and Mary went to Bethlehem was that it was Joseph’s hometown. The family line may have remained there since the days of David, and their Davidic descent could well have been something in which they took great pride.

There is one (and only one) discrepancy in the nativity narratives for which there is, in my assessment, a strong case that it is an actual mistake, and that is that Luke appears to be unaware of the flight to Egypt and implies that Jesus’ family went to Nazareth very shortly after the purification according to the law of Moses (Lk 2:39). This is, however, unlikely to be a deliberate distortion of fact on Luke’s part, since Luke shows himself elsewhere (on many occasions) to be extremely scrupulous in his reportage of historical information. There is also reason to believe that the flight to Egypt (though only reported by Matthew) is historical, as I discuss in this essay (see under the heading Joseph and Archelaus). There is also evidence that these accounts are very early (e.g. Zechariah’s prophecy in Luke 2:67-79 betrays a very second temple Jewish, pre-Christian, Messianic expectation) and that they are based on eyewitness testimony (e.g. notice the consistent one-sided reporting of Matthew’s account from the perspective of Joseph and Luke’s from the perspective of Mary, even to the extent of describing their private thoughts). Other evidence for the reliability of the nativity accounts is discussed by Lydia McGrew in this video. Plausibly, then, Luke’s source (presumably Mary) may have omitted to mention the flight to Egypt, simply moving on to discuss the later events that transpired when the family was later living in Nazareth (such as Jesus teaching in the temple at twelve years old). Luke, then, presumably created a natural transition between the pericopes that is technically in error, though resultant from a good faith oversight.

The other three items that Blais highlights pertain to the resurrection accounts. The first is “the number of people coming and going,” which I presume is a reference to the identity of the women discovering the empty tomb on easter morning (which varies between the four gospels). However, Luke makes it abundantly clear that his intention is not to provide an exhaustive list, since he says explicitly, “Now it was Mary Magdalene and Joanna and Mary the mother of James and the other women with them who told these things to the apostles” (emphasis added),” (Lk 24:10). Furthermore, though John 20:1, if read in isolation, might leave the reader with the impression that it was only Mary Magdalene who visited the tomb on the Sunday morning, we read in the next verse, So she ran and went to Simon Peter and the other disciple, the one whom Jesus loved, and said to them, “They have taken the Lord out of the tomb, and we do not know (οὐκ οἴδαμεν) where they have laid him,” (emphasis added; Jn 20:2). The word οἴδαμεν is the first person plural form of οιδα, meaning “to know,” and the word οὐ / οὐκ is an adverb that negates the verb, hence “we do not know.” Thus, Mary’s use of the plural in this verse implies that there were in fact other women who had been present with Mary at the tomb. Indeed, this incidental, casual, dovetailing between the synoptics and John supports the independence and therefore historicity of the accounts.

The next item in Blais’ list is “whether things were open or closed,” presumably a refence to the popular objection that Matthew 28:1-5 intends us to understand that the women were witnesses to the descent of the angel and the rolling away of the stone, whereas Mark 16:3-4 and Luke 24:2 clearly state that it had been rolled away already when they arrived. However, in Matthew, the passage regarding the angel (verses 2-4) is introduced by the Greek particle γάρ, “For …” The purpose of this particle is to explain the earthquake and the state of affairs as discovered by the women upon arriving at the tomb. Furthermore, Matthew employs an aorist participle, which could well be translated with the English past perfect: “… for an angel of the Lord had descended … ” Both of these considerations dissolve this alleged discrepancy.

As for who actually saw Jesus (Blais’ final alleged contradiction), there is of course no reason to demand that the gospels give us an exhaustive list of all appearances, or that they each report the same appearances as the other three. Indeed, omission is not the same thing as denial. There is nothing epistemically troubling about the gospel accounts reporting different aspects of Jesus’ post-resurrection appearances, especially given that the appearances are said to have taken place over a forty day period (Acts 1:3). In fact, the divergence between the narratives in the details they choose to report points to the independence of the accounts concerning Jesus’ resurrection.

Blais asks, “Why should we be surprised that there are some agreements, when we know that there are literary and other source dependencies between the anonymous writings of the Gospels?” Literary and source dependences between the gospels does not rob undesigned coincidences of their evidential value. Like so many critics, Blais does not seem to appreciate the value of casualness in assessing historical reliability. Indeed, an undesigned coincidence can even occur in the same document, or a writer might provide information that unintentionally corroborates (in a manner that can be detected) some fact that the said author is also aware of. However, it is evidently the case that an author may read another source and then at a later time mention something in his own writing, drawing from his own independent knowledge, that serves to corroborate a detail in the source he had read, without deliberate intent. How would such an undesigned corroboration be recognizable to another person reading the two sources? The features that allows detection of the two sources is the appearance of casualness and unconnectedness.

