Contribution
It names a pattern across law’s treatment of images and traces it through copyright. Courts treat images as transparent, so that interpretation is unnecessary, or as opaque, so that interpretation is impossible, and switch between the two as the result requires (pp. 686–687). It adds the psychology of verbal overshadowing to the critique of the substantial-similarity test, and argues from it for replacing that test with “a true “reproduction” right against exact or near-exact copying” (p. 688). The author calls the article mainly “critical and diagnostic” (p. 688).
Key insight
The two judicial attitudes to images look opposite but are one refusal. Both “the assumption that the image is the thing it represents and the conclusion that the image lacks meaning that could be analyzed are refusals to deal with the image as a separate thing, an entity with a complicated relationship to the real” (p. 702). Copyright then needs both: opacity to make an image the author’s own expression, transparency to strip it back to unprotected fact. And because judges and juries reason about works in words, the words they use change what they see. “Seeing is always selective, always shaped by context.” (p. 734).
Assumptions
- US law to late 2011. The doctrine is US federal copyright, with First Amendment, obscenity, privacy and evidence cases in Part I and a few Australian, Canadian and English cases.
- Representational images only. Nonrepresentational art is set aside (p. 688 n. 12).
- Perception is partly fixed, its meaning cultural. “there are certain features of human perception that work in predictable ways”, but what follows from them “is neither fixed nor universal” (p. 694).
- The cognitive evidence transfers. Verbal overshadowing was shown mostly for faces, and also for wine, voices and map distances, not for a car image (p. 735 n. 235). The article assumes it operates on factfinders comparing works, and flags that the effect “may be less than initially reported” (p. 735 n. 234).
Key results
The results are arguments from cases and from cited research, not new measurements. In the order the article makes them.
- Google Book Search (pp. 685–687). The proposed settlement excluded most images, which were replaced by a blank captioned “copyrighted image”. Public talk treated the text-only corpus as “the books”. This is the opening metaphor: law privileges the text when text and image conflict.
- Transparency (pp. 688–701). Pictures seem to “resemble unmediated reality” more than words (p. 689, quoting Feigenson and Spiesel). They persuade without seeming to argue. Scott v. Harris: eight Justices held a chase video so clear that no jury could find otherwise, yet Kahan and others found viewers’ readings split by demographics. “In other words, “picture” is more of a verb than a noun” (p. 701).
- Opacity (pp. 701–709). Courts also treat images as beyond words: the Eleventh Circuit held that nude photographs impart “no information” (p. 703). Obscenity law conflates cartoon images with real children (Whorley, McEwen v. Simmons). “Images, then, are both greater and lesser than words.” (p. 708). Tushnet’s remedy here: courts “could consider images as arguments, neither ineffable nor representations of reality” (p. 709).
- Copyrightability (pp. 712–716). Bleistein’s refusal to judge pictorial worth rests on images’ opacity. Alfred Bell lets an accident count as authorship. Photographs are protected almost without a lower bound, while typeface and short phrases are not. Mannion says idea and expression merge in a photograph, which should defeat protection, and then protects it. What courts protect in photographs are the marks that it is a photograph (Subotnik).
- Look and feel (pp. 716–719). The two-step test filters out unprotectable elements and then judges the whole, which lets them back in. Roth Greeting Cards found infringement where only trite words and lettering were copied. “That two works produce the same emotional state in a viewer does not mean that they are the same.” (p. 719).
- Judging images (pp. 719–723). Appellate courts re-see the works: Boisson reviewed quilts de novo after a three-day bench trial and reversed. Rogers v. Koons was judged from postcard-size black-and-white photographs of a large coloured sculpture.
- Realism is a style (pp. 723–733). Franklin Mint gave the painter’s own near-duplicate cardinal painting little protection as “realist”. But the paintings followed Audubon’s conventions, which contemporaries called falsifications. “In other words, reality is itself a style (or series of styles), the great success of which comes when we don’t notice the stylizing operations performed on the image.” (p. 729). Daston and Galison’s three regimes (truth-to-nature, mechanical objectivity, trained judgment) can each be read as protectable creativity or as unprotectable fact. “Realism, then, is a matter of perspective.” (p. 733).
