In this case of contract interpretation we must answer the following riddle: When is a photograph no longer a photograph?

* Jeffrey Mendler is a professional photographer. In August 1991 he signed a licensing agreement with Winterland, a manufacturer of screen-printed apparel. Pursuant to the agreement, Mendler provided Winterland with numerous slides of photographs he had taken of the America's Cup yacht race. Among these was an image titled "San Diego's America's Cup," which depicts the "Espana" overtaking the "Spirit of Australia" in a tacking duel. See Appendix. The license allowed Winterland the use of the photos as "guides, models, and examples, for illustrations to be used on screen printed T-shirts or other sportswear."

By 1992 Winterland had begun marketing T-shirts produced under this agreement. There were at least two different designs, each featuring drawings of two yachts with sails crossed, in a configuration modelled after Mendler's photo of the tacking duel. The drawings on these shirts are clearly identifiable as such, containing just enough detail to convey the desired image.1 After receiving some samples of these T-shirts, Mendler had no further communication with Winterland for several years.

In 1995, Mendler learned that Winterland had put out a new line of America's Cup T-shirts. While depicting the same scene as the earlier series, these shirts were made using a very different technique. Instead of line drawings, the newer shirts display a digitally altered version of the image from Mendler's original photo. See Appendix. Mendler complained toWinterland that this use of his photo was not authorized by the licensing agreement. The ensuing negotiations failed. Mendler registered his photograph with the Register of Copyrights, and brought suit for copyright infringement and related claims against Winterland.2

The case was tried without a jury. The district court held for the defendants on the copyright claim, ruling that "Winterland's use of the slides was within the scope of the license agreements." Mendler appeals3.

Contract interpretation is a question of law we review de novo. See Confederated Tribes of Siletz Indians v. Oregon, 143 F.3d 481, 484 (9th Cir. 1998). While we are wont to defer when a district court relies on extrinsic evidence in interpreting an ambiguous contract, see L.K. Comstock & Co. v. United Eng'rs & Constructors Inc., 880 F.2d 219, 221 (9th Cir. 1989), the district court here made no findings of fact with regard to the copyright claim. We know neither how it interpreted the contract nor on what extrinsic evidence, if any, it relied in concluding that the T-shirt fell "within the scope of the license agreements." Faced with a naked conclusion of law, we have nothing to which to defer.

Nevertheless, our task of interpretation is reduced substantially, because the parties agree, to some extent, about the contract's meaning. Though they dispute what they meant by "illustrations," the parties agree that the contract did not authorize Winterland to use photographic reproductions of Mendler's work. Thus, in order to affirm the district court's ruling, we must conclude that the image on the T-shirt is not a photograph.

Winterland created the T-shirt image by scanning-digitally reproducing--the photo Mendler gave them. It is conceded that a reproduction created in this fashion is photographic.4 The case does not end here, however, because another term of the license gave Winterland the right to use "whatever illustration process" it found most appropriate. Winterland was thus allowed to make a scanned image, so long as it used the image only as a "guide[ ], model[ or] example[ ]" to achieve an end result that was an "illustration" and not a photographic reproduction. The question we must answer, then, is whether Winterland's subsequent electronic modifications transformed the scanned photograph into something that was no longer a photograph.

Winterland, no doubt, made noticeable alterations to the image from Mendler's original photo. The image was flipped horizontally, so that the vessel in the foreground is on the right rather than left. The sail of this craft, cut off by the frame in the original photo, has been extended and its missing tip reconstructed. A smooth background of gently gradated blue has replaced the original sky with its strata of white and grey clouds. Shades of brown have been changed to shades ofviolet, whites to fluorescent blues. At the same time, the tonal range of the whole image has been compressed through posterization5.