Lack of Priors?

Finally, Blais comments on an alleged failure on my part to take the priors into account when performing Bayesian analyses. He remarks,

It is striking how many recent apologists use Bayesian reasoning yet ignore the prior — which means they really aren’t using Bayesian reasoning but want to just sound like they are. This is a form of dishonesty, I feel, but at best it is a clumsy analysis — and a problem that has been pointed out for the McGrews, Calum Miller here and here, Blake Giunta, and of course mentioned in my conversation with Jonathan McLatchie. The use of Bayes Factors, which also ignores priors, is indicative of this approach. Why do apologists want to ignore priors? Because what they are arguing for is, a-priori, seriously unlikely.

This is a very odd assertion, especially given how much of the debate with Pearce was spent debating how the prior probability of the resurrection ought to be determined (also discussed at length in this essay). I do not know, therefore, how Blais can reasonably assert that I “ignore the prior.” Brian Blais interestingly links to a discussion that streamed on the Digital Gnosis YouTube channel in which he participated along with Nathan Ormond, Kamil Gregor and James Fodor. The discussion interacts with the McGrews’ work (and to a lesser extent mine) on the resurrection and the epistemology of testimony, and is a full nine hours long. I viewed the first four and a half hours but the first half, I am afraid, was such poor quality that I could not bring myself to watch the second half. Among the panel’s many errors was a fundamental misunderstanding of the McGrews’ chapter in the Blackwell Companion to Natural Theology [5], since the panelists claimed (falsely) that the McGrews do not discuss prior probability in the chapter, whereas in fact the whole purpose of the chapter was to show how you can backsolve to reveal how low the prior probability would need to be in order to overcome the weight of the evidence and reduce the justified confidence in the conclusion below a given threshold (Lydia McGrew talks about this point in this video). The panelists also made ridiculous claims such that aliens raising Jesus from the dead is comparable to the hypothesis of God raising Jesus from the dead (overlooking the fact that the prior probability of God raising Jesus from the dead, given the religio-historical context, is much, much greater than the prior probability of aliens having the means or motive to raise Jesus from the dead). The panelists also brought up the ‘spiderman objection’ to the book of Acts getting hard details right (that is, spiderman comics getting details about New York city right does not entail the veracity of the stories concerning spiderman), an objection that I address in response to Bart Ehrman here. They also made the ridiculous suggestion that the reason why the gospels report women as being the chief discoverers of the empty tomb was that they were trying to pull a double bluff. The video also made false assertions such that the McGrews deny literary dependence between the synoptics (false) or that the McGrews subscribe to Matthean priority (also false). I am afraid that this discussion was incredibly poor quality.


In conclusion, I do not believe that Blais’ remarks have left so much as a dent in any of my arguments from the debate. His analysis would surely have benefitted from reading the work of myself and others, such as the McGrews, more carefully (and charitably) before offering his critiques. I would welcome the opportunity for further dialogue with Brian Blais (or, for that matter, Jonathan Pearce) so that we can iron out our differences in greater detail and make more progress than was able to be accomplished in my short debate with Pearce.


[1] John L. Mackie, The Miracle of Theism: Arguments For and Against the Existence of God (Oxford: Oxford University Press, 1983), kindle.

[2] Thomas Marshall, “Praying for a Miracle”, in Physicians Untold Stories (North Charleston, SC: CreateSpace, 2016), 115–22.

[3] Arif Ahmed, “Hume and the Independent Witness”, Mind 124, no. 496 (August 2015), DOI: 10.1093/mind/fzv076

[4] Tim McGrew and Lydia McGrew, “The Argument from Miracles: A Cumulative Case for the Resurrection of Jesus of Nazareth”, in The Blackwell Companion to Natural Theology, 1st Edition, ed. William Lane Craig and J.P. Moreland (Wiley-Blackwell, 2012), kindle.

[5] Ibid.

7 thoughts on “Bayesian Probability and the Resurrection: A Reply to Brian Blais”

  1. Pingback: Bayesian Probability and the Resurrection: A Reply to Brian Blais – Jonathan McLatchie – Tomáš Augustin :: Assorted Notes

  2. Pingback: Birth narrative Luke v. Matthew – Tomáš Augustin :: Assorted Notes

  3. Tomáš Augustin

    Dear Jonathan, first of all thanks for the articles – very thought through, readable, resourced. Now to actual comment:

    This is a second article where you refer to nativity account in Matthew as eyewitness testimony of Joseph.
    “they are based on eyewitness testimony (e.g. notice the consistent one-sided reporting of Matthew’s account from the perspective of Joseph and Luke’s from the perspective of Mary, even to the extent of describing their private thoughts).”

