Raphael and the leg of Venus: the role of expectations in representation

Raffaello Sanzio, The Triumph of Galatea
from wikipedia

As we said, one of the most common feeling expressed by the first visitors of Pompeii was, generally, disappointment. 

The common idea of a Roman town was probably based on literary rather than archaeological evidence. And because, usually, the texts that are transmitted are the official ones, Roman cities were expected to be large, monumental and majestic. 

Unfortunately, Pompeii is quite far from that model. Buildings are cramped and often brightly coloured. But, probably, the most disappointing bit for the first (and contemporary?) visitors were the frescoes.

One (only apparently unrelated) premise: it is interesting to notice the difference between what those visitors said about Pompeii when they were there and what they said about it years later, when they were relying on their memories more than their experience.

Years after his first unexcited comments, Goethe writes that Pompeian frescoes are so excellent that they could stand next to Raphael’s ones.

First, I don’t think it makes a lot of sense to assess archaeological evidence on its aesthetic or artistic relevance. This approach has caused the destruction of many Pompeian artefacts that were jus considered “not pretty enough”.

Second, if we really want to assess the aesthetic value of Pompeian frescoes (from an art history point of view or as simple observers), it doesn’t make a lot of sense to me to compare two artefacts that are so distant from each other like first century AD Roman frescoes and Italian Renaissance oil paintings.

However, if we do compare them, and assume that Raphael’s canons are a sort of archetype all figurative arts should conform to... then... well... Let’s put together the triumph of Galatea by Raphael and the fresco in the House of the Venus in the Shell. 

Fresco in the House of Venus in the Shell
Pompeii (Reg II, Ins 3, 3)
image credit

The subject is vaguely similar, but the images are very different. And I’m not saying that one is better than the other, just that they are very different and that it seems to me that the artists had very different concerns and goals.

The point is that if you’re an enthusiastic Neoclassic (or proto-neoclassic) artist that goes to Pompeii, eager to see and reproduce images that you imagine to be Raphael-like (following the syllogism: Roman art is perfect, Raphael’s art is perfect, Roman art must look like Raphael’s art) and you see our lovely Venus in the Shell... there is going to be quite a clash between your expectation and reality.

This could explain why some of the drawings of the excavation time are, again, slightly dissimilar from the original.  And we can definitely say it, as in some cases the original artefact survived and we can compare it to its documentation. 

So, let’s add some items to the list of things that affect the representation process. We don’t really know for sure what went through the mind of the first artists copying Pompeian frescoes, but we can speculate about it, especially if we focus on what, in my experience, is the element more often “touched up”: the human body (well, divine and semi-divine as well :-))

Drawing by J-C. Bellicard, 1754

We can imagine the artist spending hours and hours learning how to reproduce human bodies in the «right» way (i. e. as more realistic as they can). So, maybe, they just couldn’t help drawing in the way they have learnt and inadvertently “corrected“ the anatomy in the copy. Another hypothesis is that they are drawing not what they are actually seeing but what they were expecting (and wanting) to see.

We could even hypothesise that the artists didn’t want the public to suspect that they were not good enough at drawing. Maybe they didn’t want anyone to think “what a lousy copy! The artist is so useless that cannot even draw the leg of the Venus properly!”.

Here is an example from Herculaneum. The drawing (from the book Antiquity Recovered ) is a copy of one of the Herculaneum frescoes that are now exhibited in the Naples Museum.  All the body proportions (especially in the case of the Minotaur) look much more “harmonic” than the original. 

Fresco from the Basilica in Herculaneum
Image from the Hermitage website

Actually, in the case of Pompeii and Herculaneum there is at least another major, and slightly bizarre, reason that explains why representations are precious sources of information but have also to be treated with a certain care. For the first years it was not allowed to make copies of the exhibits in the Museum of Portici. But the temptation was too strong and some artists (like,  Bellicard, author of this one), took quick sketches that, probably, refined out of the museum, without the original in front of their eyes. According to Lyons and Reed, this would also explain why there are no bootleg copies of the artefacts positioned close to the entrance: there were always too many guards nearby.

