Sunteți pe pagina 1din 8

On the Mathematics of the Perspective of the Arnolfini Portrait and Similar Works of Jan van Eyck Author(s): John

L. Ward Source: The Art Bulletin, Vol. 65, No. 4 (Dec., 1983), pp. 680-686 Published by: College Art Association Stable URL: http://www.jstor.org/stable/3050378 . Accessed: 25/01/2011 19:26
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=caa. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

College Art Association is collaborating with JSTOR to digitize, preserve and extend access to The Art Bulletin.

http://www.jstor.org

680

THE ART BULLETIN DECEMBER 1983 VOLUME LXV NUMBER 4

savagery, and the "golden" style of Claude to depict the coming of civilization? In short, here, as in many other well-documented instances, Bingham seems to have tried to utilize the principles of high art, as he had learned them from art manuals and other sources, to immortalize the development of the Western frontier. Museum of Art, Carnegie Institute Pittsburgh, PA 15213

Discussion
On the Mathematics of the Perspectiveof the
Arnolfini Portrait and Similar Works of Jan van

Eyck
John L. Ward
The arguments and evidence presented by David L. Carleton in his note, "A Mathematical Analysis of the Perspective of the Arnolfini Portrait and Other Similar Interior Scenes by Jan van Eyck," in the March, 1982, issue are inadequate to support any of what I take to be his major claims. These are as follows: (1) Van Eyck's paintings, the Annunciation of the Ghent Altarpiece, the Dresden Triptych, the Ince Hall Madonna, the Rolin Madonna, the Arnolfini Portrait,1the Madonna of Canon van der Paele, and the Lucca Madonna, employ a perspective with "two vanishing areas centered at the two foci of an ellipse,"2 which is therefore referred to by Carleton as elliptical perspective. (2) The Arnolfini Portrait has the basic optical effects of a convex mirror, which were derived from one such as that on the rear wall of the painting.3 (3) "Jan's convex mirror was also present at the time of the execution of his other interior scenes" and its use "led to his

development of a consistent application of a mathematical theory of perspective, best called elliptical perspective."4 To substantiate his first point, Carleton presents perspective drawings of the seven Van Eyck paintings discussed. Five of these drawings reverse the layout of the paintings without explanation or apparent purpose. On the basis of these drawings, Carleton concludes that each picture has a perspective with two central vanishing areas, that these are both lowered in each subsequent picture, until the last one, the Dresden Triptych, returns to an earlier, less monumental form, and that the two vanishing areas and their sequential lowering "lead to the conclusion that Jan probably did have a mathematical theory of perspective, and that he consistently applied and developed this theory."" Carleton insists on the presence of only two vanishing areas for each of the pictures that he discusses instead of the three mentioned by G. Ten Doesschate or the four mentioned by Panofsky for the Arnolfini Portrait.6 However, the imprecision of Carleton's drawings and the omission of certain orthogonals greatly exaggerate the consistency of the convergence into two vanishing areas. To be sure, the floor and ceiling converge in two precise vanishing areas in the Arnolfini Portrait and in more approximate ones in the Van der Paele Madonna and the Dresden Triptych. But the Rolin Madonna has two vanishing areas for the floor alone, the upper one of which is slightly higher than the vanishing area for the orthogonals of the upper wall; the Ghent Altarpiece Annunciation has only one coherent vanishing area, that of the floor orthogonals,' and the Ince Hall and Lucca Madonnas are quite inconsistent in perspective.8 The Lucca Madonna is perhaps most instructive, since Carleton follows Panofsky in giving it a late date and one might expect it to be one of the clearest examples of a fully developed, mathematically consistent, "elliptical" perspective. My perspective drawing of the Lucca Madonna (Fig. 1) shows that the floor does not converge accurately to a single area (if the receding lines of the rug were projected, the disparity would be much greater), in contrast to the floor of the Arnolfini Portrait and that of the earliest work analyzed, the Ghent Altarpiece Annunciation,

II have retained the familiar title of the painting for convenience, study of its iconography("De matrimonio althoughPeterSchabacker's Reconcontracto:Jan van Eyck's'Arnolfini'Portrait ad morganaticum
sidered," Art Quarterly, xxxv, 1972, 375-98) reopens the question of the

identities. subjects' of the the Perspective Arnolfini Portraitand Other SimilarInteriorScenesby Janvan Eyck,"
2 David L. Carleton, "A Mathematical Analysisof

observesthat,with arenot citedby Carleton, on Van Eyck'sperspective respect to the two central panels, "von fiinfzehn Linien sich nicht weniger als vierzehngenau in einem Punkte schneiden"("Perspektive bei Jan van Eyck," Repertorium fiir Kunstund Bildarchitektur wissenschaft, xxxv, 1912, 28. For a perspectivedrawing of the outer left panel, see fig. 19 in my article,"Hidden Symbolismin Jan van
Eyck's Annunciations," Art Bulletin, LVII,1975, 196-220). By contrast,

Art Bulletin, LXIV, 1982, 119.