    I am wondering on which information (evidence) you arrive to this – isn’t it at minimum generally accepted tradition Joseph died before start of Jesus’s public ministry, i. e. well before Matthew has even become a disciple (not to mention well before he has penned his Gospel)?

    Thank you in advance for any clarification

  4. Brian Blais shows his bias as a physicist. Physics deals with nature at its most basic level and at that level there is often a high degree of predictability. Blais mentions the prediction that light would be bent by a massive object. That prediction was entailed by Einstein’s Theory of General Relativity, which he set out in 1915. It was possible to put that idea to the test during an eclipse in 1919.

    However, not everything in life is so amenable to experimental verification. It is believed that there has been a series of “supervolcanic” eruptions at Yellowstone National Park. So when will the next one occur? No one knows. Should we dismiss what geologists say about Yellowstone unless they can predict the next eruption? Obviously not.

    I’m surprised that Blais would challenge anyone to predict a miracle in advance. That shows real naivety. It is also naive to expect that miracles should be part of a worldview which is established in the same way as our scientific understanding of nature. Christians do have a worldview in which miracles make sense but it has to be evaluated on its own terms. Unfortunately, atheists are generally unwilling to do this.

  5. I see that Brian Blais has responded to this. I was interested in what he said about the shack in the forest scenario. Suppose you stumble on a shack in a forest which seems uninhabited. Then you go inside and discover a cup of tea that is still warm. That tells you that the shack must be inhabited. Brian Blais suggests an alternative interpretation. What if the tea was delivered by Jean Luc Picard, who then beamed back up to the Enterprise. So is this really analogous to a case where someone might be inclined to think that God has performed a miracle?

    If Jean Luc Picard is a real character, it would mean that he has the power to implant ideas in the minds of scriptwriters, producers etc. Since Star Trek is set in the future when there is far more advanced technology, this may not be impossible. There are problems, however. This kind of thing cannot be a general explanation for fiction. If I write a novel about a detective set in the present day, it doesn’t mean that there is a real detective who has implanted the idea in my mind. So is the idea just an explanation for science fiction? And is it an explanation for all science fiction or just some of it? Furthermore, why would Jean Luc Picard want people to know about him but leave them thinking he is just a fictional character? I don’t think a believer in Jean Luc Picard would be able to give a satisfying answer to those questions.

    We can agree that someone who attributes the cup of tea to Jean Luc Picard is irrational. However, it is not a mistake that anyone is likely to make. Nor is it a good analogy for a belief in a miracle-working God.

  6. Just read Blais’ response ( While I appreciate his cordiality and initiative in furthering the conversation, I was unimpressed by his arguments. In response to the Barbara Snyder case, he cites the following:

    “Greta Brandt was cured of Leukemia by aliens. Her doctor, Dr. Frans Wenderoth, confirms that she had the lethal disease and ‘something miraculous must have happened’ for her to not have it anymore. She herself claims that she was abducted by aliens and while on the ship they cured her.”

    In response, I left the following comment (awaiting moderation):

    “Some documentation for Snyder’s healing can be found here: Thus, we have a medical report, multiple independent eyewitness accounts (Marshall, Adolph, Snyder, maybe Flygare), and some weaker pieces of evidence like a newspaper clipping. The preponderance of evidence indicates that the healing probably happened. The question that needs to be addressed is what best explains it. Regarding the Greta Brandt case, I can’t find any documentation aside from articles (without citations) on suspicious websites. I can’t even confirm that Dr. Frans Wenderoth exists—I’ve used several different search engines (such as and haven’t found anyone at all by that name. Even if he does exist, I would prefer to see him make the statement for himself. In Snyder’s case, there are two qualified medical doctors who have testified to the healing, at least one of whom can be emailed. Thus, to claim that Snyder’s case is equivalent to Brandt’s is fundamentally misguided. If you have a better example of a ludicrous healing claim with the same documentation as Barbara, I would very much like to see it.”

    He also parodied certain arguments for biblical reliability and used them to argue for the historical reliability of the Harry Potter series of books. I appreciated his indication that that’s sometimes how the arguments come across, but these cases are really not parallel. He referenced “casual coincidences” between the books but neglected to provide any examples.