A mix of “cosmetic documentation”, the gap between expectation and reality, and memory-based drawing could also explain why there is so much discrepancy between the documentation of the frescoes of the Ekklesiasterium in the Iseum and the actual fragments exhibited in the Museum of Naples. In spite of the very realistic style (always beware of the realistic style!), and the fact that they were commissioned to top class professionals of the time, the drawings not always match the evidence, sometimes quite dramatically. But that story probably deserves a post on its own.

Orpheus' lost twins

The fresco in the House of Orpheus' garden
Photo from pompeiiinpictures

The House of Orpheus in Pompeii owns its (second) name to a big fresco depicting Orpheus surrounded by animals. The image is now mostly faded, especially the left and bottom areas. 

But with Pompeian artefacts there is always a chance that the object we are interested in has been somehow recorded or documented at the time of the excavations, when many features that are now lost were still visible. They might have been described in verbal accounts, photographed, or copied by an artist.

So, if we look into into libraries and archives, we do find that Orpheus’ fresco has been reproduced by Niccolini in 1854. 

It seems that, luckily, we have found the missing information we were looking for: the left side of the fresco shows a charming illusional garden, probably meant to interact, visually, with the actual one. 

If we were about to produce a digital restoration of our fresco, that would look like a precious source of information. And, it definitely is.

However, if we have a better look at our sources, we find another reproduction of the same painting, this time by Emil Presuhn, in 1878.

The two drawings are supposed to be copies of the same original. But, when you put them next to each other, something doesn’t add up...

Left: Niccolini's reproduction of the Orpheus fresco. Right: Presuhn's one

Although the general structure and theme of the composition is definitely very similar, there are countless differences in small and not-so-small details.

Some animals look completely different: the lion in Niccolini’s drawing become an elephant in Presuhn’s one, the big feline in Niccolini’s is an hippo for Presuhn and so on... The colours are also very different: Niccolini shows a green landscape, while in Presuhn we see a quite dry one, with no water streams. We might hypothesise that Presuhn was reproducing an original that was already further degraded, although there are only about 20 years between the two publications. 

But that wouldn’t explain why, on the contrary, the left part of the fresco, looks much more detailed in Presuhn’s version than in Niccolini’s one. The bottom bit, again, is only apparently similar but very different at a second look.

Which one is the right one? Well, sadly we don’t have the original anymore to say which copy is the most accurate. Possibly, they are both equidistant from the model. Or, may be, each of them has more accurate bits than the other. For sure, they are both subjective representations of the same artefact and both influenced by a long list of variables.

I’m using these two drawings to start my little list of some of the most common factors that impact on representations of cultural heritage. 

First bullet point: skills and tools.

Not everyone is equally good at the technique they are using to represent an artefact. Not all the artist are equally skilled (if we consider «skill» as the ability to reproduce the original as close as possible) as well as not all the 3D modellers have the same familiarity with the software.

Then there is training. Although the two drawings are almost contemporary, a different style is clearly detectable, possibly due to the artists different nationality (i.e. different art schools) or just to idiosyncratic differences. I am particularly fond of the lion face in both the drawings. They look very dissimilar from each other and neither of them very much Roman.

Last, the constrains that derive from the tool. For example, in this case, the very bright colours in Niccolini’s drawing and a certain flatness in both of them are probably due (I reckon) to the printing techniques at the time that didn’t allow much sophistication in colour. Likewise, our representation are influenced by the software and hardware that are available. Or, to be more precise, by the ones we can afford.

PS: I was familiar with Presuhn drawing, and I had seen Niccolini’s one before. But I have realised how different they are only when Drew, one of my supervisors, showed me the page of Gardens of Pompeii, vol 2 where they are published together. Funnily, Jashemski uses both the drawings as a source of information about birds and plants in Pompeii, without a word on their striking dissimilarities.

Robespierre's tears and other stories about the truth of images

W.J. Stillman, Picture of the Parthenon
 Athens, 1870. From Antiquity and Photography
Getty Digital Collection

I have spent the last weeks reading about representation of cultural heritage. I went a little bit into reception studies, touched art and photography history and took some dust off my semiotic background. As you can expect, the issues are many, complex and diverse. Especially when we talk about ancient cultural heritage. But there are a couple of points I want to highlight. 