3
4 5 6

Ibid., 123-24. Ibid., 124. Ibid., 121.

Ibid., 119. See Panofsky, Early Netherlandish Painting, Cambridge, Mass., 1953, I, 203, and G. Ten Doesschate, Perspective: Fundamentals, Controversials, History, Niewkoop, 1964, 139f. 7 Of course the orthogonals of the beams at the juncture of the side walls

and ceiling in the two outer panels will cross somewhere.However,in sharp contrastto the consistencywith which the floor orthogonalsare of eachbeamcross each otherwell before plotted,the threeorthogonals they meet any of the orthogonalsfrom the opposite beam. A careful is constructed deliberately analysisclearlyshows thatthe floorprojection studies and carefullyas a single system. G.J. Kern,whose fundamental

of the two beams no attemptwhateverwas madeto join the orthogonals in a single vanishingarea.Carleton's drawing,on the other perspective hand, seems to imply that VanEyckorganizedhis spaceby plottingthe convergence,on each side of the picture, of one of the many floor orthogonalswith an orthogonalof one of the beams.Such a procedure but also the strictaccuracy of the floor convergence, not only disregards implies that the artist would be more concerned over pictorial not evident to a viewerthanover those that would be disrelationships was inaccurate. evidentif the convergence of orthogonals turbingly 8 The Ince Hall Madonna,presentlyowned by the NationalGalleryof is now universally Victoria,Melbourne, by scholarsas not by recognized
Van Eyck (see U. Hoff and M. Davies, The National Gallery of Victoria, Melbourne, Les primitifs flamands, I, Corpus de la peinture des anciens Pays-Bas meridionaux au quinzieme sikcle, xII, Brussels, 1971, 29-50). Although I believe it to be a good copy after a lost Van Eyck, any conclu-

made. mustbe cautiously sionswith respectto the perspective

DISCUSSION

681

'

1 Lucca Madonna, perspective drawing which do.9 Figure 1 also shows that the sides of the throne base meet at a point immediately next to the vanishing area of the descending orthogonals of the cloth of honor,1o and that the orthogonals along the walls are entirely independent of these areas of convergence. These deviations from a two-point central perspective construction seriously weaken the claim that a mathematical system of perspective has been used, or any method based on two areas of convergence. Generally speaking, in Van Eyck's pictures the more the accuracy of the convergence of orthogonals can be visually estimated (because of length, uninterrupted visibility, number, proximity, and situation within the same plane), the more care Van Eyck takes to converge his orthogonals accurately. Hence the strict accuracy of the floors of the Arnolfini Portrait, the Ghent Altarpiece Annunciation, and the central section of the Rolin Madonna in contrast to the floors of the Ince Hall and Lucca Madonnas and the side sections of the Rolin Madonna's floor. Hence, too, the lesser accuracy, in the pictures analyzed by Carleton, of the orthogonals to the sides of the picture, which are 9 For a perspectivedrawingand analysis of the Arnolfini Portrait,see
Kern (as in note 7), 29-30 and fig. 1, or Die Grundziige der linearperpektivischen Darstellung in der Kunst der Gebriider van Eyck und ihre Schule, Leipzig, 1904, pl. Iv. For the Ghent Altarpiece Annunciation, see note 7. 10In the Van der Paele Madonna, there is also a separate area of convergence for the orthogonals of the throne base, but it is lower than that of the floor. 11I assume that these are the deleted works Carleton describes as "largescale paintings with church interiors" (p. 119) simply because they are the only other paintings with church interiors by Van Eyck, except for

