Stork Error Summary

Summary of Errors in the Data in Nine of David Stork’s Papers Proposing Alternatives to “The Hockney-Falco Thesis”

David Hockney presented detailed visual evidence,[1] and together he and I published detailed optical evidence,[2],[3],[4],[5],[6] showing optics were used for producing portions of certain early Renaissance paintings. However, David Stork has claimed in some twenty papers that there are alternate, non-optics explanations. Disturbingly, though, all of his papers rely on false assumptions and erroneous data.[†] Although I have not taken the time to critically examine all of his papers, below I summarize severe problems with the data in nine of them, and provide details elsewhere of one of the nine. Errors in three additional papers are discussed in a manuscript I wrote at the invitation of a journal editor. The interested reader can use the following summary, and a link to images showing Stork’s problematic data, to help them examine his conclusions for themselves.

[†] For one quarter, Spring, of 2007 Stanford University’s Department of Statistics employed Stork as a Lecturer to teach a one-third credit class on “Statistical Theory of Data Acquisition” that met once per week for 50 minutes (STATS 328, Wednesday, 4:15-5:05pm), so his errors listed below cannot be attributed to a lack of understanding of the principles of scientific data.

The summary below is limited to errors in Stork’s data, and does not address any of his mischaracterizations of what Hockney and I actually have proposed, errors of historical fact, misleading statements, faulty logic, etc.

David G. Stork, “Did Georges de la Tour use optical projections while painting Christ in the carpenter’s studio?” SPIE Electronic Imaging, San Jose, January 2005.

Background: Hockney briefly commented (Ref. 1, p. 128) about this painting that “the light source seems to be outside the picture” and that the two figures “were probably painted separately.” Stork presents data in this paper that quantitatively support his conclusion that the single candle within the painting was the sole source of illumination for the two figures, from which he argues that Hockney is wrong about the possible use of optics. However:

Erroneous data: The position of the boy’s knee in the shadow of his leg, where Stork has located one of his five data points, is not apparent in any image of this painting I have been able to locate. Also, the line drawn by Stork that should touch the rear left of the block misses it by a significant amount. Stork’s conclusion that there is a single light source is based on the positions of only five pieces of data, of which he has incorrectly positioned two at locations that support his conclusion.

Relevant data selectively omitted: Stork concludes from the positions of five shadows on the painting that only one light source is present. However, in addition to erroneously locating 40% of his data, he has omitted two-thirds of the other relevant well-defined shadows cast by the light sources. It is easy to show that the data he omitted from the eleven other distinct shadows are explained only if there are at least two light sources.

[Click here for a detailed analysis of the errors with the data Stork presents in this paper]

David G. Stork, “Did Jan van Eyck build the first ‘photocopier’ in 1432?” SPIE Electronic Imaging Color Imaging IX: Processing, Hardcopy, and Applications, R. Eschbach and G. G. Marcu (eds.) pp. 50-56, 2004.

Background: We demonstrated that van Eyck’s smaller drawing and larger painting consist of three large regions, and where the regions overlap with each other it is to sub-mm accuracy when the drawing is enlarged by 54%.2,4,6 The first region is the face, and the other two regions overlap to sub-mm accuracy when the enlarged drawing is sequentially shifted by 4.0 mm and 4.2 mm in directions that are not orthogonal. We argue that such high accuracy within three separate regions indicates the use of an optical projection that was displaced twice in the course of making the optical enlargement. Stork claims in this paper that instead the enlargement was made via a graphical method, with the offsets of the regions due to the artist miscounting the grids on his graph paper. However:

Erroneous data: Stork twice re-registered the two overlapping images in a way that resulted in two data points on an orthogonal grid in a “ratio within 7% of 1:2“, in agreement with his (desired) conclusion that van Eyck copied the images graphically. However, his registration of the images is inaccurate. Correct registration of the two images shows that the shifts are non-orthogonal and in a ratio 1:1.05. This difference in registration is outside the measurement uncertainty. Although the incorrect data Stork presents in this paper are in error in a way that supports his argument of a graphical enlargement, the actual data are inconsistent with Stork’s argument.

[Images showing Stork’s erroneous data in his various papers]

Thomas Ketelsen, Olaf Simon, Ina Reiche, Silke Merchel, and David G. Stork, “Evidence for mechanical (not optical) copying and enlarging in Jan van Eyck’s Portrait of Niccolò Albergati,” Proceedings of the Optical Society of America Annual Meeting 2004.

and

David G. Stork, “Optics and realism in Renaissance art,” Scientific American, 291(6):76-84, December 2004.