  7. Hi Jonathan
    I know I’m waiting for a response from you about the justice of God, but I thought in the meantime to post a response to you here on the resurrection and the McGrew article.
    I watched the whole of the Digital Gnosis episode, and I don’t think Brian Blais participated.
    I know it is a slog, but I think you would have got some benefit from watching the second half.
    In that, Fodor explains his RHBS naturalistic explanation for the evidence
    He allows, as the McGrews assume, that that gospel accounts do record some eyewitness testimony at least as far as what people THOUGHT they saw.
    I think that is a plausible model, and at least as good at explaining the evidence as the resurrection hypothesis.

    On the McGrew article, there are several problems with it
    1) It is problematic to talk about the probability of a certain alleged historical event.
    Either it happened, in which case the probability = 1, or it did not, and then the probability = 0.
    There are perhaps six different kinds of probability*, and the McGrews could have made some attempt to talk about the theory behind it, what kind of probability they were using, and how the probability of an alleged event can be between 0 and 1.
    2) The values of probability are just taken from thin air with no justification.
    3) The assumption that the gospel and Acts accounts are reliable is a big one.
    Without this assumption, the McGrews have nothing.
    Personally I think that the author of Luke and Acts makes a sincere attempt at historical accuracy, but I am not sure about the rest.
    Matthew’s account of the saints’ resurrection seems fanciful, and the story about the guards looks like it was made up post hoc.
    John 21 appears to have been inserted after the composition of the rest of the gospel.
    Likewise Mark 16.
    Mark 15 has no resurrection appearances.
    If we discount Mark 16 and John 21, and consider Matthew’s resurrection account unreliable, we are left with just Luke 24 and John 20 as strands of eyewitness testimony.
    4) The inclusion of Paul’s testimony is irrelevant as his vision is not of the resurrected Jesus, but, being post Ascension, is just a vision of Jesus like anyone might have had in the last 2000 years.
    5) The assumption of independence is crucial, as the McGrews themselves acknowledge, and Blais is right to point to this. However the arguments the McGrews use to get round the obvious problem, that the disciples’ testimonies are clearly not independent of each other, do not seem valid to me.
    It all depends what alternative naturalistic models are being compared with the resurrection hypothesis.
    If the disciple genuinely believed they had seen the risen Jesus, but were mistaken somehow, and supported each other in that belief, then there is no independence and the McGrews’ points are not valid (see next).
    The crucial inequality only applies if, under not R, the disciples were consciously aware of not R and colluding – threat of death etc would then work in favour of p(D1 & D2 | R)
    The McGrews consider hallucinations unlikely,
    However a hypothesis of empty tomb (body stolen maybe) + vision to peter + bias to believe following initial reports + socialisation and reinforcement is not unlikely, on any worldview.
    6) The McGrews keep emphasizing that there were 13 witnesses to the resurrected Jesus who were willing to die for that belief.
    But how do they get 13?
    Paul does not count, as he did not see Jesus’ resurrected body.
    I make it 11, as Judas’ replacement Matthias was not included until after the Ascension.
    Furthermore we don’t know from Acts how many of the 13(?) were actually willing to die, or indeed what their testimony was.
    Acts only records Peter and John, but mostly Peter.
    And it does not say that they were willing to die, but I guess we might grant that.
    But 2 is not 11, still less 13.
    This is sloppy from the McGrews.
    In conclusion, the article is full of holes.
    I am puzzled why I thought it was so compelling when I first read it.

    *Types of propbability
    1) Subjective belief (subjective)
    eg P(I will get a grade A in this exam), P(Liverpool will win the Premier League in 2021)
    2) Frequentist empirical based on historical data (objective)
    eg P(rain on a random day in August), P(infant dying of SIDS)
    3) Propensity of things to happen in future (objective)
    eg P(earthquake in San Francisco in 2021), P(solar event next week), P(my car will break down in the next month).
    NB I am not convinced that this corresponds to a feature in the real world.
    4) Probability of truth of alleged historical events (objective)
    eg P(the Allies won WW2), P(Caesar was assassinated on the Ides of March 44 BC), P(Jesus rose from the dead)
    5) Mathematical probability (objective)
    eg P(six on a fair die when rolled), P(random single digit number = 9)
    6) Some alleged contingent fact in the present (objective)
    eg there are aliens visiting this planet; there is a God who created the universe; there are species of insects yet undiscovered by humans.


Leave a Comment

Your email address will not be published.


Share on facebook
Share on twitter
Share on linkedin