Recently, Eastern Orthodox apologists have worked to develop a new argument for the veneration of icons. This time, they are arguing from images of praying people found in the catacombs.

This argument, which I will call the Participatory Art Argument, starts by pointing out that, in the catacombs, as I mentioned in my archaeology post, there are many *orants*, or praying figures, painted on the walls. These are people who are depicted as standing, with raised hands. Often, their eyes are also raised. This prayer posture was one of the most common (if not the most common) prayer postures in early Christian practice. These orants are thought by archaeologists to represent the souls of dead Christians.

It turns out that a few of the orants have either a Good Shepherd artwork or a Tau-Rho symbol (a symbol that stood for Christ) located above them, such as in the dome of a vaulted catacomb ceiling. This fact provides the Eastern Orthodox apologists with their argument. The argument goes basically like this:

The orants are intended to invite catacomb visitors to imitate their posture and to pray. In imitating the orants, a visitor would pray specifically toward an image of Christ, to which the orant directs them. The aim would be for the visitor to address his or her prayers to the image of Christ that appears there.^{1}See Craig Truglia’s argument and Michael Garten’s argument

Thus, to these apologists, just the placement of these orant and Good Shepherd figures suggests that Christians were intended to pray to an image of Christ when they visited the catacombs.

You may already sense that something is going wrong with this argument. However, due to the nature of the argument, it’s easier to sense that it’s wrong than to pinpoint the exact problem. That’s why, even though I think this argument is a failure from start to finish, I’m taking the time to write an entire response to it.

## Does the argument meet the criteria?

This argument can get confusing very quickly, so, first, let’s take a step back and consider what this argument needs to accomplish. Recall that, in my series on iconodulia, I have argued that, before 313, the veneration of images was not a legitimate Christian practice. We have ample historical and archaeological evidence to demonstrate this point.

So then, what would it take for an example of an image to provide evidence against my argument?

**The image must predate 313.**Otherwise, its irrelevant to this discussion.**Christians must have addressed communications through the image to the subject of the image.**Otherwise, it can’t demonstrate that Eastern Orthodox iconodulia existed.**There must be good reason to think that Christian leaders supported****the veneration.**Otherwise, it can’t demonstrate that the veneration was a legitimate practice.

So, with that in mind, let’s look at the argument. Does it meet these three criteria?

This does pass criterion (1), since the images in question are most likely pre-Constantinian. Truglia and Garten, the apologists who have developed this argument, spend most of their time arguing for why this is an example of iconodulia, and therefore would meet criterion (2). I’ll address their arguments in a moment. If their arguments succeed, then the Participatory Art Argument would also pass (2); if not, this fails to provide evidence for pre-Constantinian iconodulia.

However, even if the argument passes (2), it fails with regard to (3). There’s no evidence that Christian leaders supported the alleged veneration. In my __archaeology post__, I cited a scholar who notes that these artworks were probably not commissioned by a church. These artworks represent a few lay people who were rich enough to hire painters to paint a whole room for their tomb.

As Eastern Orthodox Christians, Truglia and Garten should know that meeting criterion (3) is of paramount importance. Just because a Christian does something, that doesn’t mean that it’s okay—it’s the consensus of church leaders that counts. However, that doesn’t seem to be on their radar.

Thus, this argument fails from the start. However, I’ll still engage with it further, because I also believe that it quite evidently fails criterion (2) as well. There’s no good reason to conclude that any sort of iconodulia is happening with these orants.

## Breaking down the argument

So why do Truglia and Garten believe that these images are examples of iconodulia? Here again is how their argument goes. Let’s number the points for easier understanding:

- Viewers are intended to imitate the orant posture and pray.
- Viewers are intended to pray toward an image of Christ, to which the orant directs them.
- Viewers are intended to pray to the image they are facing, addressing their prayers through it to Christ.

For the argument to work, all three of these points must be true. If even one of them is false, we don’t get the whole way to iconodulia, and the argument doesn’t meet criterion (2).

## What probability is needed for the argument to succeed?

Okay, then. We aren’t aiming for certainty, right? Just likelihood. So all we need is for these three points to be more probably true than not, right?

Unfortunately, no. What we need is for the *conclusion* to be more probably true than not. Let’s say that we’re aiming for 51% certainty that these orants point to iconodulia. In other words, the likelihood that all three of the Participatory Art Argument (PAA) points being true must be at least 51% (though of course that’s setting the bar pretty low).