Winterland argues that these changes have transformed the image on the T-shirt from a photograph into an illustration based on a photograph. The dissent agrees, asserting that while the T-shirt image is "obviously based on the photograph, it is not the photograph." Infra, at 2040 (Rymer, J., dissenting). The only reason the dissent gives for this conclusion is that "Winterland's manipulation of the photograph was significant." Id. at 2940. But what does "significant" mean? The dissent leaves unexplained why the changes made are such as to destroy the original image's photographic quality. If we are to give our judgment content beyond "I know it when I see it," we must attempt to articulate what kinds of changes are "significant" enough to render an image non-photographic. The contract itself does not address this issue, and neither party argues that the concept "photographic" had any idiosyncratic meaning in the context of their business relationship. It is therefore appropriate to look to common usage and understanding, taking judicial notice of such materials as may aid us. See E. Allan Farnsworth, Contracts S 7.11 (1990) ("When interpreting contract language, courts start with the assumption that the parties have used the language in the way that reasonable persons ordinarily do."). The parties have participated in this inquiry, providing supplemental briefing that addresses such topics as the history of photography and the nature of the technology utilized here.

What distinguishes photography from other visual art forms is that, as the name implies, the light itself does the writing.6 The photographer can compose the shot, but once he triggers the shutter, anything visible to the eye is captured exactly as an observer would see it.7 The reactions of the exposed film, like the workings of one's own retina, are not subject to direct control. This fact gives rise to the two qualities we most associate with photographic images: lifelike appearance and objective accuracy. The former is why we like photographs so much--they're the next best thing to seeing something in the flesh. Indeed, the association is so close that an extremely realistic drawing or painting is often described as "photographic."8 The latter is why we trust photographs-since there's no willful agency intervening between the actual scene and its recording on the film, we tend to regard photographs as infallible and unimpeachable witnesses. We don't always trust the call of the live referee; but no one argues with the photo finish.

Of course, both of these characteristics of photography are subject to important caveats. While objective accuracy is a large part of what we mean by the term "photographic," we have also long been aware that photographs are not always to be trusted. The camera may not lie, but the person who develops the film or prints the image on paper can alter what it tells us.9 Even before the advent of computers, an airbrush or a strategically placedthumb during the printing process could be used to erase a facial blemish or eliminate a purged Bolshevik. Digital technology makes such alteration child's play, and most of the photographs we see in the media today have been digitally tweaked to get the exact image desired.10 Often important elements of the depicted scene are relocated, removed or replaced entirely with borrowed images.11 Even though we are (sometimes) aware that these doctored photographs no longer accurately depict reality,12 we nevertheless perceive and identify the images as photographic13.

Nor does an image have to look perfectly lifelike to be recognized as photographic. This is most obvious with regard to color. A black and white photograph is unquestionably a photograph, even though it is dissimilar in an important respect from what we normally perceive. Similarly, a color negative doesn't lose its photographic quality when a black and white print is developed from it. This obvious truth isn't changed when, instead of removing colors, we change them. From hand-painted daguerreotypes to colorized oldies, we have always regarded images as photographic even though they contained coloring not derived from the original exposure14. Even an image whose overall chromatic appearance radically diverges from reality can be unquestionably photographic--as in the case of a negative print.

Thus, while lifelike appearance and faithful detail are the hallmarks of the photograph, an image can contain significant deviations in both respects while still remaining photographic. At the same time, an image may be extremely accurate and lifelike without being a photograph at all. A skilled artist can draw or paint an image that looks as real as a photograph--or even more so. Had Winterland engaged such an artist to create an image modeled after Mendler's photo, it wouldhave stayed within the terms of its license, no matter how lifelike or how similar to the original the image looked. This doesn't help Winterland, however, for we know that whatever photographic elements remain in the T-shirt image were not created by Winterland's artistry--they were captured mechanically in the chamber of Mendler's camera. Unlike a painting, where every detail successfully reproduced on the canvas is a triumph of the artist, here every detail that tracks the original represents the extent to which Winterland created nothing.

The contract did authorize Winterland to select the illustration process pursuant to which it used Mendler's photographs "as guides, models, [or] examples. " Winterland was thus within its rights in choosing to make its illustration by scanning Mendler's photograph and using computer software to digitally alter it to create an "illustration. " In choosing this method rather than reconstructing the image from scratch, however, Winterland necessarily took on a burden of altering the image sufficiently so it would no longer exhibit those qualities that cause us to recognize it as a photograph. This must be so, for if the use of a photographic process to reach a recognizably photographic result is authorized, the parties' avowed understanding that photographic reproductions are not "illustrations" becomes meaningless.