Background: We demonstrated that van Eyck’s smaller drawing and larger painting consist of three large regions, and where the regions overlap with each other it is to sub-mm accuracy when the drawing is enlarged by 54%.2,4,6 The first region is the face, and the other two regions overlap to sub-mm accuracy when the enlarged drawing is sequentially shifted by 4.0 mm and 4.2 mm in directions that are not orthogonal. We argue that such high accuracy within three separate regions indicates the use of an optical projection that was displaced twice in the course of making the optical enlargement. Whereas in the paper discussed immediately above Stork claimed graph paper had been used to make the enlargement, here Stork et al. claim the enlargement was made via a mechanical tool (a proportional divider). Stork asked physicist/artist Richard Taylor to copy van Eyck’s drawing using a proportional divider, and they show data from that copy that they say has sub-mm fidelity over the entire enlargement, with a slight disagreement in the region of the ear. However:

Erroneous data: Stork et al. have overlapped Tayor’s drawing onto the original drawing at a scale that is 2.6% too large. The way Stork presents his data in his OSA (Fig. 1) and Scientific American (p. 82) articles makes it appear that there is a slight disagreement in the region of the ear, which he claims indicates copying had been done using a mechanical device. However, we obtained a high resolution file of the drawing from Taylor, and when we overlap the drawing at a 2.6% lower magnification than Stork used for his fit, not only is there is no disagreement in the region of the ear, but there is also significantly improved agreement of the fit over the entire image. The 2.6% incorrect magnification Stork used for his data in these papers is outside the measurement uncertainty. Also, in addition to the data being in error, this demonstration is both irrelevant and misleading, since the evidence of the use of optics in creating the painting is that it is made up of three separate, displaced regions, each of high fidelity, not simply a single enlarged image of the drawing.

[Images showing Stork’s erroneous data in his various papers]

David G. Stork, “Optics and the Old Masters Revisited,” Optics and Photonics News, 15(3), pp. 30-37, March 2004.

Background: We demonstrated that the chandelier in van Eyck’s painting ‘Arnolfini Marriage’ was painted with the aid of an optical projection because of the accuracy of its rendition, given the complexity of the form.2,3,6 Stork claims in this paper we are wrong because the chandelier is not painted with the accuracy exhibited by examples in elementary perspective texts. However:

Erroneous data: Stork uses the agreement of his lines in Figure 6 with elementary textbook examples of perspective to conclude that van Eyck’s painting of a chandelier was not based on an optical projection. However, Stork’s lines differ systematically from ones we find when using an 8″×12″ photograph of the chandelier that was supplied to us by the museum, which is identical to the photograph they had previously supplied to Stork for his measurements. Outside of experimental error, the lines Stork drew are inaccurate in ways that support his argument in this paper.

Unrepresentative selection of data: Although that he has done so is not very clear from the text, for his conclusion that the painting deviates too much from what he incorrectly claims to be the expected symmetry of an optics-based image of a real object, Stork selected a small, 4-arm chandelier that was manufactured in the 20th century to compare with van Eyck’s large, 6-arm, hand-made, 15th century chandelier. The smaller size, simpler symmetry, and modern manufacturing techniques used all make unrepresentative the selection of this chandelier for comparison with the perfection of van Eyck’s.

[Images showing Stork’s erroneous data in his various papers]

Antonio Criminisi and David G. Stork, “Did the great masters use optical projections while painting? Perspective comparison of paintings and photographs of Renaissance chandeliers,” in J. Kittler, M. Petrou and M. S. Nixon (eds.), Proceedings of the 17th International Conference on Pattern Recognition, Volume IV, pp. 645-648, 2004 (IEEE).

and

David G. Stork, “Optics and realism in Renaissance art,” Scientific American, 291(6):76-84, December 2004.

Background: We demonstrated that the chandelier in van Eyck’s painting ‘Arnolfini Marriage’ was painted with the aid of an optical projection because of the accuracy of its rendition, given the complexity of the form.2,3,6 We introduced an analysis scheme3 for analyzing the 3-dimensional chandelier as 2-dimensional segments by individually correcting each of the six arms for perspective and overlapping them. We then pointed out from this overlap certain details of the chandelier that are optics-based, and certain other details that are not optics-based because they would have been easier for the artist to paint by eye alone. Stork claims we are wrong because the chandelier is not drawn with the accuracy exhibited by examples in elementary perspective texts. However:

Relevant data selectively omitted: Stork (and Criminisi) has omitted all of the data from the chandelier that conflict with his interpretation, and only shows the data whose deviations Hockney and I previously had shown, and offered explanations for, a year earlier.3 This can be easily seen from the obvious similarity of Figs. 3a) and 3b) of their IEEE paper to Fig. 6 of our OSA paper,3 and noting they limited their analysis to only the non-optical features that Hockney and I had previously shown. After selectively omitting the data we had previously shown to be optics-based, Stork concludes from the remaining non-optics data that optics had not been used for any portion of the chandelier.