The first one is that every representation (in any medium) is an interpretation. There is always something of the person that is producing it, the culture they belong to, and the tools that are using. This probably sounds redundant and definitely not new. Unless you are attending a 3D imaging conference. Then, you will probably hear concepts like «perfect copy» and «objective recording» more times that you are comfortable with. 

I thought academics had already agreed many years ago on the fact that photographies do lie, or, better, represent the author’s personal view of reality. Inexplicably, the concept doesn’t seem to apply to new imaging techniques such as photogrammetry or laser scanning. 

The second point is that cultural objects (well, probably all objects but let’s talk about cultural heritage) are not fixed in time. They change and evolve, like living things. It becomes even more apparent when we talk about buildings. They are planned, built, damaged, restored, altered, repurposed, abandoned.

So, actually, when we produce a visualisation of something we should always clarify which moment of the «life» of the object we are actually representing.

To deal with the chronological dimension, some 3D visualisation projects have introduced a timeline. For example, UCLA's Digital Roman Forum and Digital Karnak show how buildings have changed, as singles and in relationship with each other. In Digital Karnak, more specifically, it’s possible to pick a period and see which buildings have been built, modified or destroyed at the time.

I found this approach very informative. However, it kind of assumes that the only segment of the life of an object (or place) we’re interested in is from when it is produced to when it is abandoned or destroyed.

Fragonard, J., 1775 Travellers viewing a skeleton at
Pompeii. (Such "discoveries" were often staged
to please prestigious guests)

I think that what happens to cultural heritage when it is discovered, recorded, exhibited and disseminated is still an interesting matter. Objects keep changing, even after their uncovering. Even when they are closed in a glass cabinet. And I am not only talking about material changes, but also about how changes the way we look at them.

Let’s consider Pompeii. The circumstances of its destruction often make people think that it came out from the earth exactly as we see it now. That it emerged perfectly intact, as, to use Lazer's words, a sort of Sleeping Beauty castle. Which is rather far from the truth. A quick look at pictures of the excavations shows how much the place has been changing in the last 250 years: how much of it has been restored, adjusted, twitched, when not completely staged for the tourist’s gaze.

So, my point is, mainly, that all representations are partial and biased. And using cutting edge techniques doesn’t change it. They can be more or less accurate, but are always subjective. Which is not a bad thing per se, as long as there is a general awareness of the limits of the representation process, and, even more important, when we can compare different interpretations that enrich and complement each other. 

And I’m afraid we, somehow, ended up with multivocality and multi-authorship. Again.

To make up for this, I’m going to tell some (possibly funny) stories connected to representations and interpretations. Many, but not all of them, involving Pompeii. Starting with one of my favorite anecdote about the “truth of images” by father of cinema Sergej Eisenstein:

I cannot resist the pleasure of citing here one montage tour de force of this sort, executed by Boitler. One film bought from Germany was Danton, with Emil Jannings. As released on our screens, this scene was shown: Camille Desmoulins is condemned to the guillotine. Greatly agitated, Danton rushes to Robespierre, who turns aside and slowly wipes away a tear. The sub-title said, approximately, 'In the name of freedom, I had to sacrifice a friend...' Fine. 

But who could have guessed that in the German original, Danton, represented as an idler, a petticoat chaser, a splendid chap and the only positive figure in the midst of evil characters, that this Danton ran to the evil Robespierre and... spat in his face? And that it was his spit that Robespierre wiped from his face with a handkerchief? And that the title indicated Robespierre's hatred of Danton, a hate that in the end of the film motivates the condemnation of Jannings-Danton to the guillotine! 

Two tiny cuts had reversed the entire significance of the scene!

(Eisenstein, S. 1949. Film form: Essays in Film Theory)

What is essential is visible to the eye (if you're in my model)

I hope the ghost of Antoine de Saint-exupery won't haunt me for having played with the wise words of his Fox characters, inverting their meaning. I wasn't suggesting a triumph of shallowness and prejudice, but just introducing a brief premise about my 3D modelling process. 

F. Piranesi's hypothetical restoration of the Iseum (detail)
showing the roof's supporting structure.
From builtindex

I have said that the first phase of my modelling will focus on the present archaeological evidence in situ. I might have called it “what is still there”. The expression is not entirely correct because what I am going to model is actually what we can see of what is there (assuming that we were free to  explore the whole Iseum as much as we like). 