either non-existent or, as in the Arnolfini Portrait and the Lucca Madonna, shorter, fewer, more scattered, or otherwise less in need of treatment as a single system. On the other hand, where orthogonals at the side of the picture are more numerous and longer, and they lie within an uninterrupted plane, as in the Annunciation in the National Gallery, Washington, or the very early Madonna in a Church in Berlin, they converge to a much more accurate vanishing area. These two works, deleted from Carleton's discussion "as being considerably different in their spatial objectives,"" only clarify, by means of the greater number of side orthogonals, a tendency that is also evident in the pictures he does discuss, namely, for the orthogonals that belong to a side wall or vertical receding plane to cross the vertical axis passing through the vanishing area of the floor or ceiling orthogonals before they meet. Sometimes, as in Carleton's drawing of the Arnolfini Portrait,12 the selective diagramming of orthogonals will make ambiguous whether the side orthogonals meet in their own vanishing areas or can be divided between the floor and ceiling systems. But my drawing of the Lucca Madonna's perspective clearly shows that the side orthogonals do not belong to the same systems as do the floor or the cloth of honor. Carleton's proposal goes beyond the claim that Van Eyck used two vanishing areas, which he regards as "Jan's heritage from the Boucicault Master."'3 In his view, Van Eyck's hitherto unrecognized innovation lies in the positioning of these vanishing areas "at the two foci of an ellipse.""4 What can that mean? Where is the ellipse of which he speaks? Perhaps Carleton's description of Van Eyck's perspective as elliptical is meant as a figure of speech to suggest that, just as an ellipse is a circle stretched by separating its center into two foci, so the spaces in Van Eyck's pictures are correspondingly stretched along an axis between the vanishing areas. For example, he argues that the interior of a cube drawn in "elliptical perspective" would be considerably narrower across its back surface than if drawn in onepoint perspective. From this he concludes that "there would have been an overall approximate 43% increase in the area of the rear wall [of the Arnolfini Portrait] if Jan had used single vanishing point perspective."'5 In fact, however, nothing prohibits the redrawing of the picture in one-point perspective while retaining the proportions of the back wall. I have done so in Figure 2. If this drawing is taken as an accurate representation of the room that Van Eyck meant to depict, his treatment of the perspective has the result of flaring the receding vertical planes outward and reducing the tilt of the floor and ceiling. This explanation of Van Eyck's perspective procedure has the advantage of assuming much less deviation from an accurate representation of the original space (if there was one) than does Carleton's. Having shown to his satisfaction that Jan consistently used a mathematical perspective, Carleton next proposes to discover its sources. He begins with John White's observation of curvature in the Arnolfini Portrait floor.16 Although Carleton admits that
the porch-like space of the Maelbeke Triptych, completed after Van Eyck's death. Given the dimensions of these pictures, however (36V2x 143/8" and 12?4x 521" respectively), I must assume that Carleton meant to refer to the churches' scale. 12Carleton, fig. 8.
13 Ibid.,119. 14 Ibid.

Ibid.,120. 15i
16

John White, The Birth and Rebirth of Pictorial Space, London, 1957, 234.

682

THE ART BULLETINDECEMBER1983 VOLUMELXVNUMBER 4

there are no visible curved transverse or orthogonal cracks or any other visible evidence of spatial curvature, he argues that "assuming, a priori, that this curvature exists, we must account for its presence."'7 An attempt to explain the presence of curvature the very existence of which has not been demonstrated, and that is admittedly denied by the only visible transverse line on the floor, does not seem to be a promising undertaking. However, I will offer a suggestion below of why the picture in general may give an impression of curvature. Carleton suggests that the artist's use of a convex mirror supplies the explanation for the intuited floor curvature. To demonstrate the effects of convex curvature on the layout of the floor, Carleton constructed a model on a 1:10 scale that he photographed from in front and again through a hole in the back wall (in the position of the mirror in Van Eyck's picture). The latter photograph was taken of the scene reflected in a six-inch convex mirror placed at the front end. Carleton compares the two pictures and argues that the one taken in the mirror more closely resembles Van Eyck's painting. However, these conclusions are not justified by the evidence that he presents. Carleton begins his analysis by arguing that the presence of curvature in the floor photographed in the convex mirror duplicates an optical effect of the Arnolfini Portrait. Yet the evidence that he points to - the curvature of the transversal in the center of the floor - is unlike the Arnolfini Portrait, where, as Carleton acknowledges,18 the transversal has no visible curvature. It should also be noted that Carleton's photograph of the convex mirror reflection introduces a curvature of all straight lines except orthogonals and that none of the lines that appear curved in the photograph are curved in the painting. Carleton also cites as evidence of Jan's use of a convex mirror the fact that in the frontal photograph both upper corners of the background chair are visible, whereas in the mirrored photograph and in the Van Eyck painting "the upper right-hand corner of this chair disappears behind the head of Jeanne Cenami."19 However, the difference in overlap arises primarily from the use of a different projection point for the perspective (determined by the position of the camera in the frontal photograph and by its reflected, rather than actual, position in the photograph taken from the mirror). Specifically, the frontal photograph is taken with the camera placed just beneath the level of the crossbar of the window and to the right of the picture's center, while the reflected image has its projection point opposite the hole in the back wall. The distance of the projection point also seems significantly different, with the frontal photograph having been taken at a greater distance than the projection point of the reflected photograph. Here a word of explanation is necessary concerning the optical implications of a convex mirror in comparison with those of a flat mirror. A flat mirror creates a reflected space that is symmetrical with real space, so that, if a photograph is taken in it, the representation will be equivalent to what would be visible at a point behind the mirror equal to the distance that the camera is placed in front of it. Without changing the mirror's position, dramatic changes in the overlapping of background objects can be created by small shifts in the camera's position. The curvature of a convex mirror means, on the other hand, that the view in a photograph taken in such a mirror will be equivalent to what would be visible from a point within the imaginary sphere 17Carleton,121. The only evidenceof curvature in theflooringthatI can discoveris a barelyperceptible leftwardcurve of the centralorthogonal that extends from the dog to the shoes. However, this curve offers no supportto Carleton'stheory that the picturewas paintedfroma convex