So what happens if we only require the likelihood of each of the PAA points to be 51% or higher? In that case, we would need to multiply the probability of (1), which is 51%, times the probability of (2), which is 51%, times the probability of (3), which is 51%.

If each of them is 51% certain, the probability that all of them are true is only 13%. In other words, the PAA can completely fail, even if each individual point is more probably true than not.

So how likely does each point of the PAA need to be in order for the conclusion to be 51% likely? It turns out that each point, on average, must be more than 79% likely, in order to conclude that it is 51% likely that these orants point to iconodulia.

If the points are, on average, 80% likely, that will get us to a 51.2% probability for the entire argument. So let’s use 80% for our good round number.

Of course, even if the conclusion is 51.2% likely, that’s hardly a strong argument. A good argument on the other side could easily knock it down. However, for the sake of simplicity, let’s only hold the PAA to a 51.2% likelihood.

So now we just need to examine each of the points of the PAA. Do they appear to be, on average, at least 80% likely?

- Viewers are intended to imitate the orant posture and pray.
- In the case of (1), viewers are intended to pray toward an image of Christ, to which the orant directs them.
- In the case of (1) and (2), viewers are intended to pray to the image they are facing, addressing their prayers through it to Christ.

## The problem of intentions

The first problem the Probability Art Argument (PAA) encounters is that each of its points concerns the intentions of whoever designed the artwork. How can we be 80% certain about someone’s intentions, when we have no idea who they were, and when they left no writings to explain their intentions?

So the artist’s intentions had better be very obvious from the art itself, or the Participatory Art Argument seems unlikely to go anywhere from the start. Now let’s examine each point in turn to see if it is so obvious from the artwork.

## (1) Is it 80% likely that viewers are intended to imitate the orant?

This is the part for which Truglia and Garten have support from contemporary scholarship. It appears that some scholars of early Christian art today are interested in following the speculation that some early Christian art was intended to invite participation. It could easily be that some of the orants, especially prominent ones, were intended in part to encourage visitors to pray.

Before considering the battle won, however, it’s important to note that scholars are putting this forward as speculation, not with any degree of certainty. Not everything an academic says is intended to be taken as certain. In fact, the language Robin Jensen uses doesn’t sound like she is 80% confident:

Vividly rendered, her hands reach up as if animated and she casts her eyes upwards and to her right as if in appeal to a heavenly being. Her relatively large size and central placement make her a commanding presence in this small chamber; she is the first image that catches the visitor’s eye, perhaps meant to model the attitude they should affect as they entered her space.[

mfn]This article[/mfn]

The word “perhaps” is not usually employed for even a likelihood as high as 50%, much less 80%. So why does Jensen, an expert in this subject, use the word “perhaps”?

### Calculating probability

Though Jensen is probably not thinking in philosophical terms, we can use philosophy to understand why she is not convinced that viewers are intended to imitate the image. I’ll be using a method known as Bayes’ Theorem to calculate the probability.

The way Bayesian probability works, when comparing two hypotheses, is that you estimate what the probability of each hypothesis would be without looking at the evidence. Then, you estimate how likely the *evidence* would be, if each hypothesis were true. These numbers go together to show how likely each hypothesis is.

### Jamie’s child

For example, let’s suppose that we know Jamie has a child, but we don’t know whether the child is a boy or a girl. Then we see an American Girl doll in Jamie’s car. How likely is it that Jamie’s child is a girl?

The prior probability, before taking into account the evidence of the doll, is about 50%. After all, Jamie’s child could just as easily be male as female.

Now we need to ask how likely it would be for Jamie to have an American Girl doll in her car if her child is a boy. That’s probably fairly unlikely, although not impossible—the doll might be Jamie’s own, or the doll might be a gift for someone else, perhaps a niece. So the probability might be 25%.

So how likely would it be for Jamie to have an American Girl doll in her car if her child is a girl? Pretty high. For the sake of the example, let’s just say 90%.

Now we plug these numbers into a formula:

**P***E*(*H*) = **P**(*H*)**P***H*(*E*) **/** [**P**(*H*)**P***H*(*E*) + **P**(~*H*)**P**~*H*(*E*)]^{2}https://plato.stanford.edu/entries/bayes-theorem/

P means “probability,” E indicates the evidence in question, H indicates the hypothesis we’re discussing, and ~H indicates the alternative hypothesis.