In practice, this means that, for example, I am going to model the Purgatorium's underground space, even though it is not immediately visible and it is not accessible to modern tourists without a permit, but is potentially both visible and accessible.

On the other hand, I won't model what is not visible. To explain ti better, I can use an hypothetical example: let’s imagine that the roof of the Temple had survived. In that case, I would model only the ceiling (which is what an observer could see from inside the Temple) and the external elements such as the tiles or the architrave (which is what an observer could see from the outside of the Temple). But I wouldn’t model the (invisible) supporting structure. 

I am sure that modelling what actually allows a roof to stand (for many centuries!) is of massive interest in the study of architecture and ancient building techniques. 

As I am more interested in how a building was seen and experienced by people (either at as a Roman Temple or as a modern Touristic Attraction), I have decided  to include in my 3D model only what is accessible to the eyes of a human being that is allowed everywhere in the modelled space. 

The matter would be different if the hypothetical roof would be damaged. Then, the internal structure would be visible to visitors and part of their experience (and, therefore, object of my representation).

Everything has (at least) two faces: about walls and normals

At the walls, by S. Bakalovich
from wikipedia

 To explain how I am going to deal with constraints, I will use the simplest, and most common, type of them: walls. Although they often look just as think lines on a plan, walls cannot be represented as flat two dimensional areas. That would contradict our experience and wouldn’t make any sense geometrically, as walls are three dimensional objects with different depth. 

It is possible that at the very beginning of your 3D modelling career you are tempted, quite naively, to simply trace the shape of your walls and extrude them as a solid. In case, the puzzled expression of your supervisor and the question “what is that supposed to be? Minecraft?” will quickly tell you that you’re on the wrong track.

What you are supposed to do (or, at least, what I am going to do) is to build two separate surfaces, one for the interior and one for the exterior, without modelling what is in between (i.e. the actual thickness of the wall).

There are a few reasons why I agree it is a sensible approach.

First, walls (as well as every other constraint) may display different material or decorations on the inside or the outside. So, in the model, they will probably have different textures assigned. 

Also, as we were discussing in the other posts, there is a logical (or ontological, in the linked data meaning of the word!) benefit in it. A wall often delimits two different spaces. But, more specifically, one side of it delimits one space and the other side delimits another space (or spaces). For example, the wall that separates the ekklesisaterium from the sacrarium, is the internal southern wall of the ekklesiasterion on one side, but, on the other side, it is the northern internal wall of the  sacrarium.

It may sound confusing at the beginning, but, after you get used to that, it is a simple yet effective convention that has proven his usefulness in my previous models already. I think it will also suit well the RDF modelling and help dividing meaningfully my digital archive.

If there weren’t enough reason to justify a separate management of the two surfaces of a wall (or other constraint), game and VR engines give us another good one: normals. Surfaces are visible only from one view, either from the interior on the exterior. Even if we had a wall with the same identical texture (let’s say a generic masonry), to allow a 360° virtual view of it, we still would need two surfaces: one looking towards the interior and one looking towards the exterior. This is the only way a character (or a camera) can have a realistic view of a virtual building when it is walking into a space as well as when it is walking around it.

Academic guilty pleasures: inventing words and making lists

What I needed in order to start, was a general name to identify everything that is in model. I know that even such an apparently harmless sentence can hyde insidious philosophical controversies. At this stage of my modelling, what I want to represent (both visually and conceptually) is the archaeological evidence, i.e. what can be still seen (and potentially measured) in situ at Pompeii. Further layers of visualisation that involve a bigger amount of speculation and interpretation, will be analysed (and modelled) later on.

Classification of the Greek Orders
from wikipedia

 The terms that I am using are just draft, at the moment. I hope to find something better in the future or even to discover that someone else has already worked out a more convincing naming system and is willing to share it with me.

The word I have chosen for all the components of the model is «element». I think it is abstract and general enough to be associated with pretty much anything (in any 3D model) and it covers logically both the tangible and intangible domains. 

There are no limits to the number of elements in a model or to their size.

