Camera Obscura Project
Note: This is a work in progress. While some of the material that follows may eventually become a blog post for a very general audience, in its current form it's not public. It's intended as an overview of my progress to share with those with specialized knowledge on the topic. Any comments or suggestion are welcome. email@example.com
The Camera Obscura in 18th Century Venice
There can be no doubt the camera obscura was a well-known tool among the view painters in 18th century Venice. Reliable commentators of the period mentioned the device being used by Canaletto, Bellotto, and Guardi. Anton Maria Zanetti, an artist, writer and connoisseur who was part of Canaletto's circle, made a two-panel satirical cartoon featuring the device depicting yet another view painter, Michele Marieschi. Apparently Marieschi couldn't draw well enough to paint his own figures. Though it was not uncommon for a perspectivist to enlist the help of figure a painter to populate their landscapes, perhaps the cartoonist had in mind a higher standard: Canaletto excelled at both perspective and the figure.
The cartoon speaks volumes. A noted Venetian artist and intellectual is ridiculing another artist's reliance on a machine. Furthermore, the drawing
highlights Zanetti's knowledge of the essential elements we've come to know of surviving box-style cameras: a lens, a hinged sun shade, and a bottle of oil(?) to adhere a translucent sheet of paper to a ground glass. There's a pedestal for support too, though one that would be wholly
inconvenient to traipse around the streets and canals of Venice. And most telling, the city scene representing the camera's subject is itself the very
kind of abbreviated, blocky drawing typical of what is produced when tracing a from a camera projection. The lack of figures within that scene also
speaks to the difficulty of tracing moving subjects. Finally, the depiction of the sun's face seems to wink at the necessity of bright light for using such a device. All and all, it's clear that Zanetti was familiar with the particulars of the camera obscura and its inner workings.
But the subject depicted here is Marieschi. What of Canaletto? For him, we only need to turn to Zanetti's younger cousin, Antonio Maria Zanetti (the
younger), who wrote:
"By his example, Canaletto taught the correct way of using the camera obscura; and how to understand the errors that occur in the picture when
the artist follows too closely the lines of the perspective, and even more the aerial perspective, as it appears in the camera itself and does not
know how to modify them where scientific accuracy offends against common sense. Those learned in the art will understand what I mean."
Here again, the speaker knows his subject (but otherwise refrains from ridiculing Venice's greatest view painter). Having now made three trips to Venice with my own home-built machine, I consider myself somewhat "learned in the art“ of using the camera obscura. I believe the errors Zanetti the Younger refers to, is when tilting the camera up, as one must do frequently in Venice to see the tops of churches and the like, vertical plumb lines of buildings converge on the flat plane of the projection. Such convergence may be "scientifically accurate" according the the rules of linear perspective, but it's not what the human eye experiences. The "offense" to common sense, I believe, is since plumb lines of
actual buildings do not converge, why draw them that way? With the exception of the occasional taper of the campanile in St Marks Square, no other
vertical lines converge in a Canaletto drawing or painting (or any other 18th century view painter as far as I know). The mention of "aerial
perspective“ is interesting too. It could refer to the way parts of the projection blur either due to focus issues, or poor quality of the projection in general.
These two sightings alone, at the very least, are enough to place the camera firmly in the hands of 18th century Venetian view painters. The most
convincing aspect of these depictions is the speakers in-depth knowledge of the device's workings — and notably — its many limitations too.
Canaletto and the Camera Obscura
A Canaletto biographer, the late J.G. Links, has noted that almost every modern commentator on Canaletto, sighting 18th century chatter like that of the above assumes the device was indeed a part of his process, but he then adds "to what extent they are justified in doing so is a difficult matter to decide". His reticence to fully accept the tracing method begins with the fact that few of Canaletto's paintings, if any, are in any way scientifically accurate in terms of the topographical outlines: buildings have been moved, towers and chimney's heightened, multiple viewpoints blended into one. To the casual observer, these "inaccuracies" are undetectable, for the our brains are not hard-wired to set off alarm bells when topographic details are off. Just as we can usually recognize an old friend who we've not seen in years despite many changes in their facial features, it follows that we would have the same flexibility in assessing the general layout in town squares we've only briefly visited. Thus, even when we become aware that Canaletto has “moved the furniture around“ (as some have described it) the sensation of ultra-realism of his best
painted works still endures.