So our new formula looks like this:

The probability that Jamie has a girl, given the doll = 0.5 * 0.9 / (0.5 * 0.9 + 0.5 * 0.25)

And it turns out that the answer is 78%. Given the American Girl doll, it is 78% likely that Jamie’s child is a girl and 22% likely that her child is a boy.

### Back to PAA (1)

So what are the two options that are being proposed for the intention of the artist?

The other option that Jensen proposes, which is consistent with contemporary scholarship, is as follows:

Those coming to pay their respects on the anniversaries of death, or other traditional occasions for visiting the graves of friends or family members, would be reminded of these Bible stories and be prompted to trust in God’s promised salvation. Yet, it may be simply that all these orant figures more broadly signify the virtue of piety and the importance of engaging in prayer

So let’s compare the following two options:

- The large, prominent orant is a reminder of the importance of prayer.
- The large, prominent orant is a reminder of the importance of prayer and an invitation to model her attitude and pray right there.

It should be immediately obvious that (B) the claim of the PAA is a more extensive one. It proposes a more specific and complex motive for the artist, and therefore is less likely from the start.

So let’s set the prior probability of (A) at 60% and the prior probability of (B) is therefore 40%.

Now, how likely is it that we would see the large, prominent orant, if (A) is true? This would of course be 100%, because if the orant is intended as a reminder of the importance of prayer, it of course must exist!

In the same way, the likelihood of the orant given that (B) is true is also 100%.

So let’s do the math to find the likelihood of (B):

The probability of B given the orant = 0.4 * 1 / (0.4 * 1 + 0.6 * 1)

If you’re good at math, you can probably already see that the answer is 40%. That’s why Jensen uses the word “perhaps.”

So what just happened in the math? The probability of (B) started out as lower than (A), because (B) was a more extensive hypothesis. If it had turned out that the orant was better evidence for (B) than for (A), we might have ended up with a higher probability for (B). But since the orant is equally good evidence for (B) and (A), it didn’t do anything to raise the probability for (B).

So the probability of the first point in the Probability Art Argument doesn’t come anywhere close to 80%.

## (2) Is it 80% likely that viewers are intended to pray toward an image of Christ?

Let’s consider the second point. How likely is it that, if viewers are intended to pray, they are intended to pray toward the Good Shepherd?

If we just look at the example that Truglia gives, where an orant is actually looking upward toward the Good Shepherd, this would seem very likely. However, we also need to consider that there are other cases of prominent orants that are not looking at a Good Shepherd; thus, imitating them would not lead to looking at a Good Shepherd. So this one could be a coincidence—after all, Christ is the center of our faith, so we would expect an image of the Good Shepherd to be in a prominent place wherever it appears, whether or not it would be venerated.

Because this point has multiple pieces of evidence, some of which point one way and some of which point the other way, I’ll just grant (2) a high level of probability: 90%.

## (3) Is it 80% likely that viewers are intended to pray to the image?

Now let’s consider the final point. How likely is it that, if viewers are intended to pray toward the Good Shepherd, they are intended to pray *to* the Good Shepherd?

On the face of it, before examining the relevant evidence, this would seem fairly likely: perhaps 80%. So let’s examine the evidence and see whether that supports or detracts from the hypothesis.

The big question is whether the designers of these artworks accepted the Eastern Orthodox theology of images. Did they think of images as a means of communicating with Christ? Of course, it’s easy for the Eastern Orthodox to assume so, but we need to look at the evidence before just assuming anything.

The most relevant evidence for someone’s intentions is what they themselves say that their intentions are. Of course, we have no record of the intentions of the Christians who designed these artworks, but we do have records from representatives of Christianity during this time period. In my article on the early church fathers and icon veneration, I showed that the Eastern Orthodox philosophies underlying iconodulia did not exist before Constantine. The early Christians consistently believed that to pray to images was to pray merely to the material that made them up, not to the subject in the image.

The likelihood that the early Christians would have been so consistent, if Christians were venerating the Good Shepherd image, is very low, although of course it’s possible that there were lay Christians who were disobedient to the consensus of early Christian leaders. Let’s set the likelihood at 10%.