- Verbal overshadowing (pp. 733–738). “Hearing a word made otherwise invisible objects visible.” (p. 734, quoting Lupyan and Spivey). Producing a verbal description lowers the accuracy of memory of a nonverbal stimulus but not confidence, and shifts evaluations toward extremes. Others’ descriptions also change perception. So “the ways in which the witnesses and lawyers talk about the works at issue and direct factfinders’ attention to specific features will quite literally change how the factfinders see the works” (p. 737). Of Boisson: “it might be more accurate to say that the appellate court and the trial court saw different works” (p. 738).
- Solutions (pp. 738–740). Lemley’s dissection and Fromer’s written claims would add more words. Instead: “abandon substantial similarity entirely” (p. 738). The reproduction right covers copies so near that observers “would be inclined to see two works as the same”, judged by differences, not similarities (p. 739). Non-exact cases go to the derivative-works right, with a market-definition question. No protection for style. There is, she says, “no natural law of substantial similarity” (p. 740).
- Comics and authorship (pp. 741–751). In Gaiman v. McFarlane Posner made the writer of a stock description co-author of the drawn Cogliostro, reasoning that “A reader of unillustrated fiction completes the work in his mind; the reader of a comic book or the viewer of a movie is passive.” (p. 745). Tushnet answers with fan works, which complete visual works on their own pages and screens (pp. 746–747), and with MGM v. Honda, where visual concreteness widened protection instead. “What we are really protecting in visual characters, it turns out, are words instantiated in images.” (p. 749).
- Fair use (pp. 751–755). The text model favours partial quotation, but criticism of an image often needs the whole. The redrawn Zapruder frames in Time v. Bernard Geis were still sued on. Cariou v. Prince (district court) treated a photograph as both transparent truth and the artist’s unique vision. Factor two, the nature of the work, could carry medium-sensitivity.
- Conclusion (pp. 755–759). Naïve realism about images “cannot simply be transcended” (p. 757, quoting Feigenson and Spiesel). Judges should borrow art, film and narrative theory as they borrow economics. Showing students many myopic world-view pictures changes their Steinberg judgment from infringement to shared style (p. 759).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Courts oscillate between treating images as transparent and as opaque, unpredictably | moderate | many cases across evidence, First Amendment, obscenity, privacy and copyright; selected, not sampled |
| C2 | Copyright protects images on looser terms than text in copyrightability but looser or tighter in infringement as the case requires | moderate | Bleistein, Alfred Bell, Mannion, typeface rule, Franklin Mint, MGM |
| C3 | Realism is a style, so the idea/expression line cannot be drawn by a work’s realism | strong as argument | art history (Audubon, Jakobson, Snyder), Daston and Galison |
| C4 | Verbal description in litigation changes what factfinders see and remember | weak to moderate | verbal-overshadowing studies on faces, wine, voices, maps; not tested on legal comparisons; replication caveat (n. 234) |
| C5 | Dissection and gestalt judgment cannot be combined | moderate | C4 plus the self-contradiction of current tests |
| C6 | A reproduction right against near-exact copies plus a derivative-works right would be fairer and more predictable | weak | argued; implementation not addressed (n. 258) |
| C7 | In mixed-media authorship disputes the writer is favoured over the artist | weak to moderate | Gaiman and a handful of joint-authorship cases |
Concepts
- Transparency. Treating an image as a direct copy of reality whose meaning is obvious and needs no interpretation.
- Opacity. Treating an image as beyond analysis in words, so that interpretation is impossible.
- Look and feel. The gestalt comparison of whole works in the second step of substantial similarity.
- Verbal overshadowing. The loss of accuracy in memory or perception of a nonverbal stimulus after it is described in words.