But as J.G. Links well knew, it's not the paintings we must look to find evidence of tracing, it's the drawings. And not just any drawings, but a very particular set of drawings found in a sketch book now housed in the Academia. These drawings are not compositional studies, because they are drawn in sections, like panels in a stitched-together panorama. They rarely include moving subjects, such as people, boats or water. They are topographical schematics of the Venice cityscape, made from identifiable vantage points within the city. The drawings (in ink) include notes written in Canaletto‘s hand about building color, names of landmarks, and other details, but there is rarely any indication of light and shade. And to many observers, they have the blocky, mechanical characteristics of tracings. However they were created, there's little doubt that they served as pictorial field notes for what was to become his finished paintings or engravings.
There are also many surviving examples of another style of drawing (above) which clearly represent a later step in Canaletto’s process. These are fully realized designs; finished ink drawings that include figures, boats and any topographical modifications that ended up in his paintings and engravings. Recently, the Queen's Gallery in London mounted a Canaletto exhibition that included many of these interim designs. The curators, having detected carefully ruled pencil lines under theses ink designs, concluded that they could not be camera tracings. This discovery, in turn, has been touted in the press as proof Canaletto did not use a camera obscura. But this assumption is wide of the mark: the designs are the point in process where he's made all the deviations from the "facts" contained in his sketchbook in the interests of composition, balance, clarity and other considerations, so the new findings do not move the needle on the debate at all.
Before we jump into my experiments with the camera obscura in Venice, let's just look at a comparison of the four-panel sketchbook drawings of San Simeone Piccolo with the actual site. As luck would have it on one of my trips there, a vaporetto (waterbus) station, which normally blocks Canaletto's drawing position from across the canal, had been temporarily removed during an extensive renovation. A construction fence was still in the way, but I was able to hold my camera above the fence from where Canaletto's viewpoint.
While there are errors, the second panel from the left is the most precise, but it goes down hill from there with the other three panels (see below). The first panel on the left has pretty good agreement with window heights, but the spaces between the windows are off and the foreshortening of the building looks like it was taken from a separate view point. The two panels on the right, which are the overleaf of the first two panels, have to be resized: about -10% on the second half of the church to -20% on the last panel. This means, had he used a camera obscura, he would have used three different lenses, which makes no sense at all. However, all is not lost, as it's possible he used the camera to takes some measurements for panel number two, and extrapolated those horizontal measurement of the building on the left, then had some difficulty extrapolating measurements when flipping the page over in the last two panels. If this seems like too many apologies in order to cling to the camera theory, that's understandable, but hang in there, it may make a little more sense later on.
I've included another copy of the Canaletto's later design, sized to match the church in the photo. Keep in mind, I resized Canaletto's sketches in order to try to fit them to the actual buildings. Canaletto did not do that, so the building in the last panel is way too large in comparison to the church. He extended the building on the far right, the lantern on top of the church has been reduced in size, and the "errors" regarding the foreshortening of the building on the far left remain.
Above is another look at the sketchbook and the companion guide book, which for the purposed of my investigation, amounted to a choose-your-own adventure book: all the translations were provided, the approximate locations of each set of sketches, and a running commentary, pro and con, on the camera debate among other comments were all included.
Here, I'd like to make a note about lenses and how they would impact on the scale of drawings in Canaletto's sketchbook. We usually can determine the place were Canaletto stood by following the sight lines of the landmarks in his drawings. When you know where he stood, and if you have the size of his drawing (which is provided by the actual-size facsimile), you can determine focal length of the (presumed) lens he used. The companion guide did not go into that level of detail, so using Google maps, I predetermined the focal length of the lens used at one of the more promising sites, the Campo San Giovanni e Paolo, the location of our next experiment. It turned out to be a good match, the projection in my camera scaled to Canaletto's sketchbook drawing pretty closely (see the his sketchbook upper right and my corresponding drawing lower right). In other words, if he traced his sketchbook drawings from projections, the drawings are a record of the focal length lens used. If, like at the San Simeone Piccolo site, you find drawings multiple at scales, you have have a problem, because it would be a really impractical method. I brought just two focal lengths (shown above), which require that I also build an extension bellows for to accommodate the second one.
SS. Giovanni e Paolo, with its the much studied associated drawings from the sketchbook is a good place to start as much of this campo and the surrounding buildings have not changed since the 18th century. The four sketches above, which make up two spreads of the book, are thought to be four separate tracings, with each view obtained by swinging a camera obscura – as one would do with a modern camera – to achieve a panoramic view.