The likelihood that the early Christians would have been so consistent if Christians were *not* venerating the Good Shepherd image is very high: let’s say 90%.

So our formula is as follows:

The probability that the truth of (1) and (2) indicates veneration of the Good Shepherd = 0.8 * 0.1 / (0.8 * 0.1 + 0.2 * 0.9)

That comes out to 31%.

## Combining the probabilities of the Participatory Art Argument

Let’s review the probabilities we have:

- (40%) Viewers are intended to imitate the orant posture and pray.
- (90%) In the case of (1), viewers are intended to pray toward an image of Christ, to which the orant directs them.
- (31%) In the case of (1) and (2), viewers are intended to pray to the image they are facing, addressing their prayers through it to Christ.

When we multiply these together, we have an overall probability of 11%. In other words, the Participatory Art Argument is a complete failure.

What just happened above should show why, even though this argument fails, it is nonetheless difficult to articulate the reason. There are so many different probabilities that make such a big difference, and it takes a careful assessment to show why the probabilities combine to disprove the PAA.

### The experts

This is precisely why the experts in the field of early Christian art don’t use this argument, although their sense of the evidence is well-developed enough that they don’t need the philosophical rigor that I just provided.

For example, in this very same article cited by Truglia and Garten, Robin Jensen writes,

While representations of popular scripture passages were unlikely to be the focus of devotional prayer or veneration in the way that later portraits of saints, Jesus, or the Virgin Mary came to be, it is incorrect to see them as playing merely didactic or catechetical roles, as if they were simply visual aids to written texts. Rather, the limited repertoire of the images, their spatial context and compositional arrangement, as well as stylistic aspects of their fabrication, all suggest that early Christian artifacts played a role in shaping and reflecting on certain religious practices.[

mfn]This article[/mfn]

In other words, she does not see this consideration as evidence for iconodulia.

## Conclusion

It’s a little sad to know that I’ve spent an entire argument responding to an argument that is so weak. But, if nothing else, I hope any readers who had an intuitive sense that the PAA fails now have the words to explain why. I also hope that this will help model for readers how one can assess the claims of an argument with philosophical rigor.

- 1
- 2

LandonIsn’t this getting Bayesian historical methodology a bit backwards? You have two hypotheses: veneration pre-313 and no veneration pre-313 (call these V and NV for short). You have a prior for each hypothesis (call the priors P(V) and P(NV)). Now you introduce the historical evidence of an orant. Which hypothesis should we prefer after we consider this evidence? You can rearrange Bayes to show that P(V | orant) > P(NV | orant) iff P(orant | V) P(V) > P(orant | NV) P(NV). Now, of course you and the Orthodox have different priors. But to get something prior-independent you can assume the priors are equal (say this is the first piece of evidence we consider), then you have that the orant increases the probability of V over NV iff P(orant | V) > P(orant | NV). In words, which hypothesis does the orant fit better with? All else being equal, does the orant fit better with V or NV – which hypothesis has a higher explanatory potential? If it fits better with V, then that piece of evidence favors V over NV.

In your approach, you aren’t considering the orant independently of other pieces of evidence. In your last step, you take it as a given that “the Eastern Orthodox philosophies underlying iconodulia did not exist before Constantine.” You

mightbe able to argue that this is not “technically” begging the question, but it is very close, and it certainly does violate independence. Aside from this issue, what you have shown is that the probability of any one piece of historical evidenceprovingany moderately complex hypothesis is quite low. It is in fact muchlowerthan 11% in general (for specific evidence-hypothesis combinations it might be higher). This is why considering the explanatory potential is more useful.Of course, once you start independently considering multiple pieces of evidence in this way and guesstimating the likelihood ratios, you could very well come to a

probabilisticpreference for NV over V – I’m not weighing in on the outcome of that entire process (or even on what the “correct” likelihood ratio for this specific piece of evidence is).Lynn MartinHi Landon,

My guess is that you know much more about Bayes than I do, so you’re probably right that I’m stating things in a nonstandard way. My aim here is mainly to give some clarity to the reason why this argument is so weak. There are several assumptions required to make this argument, and I don’t think they work. I’m not intending to approach this particular argument without context, because I think the context makes all the difference.

Perhaps I shouldn’t go into Bayes, but instead state my argument without it.