2 Arnolfini Portrait, tracing from a projected slide with orthogonals redrawn so as to converge in the Passion scene painted directly beneath the convex mirror

created by extending the curved mirror surface until it meets itself. With a fairly small convex mirror, the surprising result is that wherever a viewer moves there is little change in the overlapping of forms by one another (except for the viewer's own reflected overlap). This is because the viewer's visual position is in effect confined to the space within the sphere. As the size of the sphere increases (or the curvature of the mirror lessens), changes in the position of the viewer or camera will produce greater changes in the perceived position of things. One of the consequences of compressing the space reflected in a convex mirror within a smaller compass than that within a flat mirror is that the viewer's spatial position (i.e., the projection point of a photograph taken from this position) is brought forward. Had the projection points of Carleton's two photographs been identical, the degree of overlapping would also have been identical. Or, rather, this would be true if the figures had been kept in the same place. Their altered positions, clearly evident from their relation to the orthogonals and transverse crack, also contribute to the difference in overlap. The most confusing of Carleton's arguments for Van Eyck's use of a convex mirror arises from the extensive floor space in his mirrorsince in convex mirrors reflectas straightlines. This orthogonals fact is demonstrated by Carleton's fig. 15.
18Carleton, 121 19Ibid., 123

DISCUSSION

683

reflected photograph which, contrary to visible evidence, he takes to resemble the painting's foreground space. He writes: Now the floor spaces of the painting and Figure 4 [I assume that Figure 15, the reflected photograph, is meant, since Figure 4 is a perspective analysis of the Ince Hall Madonna] are much greater than the ceiling spaces. Hence, if a flat plane of glass were connected to the foremost edge of the rafters and to the foremost edge of the floors of both the painting and Figure 4 [sic], then the two figures would have to lean back on their heels to avoid being severed ... Thus a convex surface serves as the picture window through which we view the couple in the painting and in Figure 15, a surface not unlike that of the convex mirror on the wall of the painting.20 This reasoning is apparently based on the assumption that the rafters meet the picture plane where they touch the top of the painting and that, since they cannot reach to the front of the space, the picture plane leans into the picture to meet them. But nothing requires that the rafters meet the picture plane at the top of the picture. If the picture frame were, for example, intended to suggest a doorway such as the one reflected in the convex mirror, then the beams would have to extend well above the top edge of the panel before they reached forward to the picture plane, since the top of the door seen in the mirror appears to be more than three feet below the ceiling. Multiple horizontal and vertical lines in the picture confirm that it is not a view through a curved mirror, much less one that is tipped. Early documents indicate that in its original state the painting was covered by shutters painted to imitate speckled marble.21These shutters would have emphasized the flatness of the picture plane and the rectilinearity of the interior space. After some further observations, Carleton concludes by noting that the Arnolfini Portrait "has the basic optical effects associated with the wide-angle lens ... There is the tendency of the figures to protrude from the picture space, the effect of tremendous depth in a narrow space, and the compacting of the background. Also, there is the presence of curvature and the radical divergence of both upper and lower orthogonals."22 The same effects, he asserts, are visible in convex mirrors, and Jan's use of one in the construction of the Arnolfini Portrait "led to his development of a consistent application of a mathematical theory of perspective, best called elliptical perspective."23What connection there is between Jan's presumed use of a mirror and his development of an elliptical perspective system is never explained. Carleton's photograph in a convex mirror, of course, produces a perspective with a single vanishing point for the orthogonals. Even a mirror of elliptical curvature, which did not exist at the time, would produce a single vanishing point for orthogonals. Still, if Carleton's argument is unpersuasive, it may yet be useful to ask whether any of the properties of convex mirror reflections do in fact appear in the Arnolfini Portrait and, if so, why. Most evident is what Carleton calls "the tendency of the figures to protrude from the picture space." A possible cause of this effect not noted by him may be the tilting of the man's vertical axis to the left (depending on how his weight is distributed