Pertinent to Canaletto's possible viewpoint, the inscription in the upper right of the fourth panel reads: view of Campo SS. Giovanni e Paolo from Ca' Dandolo toward Barbaria dele Tole first part and right.
It would appear "postion A", at the door of the small house above is Ca, Dandolo, the location Canaletto stood to draw (or trace) with the aid of a camera obscura projection. I could not get access to the building, so I hired a boat and took a series of digital photos from that spot to compare with the presumed camera drawings.
The alley just to the right (position B), was the closest I could get to set up to test my camera obscura. I also took digital photos from there to compare with the resulting drawings to check the validity of comparing tracings made with a simple, biconvex lens – as I assume would be the case for an 18th century camera – a modern photograph made from the same position.
We will first look at Canaletto's drawing against a photograph taken from position A.
I'm starting with the second spread (drawings 3 & 4 in the sequence), which I made slightly transparent to compare with the modern photograph underneath. Click and drag the center line back and forth to further study the alignments. The drawing matches the dome, though not the cupola (a replacement?). Most of main contours of the church line up pretty well, as do the horizontal divisions on the plinth supporting Verrocchio's statue of Colleoni. There are discrepancies within the conforming areas, but overall it's a surprisingly good fit given the unknown conditions of Canaletto's camera, his exact position, and the many variables that could occur in shooting the modern photograph.
The row of houses on the far right have some overall perspective agreement, but on closer inspection, it would appear much of the details were drawn with a considerable amount of guesswork (or invention), assuming of course, that the houses have remained mostly unchanged. Canaletto's rendering of the balconies (far right) are miniaturized, though they otherwise follow the design and perspective of the corresponding "real world" balconies pretty well.
It's temping to think Canaletto's dotted line, angling down from the right side of the church to the tip of the present-day chimney on the far right, is the very same feature he was referencing 275 years ago, but that may be just a coincidence. It does align with a chimney in his sketch, but it's one of a very different character.
UPDATE: Since I wrote the above, I've resisted the site twice. Once to attempt a freehand drawing of the church. It's quite massive, and looms high above you, and I found it nearly impossible to size it to the small page. The second time I made an attempt using proportional dividers as a guide and had much better luck, but my accuracy was was not as close as we see here. I would like to see how a more experienced urban sketcher would fare at this location.
For this spread, we can see pretty good conformity to the right of the church door. In fact, as has been noted by others, the two bricked-up arches on the right, and what looks like a crumbling section of the church just under the eave directly above, is a repeat of a detail in the previous spread. That triangular "crumbling" section below the eave curiously reveals a tiny arched window, or niche of some sort. Though indicated with just a few squiggly pen lines here, in the final painting, this bit of rustication is lovingly detailed with individual bricks.
In completing the rest of the church, Canaletto seems to deliberately subvert the natural perspective of the facade, a recession that would be clearly evident to any viewer standing in this spot. It's as if he was visualizing the schematic of the finished painting on the fly. Nonetheless, he has the correct number of arches above the door, carefully noted on the sheet with the number "7". However, ten arches appear in the final painting.
The Scuola di San Marco to the left of the church is not drawn. In that space is a sketch of one of the spires on the top of the church. Some suggest that Canaletto repositioned the camera to grab that reference. I did not attempt that, but given Canaletto's demonstrable drawing skills, with measurements taken from the lower part of the church, he could have estimated the rest.
Canaletto's presumed camera (above left) is a "box style" with an internal mirror, and requires translucent paper (such as oiled paper), which is laid on top of a ground glass. The resulting drawing ends up being reversed left to right and would require that he flip over the paper to correct it, then redraw it in the sketchbook by using a carbon paper-like transfer, pin pricks, or some other means. My design, though of rigid construction is a "tent style" camera with an external mirror on top, very much like Joshua Reynolds' camera shown on the right. With such a design, Canaletto could have drawn directly in an open sketchbook, as is shown in the photo above center (the image you see is an actual projection), or on separate sheets of paper. The mirror on top rotates so the operator has a choice of facing the subject, but then must view the projection upside down (much like a large format view camera). Alternately, with the mirror swiveled 180 degrees, one can view the projection right side up, but the subject would then be behind the operator (as how Reynold's camera is shown).
Above (left) is my camera after further modifications, set up in the alley (position B). My camera is a bit unconventional as I needed it to be as lightweight and compact as possible, but it works perfectly well for the purposes of the experiment. I choose to view my projection upside down so the subject would be in front of me as I drew. On the right is the base of the camera that served as a platform for drawing details by eye that I could not make out in the camera.