Viewing the problem through this lens, we conclude that the alterations made by Winterland failed to destroy the essentially photographic quality of the image on its T-shirt. Were this question to hinge solely on the appearance of the T-shirt image when viewed alone, the case might be a close one. Changes in color alone do not render an image any less photographic, but here the addition of posterization has produced an effect such that at first glance it is unclear how the image was created. The question, however, is not whether the T-shirt image is readily recognizable as a photograph standing alone. To evaluate the degree of accurate, lifelike detail an image contains, we must necessarily compare it to the original. See Appendix.

Once we do this, all doubts disappear. The precise shapes of the two boats, their positions in the water, their spatial relationship to each other--all remain perfectly distinct and (apart from the horizontal flip) identical to the original. Though somewhat washed out by the posterization, even most of the finer details of the original photo--the stitching and insignia in the sails, the positions of the crew members, the reflection of a boat in the sun-dappled water--remain visible and unaltered. The smoothing out of the background and reconstruction of the sail tip are within the range of cosmetic retouching we see in media photographs every day. Apart from the sail tip, none of the elements of the T-shirt image that can be said to "illustrate" anything were added by Winterland--they were simply scanned from Mendler's photo. Despite the differences in appearance, no one familiar with the original can fail to recognize this.15 The T-shirt image thus remains essentially what it was the moment it was transferred from Mendler's slide to the hard drive of Winterland's computer: a photographic reproduction. It is now a filtered, posterized reproduction--but photographic nonetheless.

As we find that Winterland's use of the photo exceeds the terms of the license, it was an unauthorized use and therefore infringes Mendler's copyright. We REVERSE and REMAND for a determination of damages.

[Tabular or Graphical Material Omitted]

The real riddle in this case is: When is a record no longer a record?

The majority opinion cites no fewer than two web sites, one computer software user's guide, one book, two dictionary definitions, and six newspaper or magazine articles -none of which was referred to, introduced, validated, used or argued in the district court or to us. While it makes for interesting reading, I have no idea whether the parties' intent was shaped by the existence of a "cottage industry dedicated to helping people improvetheir family photos by sending that better forgotten ex-spouse down the oubliette of history, " much less colorized pictures of snow boarders with green hair and bright pink tongues. These things were not in the record, and I don't even know whether they existed at all when the contract was formed in 1991. In any event, these data are not the usual stuff of contract interpretation.

I would instead use more conventional tools to ascertain what the parties meant when they allowed Winterland to make "illustrations." See Cal. Civil Code SS 1636, 1647; City of Atascadero v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 80 Cal. Rptr. 2d 329, 349 (Ct. App. 1999). Hunter argues that the scope of the copyright license was limited to "cartoonstyle" or "graphic" illustrations. He also contends that the parties did not intend for Winterland to use computer-scanned images of his photographs. I don't agree. While it is clear that the parties intended the term "illustrations" to be limited to "graphic" illustrations (thereby excluding photographic reproductions), the evidence does not support Hunter's further limitation to "cartoon-style" illustrations. Nor does it support the exclusion of computer-scanned images as "guides, models, and examples" for computer-created artwork. The contract allows Winterland to "use whatever illustration process it finds most appropriate." Winterland had the technology (albeit less sophisticated) to scan and manipulate images at the time of the contract; thus, this case is distinguishable from Cohen v. Paramount Pictures Corp., 845 F.2d 851, 854 (9th Cir. 1988), in which the relevant technology, home videocassette recorders, did not exist in any form at the time of the contract. Further, Winterland could have used the same technology to produce simple line drawings that Hunter admits are within the scope of the license.

As I see it the issue is not "when does a photograph stop being a photograph," rather it is whether this particular digitally-scanned and manipulated image is within the scope of the license. Having reviewed the record and exhibits, I am not firmly convinced that the district court erred in finding that the "Cross Sails" image is within the scope of the license. Winterland's manipulation of the photograph was significant. It was flipped horizontally, one sail was elongated, colors were changed dramatically, the sky was redrawn, and it was posterized in such a way as to destroy and compress tonality. While the resulting image is obviously based on the photograph, it is not the photograph. Rather, the photograph was used as a guide or model to produce a graphic illustration of sailing.

Given that Winterland did not infringe upon Hunter's license, I do not believe that the district court erred in finding that the San Diego Yacht Club was not liable for infringement. I would, therefore, affirm.