After a certain amount of head-banging against the screen of my computer, I think I have identified three different categories of elements. I have tried them against both the Iseum and the House of Orpheus. It seems to me that they kind of work, and cover all the possibilities in a relatively sensible way. But I wouldn’t be surprised if someone pointed out some discrepancy. 

The three groups are:

1) spaces

2) constraints

3) transitions

To be completely honest, this is probably all I need to assign meaningful names to the elements in my 3D model or to catalogue my digital archive of resources. But I thought it was useful to refine the classification a little bit, thinking of a future (potential) higher complexity. 

(The fact that I actually enjoy making lists has absolutely nothing to do with it)

The classification itself, is very likely to be influenced by my 3D modelling experience and by the way I had been taught to approach ancient buildings (especially Pompeian ones) and their components.

Plaster cast of a Pompeian door.
Photo by Günther Einhorn
via Pompeiiinpictures

I call «spaces» the elements that allow activities to happen within them. Very straight forward examples of spaces are all sorts of rooms but also gardens, portici, courtyards, etc...

«Constraints», on the other hand, are elements in which activities cannot take place, and usually serve as boundaries for spaces. The most common type of constraints are walls, but colonnades, podia and roofs will be considered constraints as well. A space is supposed to be delimited on each of its sides, although not all the constraints are tangible. A change in the mosaic pattern can delimit a space, even if there are no walls to mark the separation. To be precise (and slightly pedantic) the tangible constraints can be divided into permeable and impermeable. To the latter clearly belongs plain walls or high gates. To the former, belong colonnades. Although some constraints (such as colonnades) are physically permeable, I suspect that they were not always considered as such in practice. I can hardly imagine someone being allowed to jump down the Temple of Isis’ podium  through the pronaos’ colonnade. However, I will consider only the material qualities to assign the constraint to one or the other category. All the ones we have mentioned so far are permanent constraints (both tangible and intangible), but we know that Romans had temporary ones as well, such as curtains or removable fences. The existence of temporary constraints implies, I guess, the existence of temporary spaces. As I am currently working only on the present archaeological evidence, temporary constraints or spaces are not object of my attention now. 

Last, «transitions» are those elements that connects spaces and, as such, they do not belong to either of the two elements they connect, but are independent. We will consider two kinds of transitions: the ones that allow physical access from one place to the other (such as thresholds or stairs) and the ones that allow visual access from one place to the other (such as windows). Likewise the other two elements, they can be permanent (a stone staircase) or temporary (a ladder). 

Spaces, constraints and transitions can all have «features». Features is used, as the dictionary suggest, to indicate “a distinctive attribute or aspect of something”. 

Features often pertains to the decoration of a space, or a constrain, or a transition. It is easy to identify constraints features, as they are basically contiguous. For example niches in the walls, engaged columns, mosaic floors. But also transitions’ features are usually easy to spot (windows moulding, or doors). Stand-alone features, that have no physical contiguity with any other element (such as altars or herms), will be considered features of the space in which are situated. So, for example, the altar in front of the Purgatorium in the Iseum is a feature of the cavaedium. 

Circular colonnade in the Temple of Vesta, Rome.
Photo by E. Trutat, Bibliothèque de Toulouse

Any element can be divided into sub-elements, according to the level of granularity that the research requires. The sub-elements can be either areas or constraints. For example, in the Iseum, the element space temple can be divided into sub elements pronaos and cella (spaces) and roof and podium (constraints). The Temple also has quite a few transitions: staircases, thresholds, etc...

The same constraint can delimit more than one space, especially when they are subelements of the same space. In the element Temple, constraint roof delimits both sub-elements space pronaos and cella.

It could be argued (especially after a couple of drinks) that the part of a constraint, let’s say a wall, that delimits a space, is conceptually different from the part of it that delimits a contiguous space (such as the wall shared by three of the private rooms in the Iseum). 

In this case, 3D modelling’s logic has stepped forward helping me in two ways. In the first place, reminding me the practicality (even the “materiality” if I can use this word) of the process. Digital 3D models should be optimised for Real Time engines or any other platform that makes them explorable by the users. Therefore it is crucial to keep the number of polygons under control. The less polygons, the faster is the loading process and smoother the character/camera movement.