Although I had a sketchbook that I could place inside the camera, doing so introduced errors as the pages would bounce up and down a bit, distorting the projection. I chose instead to draw on a flat sheet. The skeletal drawing (above left) is about all I could get from the projection. Although I had much better luck in other locations where I had strong, raking sunlight, it turned out I really only needed the basic outlines you see here. In fact, I erased much of the drawing on the left, leaving only faint lines, because I knew I would need to correct the slight taper of vertical lines resulting from the slight upward tilt of the camera. The drawing on the right is the result after I filled in details by eye. I had some difficulty with the horse, but I was very surprised how little I needed of the initial tracing to to achieve accuracy with most of the other internal elements by eye.
Here you can scroll back and forth between my camera drawing, and my camera-assisted freehand drawing overlaid on a photo I took by holding a digital camera in front of the camera obscura. Even though I had the advantage of control over all the variables in this comparison, there are still some errors due to bumping the camera while drawing. It's sometimes hard to notice misalignments because pencil lines disappear among the lights and darks of the projection.
The Canaletto historian, J. G. Links, stated that an 18th century lens would project "just what the human eye normally saw from the same viewpoint, no more, and not much less". With my simple, bi-convex 50cm lens, the projection was definitely less than what could be seen by eye. I could not, for example, make out that there was an inscription on the front of the plinth, let alone read it, but once out of the camera I could see it pretty well. Thus confronted with new details, I wanted to draw them as precisely as possible. Canaletto, on the other hand, seemed to be moving quickly, using lots of rapid lines and a kind of short hand notation for indicating such things as the capitols of columns. Where I fussed over the horse with a lot of erasing and redrawing, he got the full force of the horse's attitude with just a few pen strokes. (Note: his horse is seen more from the side due to his position a meter or so to the right).
As far as I know, there is no surviving interim design, so I've joined all 4 of Canaletto's sketches together (overlapping the "repeat" sections of the second and third panel in the sequence) to show how his "fact based" sketches differ from his idealized vision for the final painting. I've sized the sketch to the church facade, which Canaletto has chosen to show frontally, without perspective. Yet, the windows, doors, and arched insets all retain the oblique perspective of his sketches. Likewise with the South side of the church itself, which would not otherwise be visible if the church is viewed from the front. The Scoula di San Marco on the left, is an idealized perspective: one that would be plausible from a frontal viewing of the church, but in fact would be unattainable with a camera from any position. Thus, like in so many of his paintings, he's representing two view points as one.
If that were not enough, as Links pointed out, Canaletto "perversely" chose to show the sun shining from the north, an occurrence that could have never have happened. This means every highlight and shadow in this picture is an invention. He would have had to extrapolate how light would interact with every stone, every column and each individual brick. Links has also said the shadow on the front of the church is unconvincing. I agree. It should take on the irregularity of the San Marco facade. But I do love the way he handled the illuminated shadows on that facade, as if it was washed with a soft light bouncing off the sunlit church.
Clearly, if Canaletto used a camera, it was an ingredient, not the recipe. The most convincing evidence within the narrow scope of this experiment is sketch #3 in the sequence. There's an intriguing possibility he indeed used a box style camera. After tracing on the transparent medium, he could have then flipped over the sheet, transferred some basic contours to the third panel, do the same for the "repeat" on the second, then extrapolated all the rest. However, as conforming to the actual site as that segment is in the sketch, he ultimately made changes no camera could accomplish. The statue is taller and bolder. The arched windows that flank it have been reduced so as not to compete. The dome has been enlarged, elongated, and moved. The south side of church has been compressed. Finally, comparing the painting to the photograph, you can see how his manipulation of perspective of the scoula on the left and the reworked bank of houses on the right, he literally framed the scene. The often used analogy of a theater set could not be more apt. We get the best seats in the house – an orchestra and a box seat.
A provisional conclusion: If Canaletto used the camera obscura as his contemporaries claimed, it only got him so far: a few key marks to get the general outline of a tricky perspective, the rest filled in by eye, and further extrapolated from there to neighboring buildings. Recording only light marks would allow him to easily subvert that which "offends" the eye, such as the converging lines that occur when pointing the camera up, or angling the camera obliquely toward the facade of building (Canaletto favored one-point perspective such as in the example of San Giovanni e Paolo above). Once his sketches had been transferred to the design or painting,
Coming soon: an analysis of the region around S. Simeone Piccolo