Failure to acknowledge prior work: Stork takes credit in both papers for an analysis scheme that used “rigorous computer-vision algorithms to ‘undo’ the perspective in each arm; we then placed the corrected images atop one another.” This analysis scheme of overlapping the perspective-corrected arms to reveal similarities and differences is identical to the analysis scheme Hockney and I used a year earlier,3 but Stork claims credit for it in both papers.

[Images showing Stork’s erroneous data in his various papers]

David G. Stork, “Were optical projections used in early Renaissance painting? A geometric vision analysis of Jan van Eyck’s ‘Arnolfini portrait’ and Robert Campin’s ‘Mérode Altarpiece’,” SPIE Electronic Imaging, Vision Geometry XII, L. J. Latecki, D. M. Mount and A. Y. Wu (eds), pp. 23-30, 2004.

Background: We demonstrated that the latticework in a portion of Campin’s painting is made up of three segments, corresponding to the depths into the painting where the depth-of-field of his lens would have forced him to refocus. From measurements on this portion of the painting we calculated the diameter and focal length of the lens.3,6 Stork claims that instead the latticework was laid out with a straightedge. However:

Erroneous data: Stork writes that “Notice that every slat that crosses the “break” in the trellis is in fact perfectly straight”, consistent with his desired conclusion that the painting was constructed using a straightedge. However, we previously published3,6 that every slat is kinked by 5–8o where it crosses the breaks in the trellis, in agreement with the use of a lens. The image of the painting that Stork includes in his paper is too small for the reader to see that his data on the linearity of the slats is simply false.

[Images showing Stork’s erroneous data in his various papers]

David G. Stork, “Asymmetry in ‘Lotto carpets’ and implications for Hockney’s optical projection theory,” SPIE Electronic Imaging, San Jose, January, 2005.

and

Christopher W. Tyler and David G. Stork, “Did Lorenzo Lotto use optical projections when painting Husband and wife?,” Proceedings of the Optical Society of America Annual Meeting 2004.

Background: We demonstrated that the perspective errors in an octagonal pattern and in a linear border pattern of the carpet in this painting can be self-consistently calculated to an accuracy of 0.5% from the laws of geometrical optics.3,6 Stork shows a detail of a distorted octagonal pattern from a photograph of an actual carpet and argues that this distortion indicates the features in the painting are simply due to manufacturing defects in the carpet that Lotto used for his subject. However:

Relevant data selectively omitted: Although photographs of at least 100 similar carpets from the period are readily available (e.g. Volkmar Gantzhorn, ‘The Oriental Carpet’, Taschen, 1991), Stork used as his entire data set only one portion of one unrepresentative carpet. Although he only shows a detail in his paper, a photograph of the full carpet he selected shows it to be significantly warped after 600 years of wear and weathering in a way that makes the detail superficially appear to be consistent with his desired conclusions. Even though the obvious warping of this particular carpet makes it an inappropriate choice for analysis, it is very easy to show that if the warped outer border is simply “morphed” back to being rectangular, the distorted octagonal feature he claims supports his conclusion ceases to be distorted.

[Images showing Stork’s erroneous data in his various papers]

  • Errors in three additional papers can be found in a manuscript I wrote at the invitation of the editor of a journal after she become aware of problems with Stork’s publications.

To summarize, Stork’s alternative, non-optics explanations are based on the false assumptions and erroneous data outlined above. In every case, the actual data is consistent with the optics thesis put forward by David Hockney and me.

Publications on this topic by David Hockney and Charles M. Falco

[1] David Hockney, Secret Knowledge: Rediscovering the Lost Techniques of the Old Masters (Viking Studio, 2001).

[2] David Hockney and Charles M. Falco, “Optical Insights into Renaissance Art,” Optics and Photonics News 11, 52 (2000).

[3] David Hockney and Charles M. Falco, “Optics at the Dawn of the Renaissance,” Technical Digest of the Optical Society of America, 87th Annual Meeting (Optical Society of America, 2003).

[4] David Hockney and Charles M. Falco, “The Art of the Science of Renaissance Painting,” Proceedings of the Symposium on ‘Effective Presentation and Interpretation in Museums’ (National Gallery of Ireland, 2004), p. 7–11.

[5] David Hockney and Charles M. Falco, “Optical Instruments and Imaging: The Use of Optics by 15th Century Master Painters,” Proceedings of Photonics Asia 2004: Sensors and Imaging (SPIE, 2005).

[6] David Hockney and Charles M. Falco, “Quantitative Analysis of Qualitative Images,” Proceedings of the IS&T/SPIE Symposium on Electronic Imaging Science & Technology (SPIE, 2005). p. 326.

<< Back to Technical and Historical Support for, and Objections to, Our Thesis