Also, a limited number of polygons ensures a better control on the file for both the original author and the potential other researchers that want to build on top of it.

On the other hand, I will deal with the conceptual separation of constraints through their internal and external surfaces.

If what I have just written sounds a bit sibylline, I am going to explain myself better in the next post.

Breaking down reality

Greek philosopher Parmenides
image from wikipedia

Dividing the Iseum in its main components and work out a naming convention that made sense in the 3D modelling context as well as in the RDF one, seemed a sensible starting point to me. 

However, reality is a continuum and every attempt to break it down according to our logic(s), and partly to our needs, is bound to result in long hours of discussion, over-consumption of coffee, and a severe headache. The meaning of words and concepts gets challenged over and over again in endless loops, going inevitably back to pre-socratic questions about «being» and «existing». 

The point is that every division and classification is artificial, no matter how much we are fascinated by The Myth of Perfect Consistency or how much we try to be objective.

Objectivity is, probably, a common self-delusion among ontologists. In my opinion, personal and cultural biases can be partly counter-balanced through collaboration and the expression of multiple hypotheses and interpretations. However, at this stage, I have to work on my own in order to produce a proof of concept. So, the best I can do is to record my methodology, so that I can revise it after having received feedbacks.

About naming convention: I was tempted, at the beginning, to use human readable names. I thought it would have made my 3D model much more accessible and easy to look at for other modellers. I had even thought of using Latin to label well identified areas, and more generic names, always in Latin, for the one that have not been identified yet. After discussing it, I have decided to abandon the idea for two main reasons:

Colour map of the elements of the Iseum

1) it is not always so straightforward to label a space. The fact that I am mainly working on the Iseum can be deceptive under this respect. Roman sacred architecture tends to be quite formalised. So there is little to argue if I want to call the pronaos «pronaos». However, it’s not the case for every building in Pompeii. 

A look at my comparative example, the House of Orpheus, proved to be very useful. There are many spaces that do not have an agreed identification, and different scholars refer to them with different names. Not to mention that the upper story(ies) are entirely hypothetical. 

2) Giving a meaningful label means already to make assumptions about the use of a space. And I am trying very hard not to express interpretations at this level. 

In the end, I have opted for a more abstract convention and I will simply use letters (in the English alphabet) to label the components of my model. 

Through RDF triples, I will then express that, for example, room A is type kitchen according to scholar 1 and that the same room A is type storeroom according to scholar 2.

About identification of elements: In the case of the Iseum, the identification of the main components didn’t seem very difficult to me. The different blocks stands out fairly clearly. Even before making (or agreeing on) assumptions on the use of the spaces, they look pretty much neatly separated from a geometric and structural point of view.

I’m using the first published plan of the Iseum ever published (Saint-non’s one) to illustrate the first division into the main elements. In order to avoid confusion, I have erased from the original map the letter-based naming convention that Saint-Non himself used. Next to the letters, I am listing in the legenda also word labels for the sake of clarity, following a widely accepted convention about the Iseum.

Some of the elements can be divided in sub-elements or grouped in super-elements (the relationship being very easily expressed through properties such as :isPart Of or :hasPart). Some of this relationships are very quickly recognisable looking at the plan. The element C (Temple), for example, can be divided into element H (pronaos) and I (cella). The latter, also contains a sub-element J (cellar). Likewise, space G can be divided in several sub-elements.

Close up of the Temple of Isis

However, buildings are not flat, and spaces are not divided only along an horizontal axis. If we look at a cross-section of the Iseum, then we will notice that the Temple is not only made by pronaos and cella, but also by a podium and a roof. And the Purgatorium has a ground level room as well as an underground one.

Last, we can hypothesise the existence of a second story for at least one of the rooms of the private area, although it isn’t in any plan or blueprint of the Iseum.

So, the whole list of the elements in which I have divided the Iseum is a bit longer than the one showed above, but it is probably not interesting to publish it here.

If reading this post you have started mumbling things like: “do a roof and a room belong to the same category”? “what about stairs?” “what about the altars and the statues?”, I know exactly that feeling and this is why I have spent a few nightly hours searching for some convincing answers (that I will publish soon).