on his legs, which is somewhat ambiguous), which produces a hint of perspectival convergence downward toward a second vanishing area (that is, one in a different direction). Since the bride's feet cannot be seen, her participation in this effect is even less clear. The effect of tipping forward is also suggested by the perspective of the pattens in the lower left corner, which appear to sit on a floor that tips drastically forward. The combination of the apparently tilted axis and tipped floor do indeed create a sense that the man's head is closer to the picture than his feet are. However, the tendency of the floor to tip and the figures to loom forward is not exclusively explainable by the use of a curved mirror or a wide-angle lens, which Carleton also compares with the Arnolfini Portrait in its optical effects.24 Comparable effects can be observed in the pictures of contemporary realist Philip Pearlstein, who remarked in a recent interview, "I seem to do a lot of what a wide-angle camera lens does. But that's simply because I work so close to the model."25 It is my belief that the magic of the Arnolfini Portrait space is not "done with mirrors," but instead can be accounted for on the basis of Jan's use of information derived from direct perception of the world. I will not attempt to consider the question of whether the scene ever existed in approximately the form in which it appears in Jan's painting or whether it is partially or completely contrived from his mind. In any case, his space, like that in his other paintings, is based on the scrupulous observation of the surrounding world in the absence of a comprehensive theory of perception and with the use of a very close station point. In order to demonstrate the perspective effects of a close viewing distance, I constructed a model of the scene depicted in the Arnolfini Portrait that measures 11?" (height) by 13" (width) by 18" (depth). The depth does not represent the whole depth of the room; rather, the front edge of the model cuts through the nearest window (the presence of which is visible in the mirror at the back of the painting). This was necessary to allow the camera to be placed in the proper position. It was necessary to make solid figures in order to measure the perspective effect accurately. The model was photographed with a 35mm camera with a 24mm lens at a distance of 6" from the figures and with a vertical visual angle of 700, and agaifi with a 50mm lens at a distance of 19" from the figures and a vertical visual angle of 26V2?. Note, however, that the 24mm lens could have been used for both distances by enlarging the negative until the figures were the size of those in the other photograph and by cropping the picture to match the painting. The only observable difference that the 50mm lens produces is an increase in sharpness, because at an equal distance the 24mm lens takes in a much wider visual angle and consequently requires cropping and enlargement. To put the matter differently, Carleton is correct in stating that a series of optical effects that he observes in the Arnolfini Portrait is "associated with wide-angle lenses"26 - but this common association is based on the mixing together of some very different things. One property - curvature - is in fact an optical property of curvilinear wide-angle lenses. Such lenses are designed to deform all straight lines systematically, except those passing through the picture center, in order to minimize the distortion toward the edges of a wide-angle perspective image when

20 21

Ibid., 122.

23 24 25

Ibid., 124. Ibid., 123-24. S.S. Shaman, "An Interview with Philip Pearlstein," Art in America,

See Martin Davies, The NationalGallery,London,Les primitifsfla-

au mands, I, Corpusde la peinturedes anciens Pays-Basm&ridionaux


quinzieme sikcle, ui, Antwerp, 1954, 125-26.
22 Carleton, 123-24.

LXIX, September, 1981, 121.


26 Carleton, 123.

684

THE ART BULLETINDECEMBER1983 VOLUMELXVNUMBER 4

3 View of reconstruction of Arnolfini Portrait with vertical visual angle of 700 it is viewed away from the (very close) projection point. The wide-angle lens I have used does not produce this curvature. The other optical effects associated with wide-angle lenses mentioned by Carleton are simply effects of natural wide-angle perspective after it has been translated into pictorial perspective. In other words, they are not a function of the lens but of the closeness of the station point or, more exactly, the wideness of the visual angle needed to take in the subject from a close station point and of the spatial deformation created by viewing a close-up picture from too great a distance. Since the effects of wide-angle photography (apart from curvature) are not a function of the lens but of the projection of a wide visual angle perspective onto a flat surface, a close-up photograph of my model of the Arnolfini Portrait should provide an effective means of testing whether spatial properties of the painting arise simply from the proximity of the artist to his subject or whether it is necessary to hypothesize the intervention of a curved mirror. At the same time, if it cannot be assumed that Van Eyck had formulated a comprehensive perspective theory, the absence of any tangible picture plane - such as a mirror - against which to check his empirical observations should lead us to anticipate deviations by the painter from a perfectly systematic perspective. The close-up 27The back wall and the furnishings at the back of the room were originally designed with a slightly greaterviewing distance in mind. would raisethe top of the wall and Bringingthe wall forwardabout 1v/2" the high-backedchair and lower and enlargethe low-backedseat, all of which would improve the accuracyof the correspondence to the paintbut VanEyckhas shifted ing. My carpetis alignedwith the orthogonals, his to blunt the force of the convergence. It shouldalso be remarked that even with the wall moved forward,therewould be 31"betweenthe head of the bed and the back of the wall in my model. Carletonapparently uses the lengthof the bed as a meansof establishing the distancefromthe

4 View of reconstruction of Arnolfini Portrait with a vertical visual angle of 26V2o photograph (Fig. 3), in comparison to the middle-distance photograph (Fig. 4), does in fact reproduce most of the properties singled out by Carleton as indicative of the use of a curved mirror, namely, the tendency of the figures to protrude from the space, the increased effect of depth, the position and angle of intersection of the picture top with the bed canopy, and the relation of the bride to the chair that she overlaps. The upper edge of the photograph does not cut off the top edge of the window, but this is because I constructed the window so that there would be a wall area above it as Van Eyck's mirrored window shows. The fact is, however, that there is an irreconcilable contradiction between the window as shown in Van Eyck's painting and in its painted reflection: it is not possible to construct it as it appears in the picture and still include wall area above it as in the reflection. Carleton achieves his configuration without taking into account the information in the mirror. Had I done likewise, I would have obtained similar results, as can be seen by projecting the window in Figure 3 upward. Numerous discrepancies exist between the photograph of the model and the original painting, but most of them could be rectified, given enough time and patience, by adjusting the size and placement of the objects and surfaces and the location of the camera.27 wall to the bride; however, the result is that, even in his mirrored photograph,the space appearsconsiderablymore shallow than in the painting.Since Van Eyck undoubtedlymeantthe head of the bed to be as being next to the wall (as it appearsin a very similarconunderstood figuration in Petrus Christus's Virgin and Child in Kansas City in Max J. Friedlinder, [reproduced EarlyNetherlandish Painting,I, 1968, Leydenand Brussels, pl. 109]), it seems most likely that, if an actual chamberexistedas VanEyck'smodel,he movedthe bed forwardso that it would form a moreeffective spatialsettingfor the brideby defininga volumethatseemsto enhancehercompactmass.

DISCUSSION

685

The groom's apparent tilt in the painting was replicated by tilting the camera forward. The result is the creation of a second, downward vanishing point for all vertical lines and of a forward inclination to the space. Although this effect may be present in Van Eyck's bridegroom, it does not occur in the lines of his room (as it does in the photograph). Presumably, the tendency to represent vertical edges by vertical lines is so natural and pervasive that, at least for geometric forms, any alternative rarely suggests itself.28 The photograph points out other discrepancies in Van Eyck's perspective: the dog is viewed from a distance corresponding to a position somewhere between the viewing point of my two photographs, as is the patten nearest to the man's foot. But the second patten is even more dramatically tipped than the corresponding one in my close-up photograph, and this also seems true of the groom's closest foot. Moreover, the chandelier is tipped in perspective even more strongly than mine is, even though the perspective of the ceiling beams is much less steep. These discrepancies support my belief that Van Eyck's perspective in the Arnolfini Portrait is based on a collection of direct observations that are extraordinary in their perceptiveness, but not subject to a uniform system or theory. Since there are no curved lines in the Arnolfini Portrait and since the other resemblances between the painting and the reflection of an equivalent scene in a convex mirror have been shown to be the result of the close perspective, Carleton offers no credible evidence for his theory that a mirror was used. Still, the popularity of the theory29suggests that before dismissing it, we ask whether any supporting evidence not cited by Carleton might exist. And, since Carleton acknowledged that "the interpretation of the floor as showing curvature will remain essentially intuitive,""30 we may ask whether any basis for this intuition, other than those already discounted, can be discovered. It is un-

deniable that Jan was fascinated by curved mirrors and reflections; his paintings show that he carefully observed the general effects of curvature on the reflected forms, and the mirror retained for him the full force of medieval symbolism.'3 Furthermore, in the Arnolfini Portrait the forward horizontal edges of the small stand by the window are slightly inclined toward the left and there appears to be a barely perceptible convergence of the masonry joints of the window frame toward the left, so that both surfaces appear slanted slightly out of the picture plane. Similar effects can be found in peripheral objects in other Van Eyck paintings.32 Such slantings away from the picture plane at the edges might conceivably be understood as vestigial traces of an effect noted in a convex mirror. However, when the rules required to produce an accurate perspective picture are not known or are disregarded, the tendency to slant peripheral surfaces away from the picture plane will often result spontaneously from the transferral of the painter's direct observations to the picture surface. John White has studied this phenomenon in the fourteenth and fifteenth centuries33(when, of course, finished paintings were rarely done directly from observation, but may nevertheless have resulted from the synthesis of a multitude of observations), and it may also be observed in pictures by nineteenth-century painters such as Van Gogh34 and C6zanne35 and in pictures by contemporary painters such as Rackstraw Downes36 and Bruno Civitico.37Painter Robert Hansen has written an article proposing the availability of "subjective curvature" to direct perception.38 But whether or not an artist perceives straight lines to be curved, and despite the fact that in an accurately painted perspective picture with a central vanishing point no convergence of transversals or vertical lines is needed, the tendency for such peripheral convergence to occur spontaneously in wide-angle paintings done from direct observation is very great.39

The most notableuse by paintersof non-verticallines to depict verin Renaissance ticalsoccurred and post-Renaissance ceilingpaintings,in which the position of the work createda dichotomybetweena depicted verticaledgeand the actualpositionof the line, whichcouldno longerbe vertical.Photography has createdthe othermajorgroupof picturesthat does not always keep verticallines vertical;this is because,amongother with which a paintertendsto align things, in contrastto the spontaneity his pictureplanewith a verticalaxis (orto compensate for any discrepancies), a photographermust accomplishsuch an alignmentwith great deliberateness.
28

Portrait, therewere a lost Fazioas containing paintingof bathingwomen,described by Bartolomeo a mirrorthatshowed the figuresfrombehind,anda secondlost painting of a singlewomanat her bath with a mirror, knownfromits inclusionin a "picture gallery"paintingby Willemvan Haecht.Therearealsoreflected imageson the armorof Saint Georgein the Vander PaeleMadonna, the Knights of Christpanel of the Ghent Altarpiece,and the Dresden Madonna'sSaint Michael.Van Eyck also made four separateuses of a text, takenfromthe Bookof Wisdom,in whichDivine Wisdom(equated as the "speculumsine maculaDei by him with the Virgin)is described Majestatis."

In additionto the mirror in the Arnolfini 31

29Other scholars who accept the idea that the Arnolfini Portrait derives

from a convex mirror reflection are Decio Gioseffi, "Perspective,"


Encyclopedia of World Art, xi, 203, Elisabeth Dhanens, Hubert and Jan

van Eyck, New York, n.d [1981], 204, and Jean Lejeune,"Jean and MargueriteVan Eyck et le roman des Arnolfini," Commissioncommunale de l'histoire de l'ancien pays de Liege, Documents et memoires,

32This can be observed, for example, in the stand in the Ince Hall Madonna,in the chairof the SaintJeromein Detroit,in the transverse axis of the basin in the LuccaMadonna,and, most dramatically, in the
open chest of the early illumination, The Birth of Saint John the Baptist

fromthe Turin-Milan Hours.

fasc. xi, Liege, 1972; "A proposde Jeanet Marguerite Van Eyck et du
'Roman des Arnolfini,'" Bulletin monumental, cxxxiv, 1976, 239-44. Le-

he believesthat the jeune'stheoryis by all odds the most controversial:


men in the Arnolfini Portrait and in Van Eyck's Man in a Red Turban

and he accountsfor (NationalGallery,London)are both self-portraits, the great discrepancyin their appearances by theorizingthat Van Eyck paintedhimself in the double portraitwith the aid of a convex mirror andsubsequently ob(butdid not use it for the otherpartsof the picture) tained a flat mirrorin which he paintedhimself as the Man in a Red Turban. nose and Lejeunearguesthat this explainsthe moreprominent full mouth in the earlierwork. To demonstrate his theory, he had a sculptorcreatea likeness of the laterportraitwhich was photographed with a curvilinear the appearance of the wide-anglelens to approximate
Arnolfini Portrait head. 30Carleton, 121.

33White (as in note 16). See also Panofsky's classic and controversial study, "Die Perspektive als 'symbolische Form,'" Vortrage der Bibliothek Warburg, 1924-25, Leipzig, 1927, 258-330. 34See J.L. Ward, "A Reexamination of Van Gogh's Pictorial Space," Art Bulletin, LVIII, 1976, 593-604. 35See Norman Turner, "Subjective Curvature in Late Cezanne," Art Bulletin, LXIII, 1981, 665-69. 36 See the painting reproduced in Newsweek, xcix, June 7, 1982, 66. March, 1982, 37See the paintings reproduced in American Artist, XLVI, 43f.
38 R. Hansen, "Hyperbolic LinearPerspective," Journal of Aesthetics and Art Criticism, xxxII, 1973, 147-61. 39 Reasons for this are offered in Ward (as in note 34).

686

THE ART BULLETIN DECEMBER 1983 VOLUME LXV NUMBER 4

In summary, all available evidence points to the conclusion that Van Eyck did not paint the Arnolfini Portrait with the aid of did not have a perspective theory, and did not a convex mirror,40 consistently converge his orthogonals in two vanishing areas positioned at the foci of an ellipse. A careful analysis of the perspective in all of his pictures indicates, first, that he recognized early on that orthogonals in a single plane converge to a point; second, that he only bothered to construct this convergence carefully when its accuracy was visually significant, and, third, that his approach to perspective varied with the expressive requirements of a given picture and of a given area of a picture. Thus, the high position of the vanishing area of the floor orthogonals in the Ghent Altarpiece Annunciation and the Rolin Madonna - Van Eyck's only interiors with exterior views results from the need to lead the viewer into the outside space. By contrast, in the Arnolfini Portrait and the Van der Paele and Lucca Madonnas, the desire to achieve a sense of monumentality and stability while retaining the effect of a closeup perspective leads him to lower the vanishing area for the floor well below the vanishing areas for the orthogonals of the receding top and side planes. This has the effect of reducing the effect of the floor tipping forward, a characteristic of wide-angle pictures when they are viewed from a position further back than the (very close) projection point, 41while preserving the intimacy of the close-up. view. I would like to conclude my discussion of Carleton's study by remarking on its valuable aspects. The idea of constructing a model to test his theory concerning Van Eyck's methods of space construction is an original contribution and inspired me to try my hand at it. Especially valuable is the comparison that his model permits between the reflection in the mirror on the back wall and that in Van Eyck's mirror, since it makes visible the deviations in the painted reflection from the relationships that would have been reflected in an actual mirror in Van Eyck's room. University of Florida Gainesville, FL 32611

Reply David L. Carleton


My hypothesis on "elliptical" perspective in Jan van Eyck's room-size interior scenes has been questioned by Ward in three primary aspects. Ward basically says: (1) Jan did not consistently use two vanishing areas in the construction of these works; (2) the optics of the Arnolfini Portrait show little if any relationship to the optics of a convex mirror; and (3) the convex mirror did not lead Jan to a consistent use of any type of perspective of these interior scenes. Regarding the first aspect, Ward observes correctly that the Arnolfini Portrait, the Van der Paele Madonna and the Dresden Triptych have separate vanishing areas for their upper and lower structures.1 The four works questioned by Ward are the Annunciation in the Ghent Altarpiece, the Rolin Madonna, the Ince Hall Madonna and the Lucca Madonna. He provides a perspective drawing only of the Lucca Madonna, which he says has a random scattering of vanishing points, especially in the area of the floor.2 I will show that Ward's conclusions regarding this work and the first aspect in general are the product of a form of circular reasoning, wherein he derives a random pattern of vanishing points using a random pairing of orthogonals. In order to analyze the perspective of Jan's interior scenes meticulously, agreement first must be reached as to which orthogonals are of primary value and which orthogonals are of secondary value. Ward agrees with this premise, saying that certain orthogonals are: "... shorter, fewer, more scattered, or otherwise less in need of treatment of a single system." I classify Jan's secondary orthogonals as those which were probably not used in the actual structuring of his rooms, which was probably the first thing he did in the construction of these interior scenes. Most of these secondary orthogonals are either short in length and/or placed to the sides of the rooms. Jan generally drew these orthogonals without using vanishing points. It then follows that a primary orthogonal is one that plays a dominant role in determining the structure of the parallelepipedthat defines the space of the room. Examples are corner edges, floor lines, roof lines, and certain window edges. Jan generally uses these orthogonals in symmetrically matching pairs. The Lucca Madonna will suffice to demonstrate both those orthogonals which have primary importance and those which have secondary importance. I have prepared a perspective drawing of the Lucca Madonna by using digital entry of Ward's drawing into a computer which projected the orthogonals in symmetric pairs for their vanishing points (Fig. 1). For the upper portion of the painting, I consider the outer edges of the canopy to be of primary importance. These converge in the vicinity of the knee of the infant. Ward's perspective drawing indicates this. His drawing also indicates that the inner orthogonals of this canopy converge in the same vicinity. Ward has shown this by matching these orthogonals in symmetric pairs, a procedure that he does not use in the lower portion of the painting. These four orthogonals largely define the upper space of the room. The only
reversed several perspective drawings during the editing process of my note. However, this does not have any effect on the hypothesis of elliptical perspective. 2Ward says that I agree with Panofsky on the date of the Lucca Madonna. I did not in fact take a position on its date but instead pointed out the various ramifications of the different dates in contention for the work.
1

Even the representation of the reflection in the depicted mirror seems not to have been done directly from a mirror, in view of the numerous points at which it fails to correspond to the scene it mirrors. The most notable discrepancy in the depicted optics is the forward edge of the stand at the left, which remains resolutely horizontal, whereas the corresponding edge in Carleton's mirror (fig. 18) follows the curve of the mirror.
40

See M. H. Pirenne, Optics, Painting, and Photography, London, 1970, and J. L. Ward, "The Perception of Pictorial Space in Perspective Pictures," Leonardo, ix, 1976, 279-88.
41

S-ar putea să vă placă și