- Truth-to-nature, mechanical objectivity, trained judgment. Daston and Galison’s three standards for a scientific image: the type, the untouched record, the expert’s enhanced image.
- Thick and thin copyright. Highly creative works get broad protection; minimally creative ones are infringed only by virtually identical copies. The reform extends the thin rule to the reproduction right generally.
Connections
It builds on Feigenson and Spiesel’s Law on Display for law’s handling of images, on Silbey, Mnookin and Kahan on film and photographs as evidence, on Adler and MacKinnon on images in First Amendment and pornography law, on Farley, Yen, Subotnik and Amy Cohen on aesthetic judgment in copyright, on Daston and Galison’s Objectivity, and on Schooler’s verbal-overshadowing research. It cites Lessig’s Remix (LIT-1298) three times: for text as the elite medium and video as the vernacular (p. 710 n. 125), for Read/Write and Read/Only culture, whose appeal it traces to the special status of reading (p. 752 n. 313), and for freedom to quote as the foundation of textual fair use (p. 753). Christina Spiesel published a reply in the same volume, which this record has not read.
Bearing on the record
- QUESTION-002 (what survives a rendering in another medium). This is the main point of contact. Tushnet’s subject is the rendering of an image into words, which courts do whenever they decide an image case. She treats it as a translation that is “flattening and distorting” (p. 694), and cites Kasunic that it is “necessarily subjective and incomplete” (p. 702 n. 82). The verbal-overshadowing evidence adds a point the record lacks: the rendering acts back on the reader’s grasp of the source. A factfinder who has described an image, or heard it described, sees the image differently afterwards. What survives a rendering can then depend on whether the receiver also meets the source, and in which order.
- CLAIM-076 and CLAIM-087 (cross-modal chains). CLAIM-076 says each interpreter in a cross-modal chain, such as a captioner, may add motives and relations not present. Tushnet supplies legal cases of it: Roth, where shared words drove a finding about drawings, and Gaiman, where a writer’s stock description is credited with the drawn character. Her Barthes quotation, that to describe a photograph is “to change structures, to signify something different to what is shown” (p. 734), is the strong form. CLAIM-087’s claim that cross-modal reconstruction can improve access gets no support here; the article is about loss and distortion from image to word.
- CLAIM-220 (task-relevant communicative distinctions). A court comparing two works is a receiver with a stated task. Tushnet’s point that “the appellate court and the trial court saw different works” (p. 738) bears on how CLAIM-220 names its population of receivers: two receivers with the same task and the same objects recovered different distinctions, because of the words that came with the objects. It does not test CLAIM-220.
- CLAIM-210 (sharing a kind versus being a rendering). Franklin Mint is close to a legal case of it. The third cardinal, painted in court without looking at the first two, shares the kind of the first without being a rendering of it, and that is what the jury used to find no infringement. Copyright’s independent-creation defence draws the same line. Tushnet’s own reform, judging near-exact copies by differences, is a test of rendering, not of kind.
- CLAIM-002 (identity across retellings). Character copyright is a legal attempt to say what persists across versions: Sam Spade, Tarzan, Superman as drawn by several artists and played by several actors (pp. 744, 750). Tushnet shows the courts have no settled account, and that a character’s protection “stems from the image but goes beyond it” (p. 750). That fits CLAIM-002’s family of relations better than any one of them, but the article does not argue it.
- NOTE-1023 (Lessig). Lessig argues remix needs the original because its meaning comes from the reference. Tushnet adds the legal and perceptual side for images: criticism of an image needs the whole image, and a redrawing, as in the Zapruder case, loses the persuasive force that was the point of copying (p. 753). Both say the artifact, not only its content, carries the meaning. Tushnet also argues, against Posner, that fans complete visual works on their own pages, which is Lessig’s RW culture as a reply to a court (pp. 746–747).
- CASE-100 and CLAIM-332. No real bearing. The fan-work point is about audience creativity, not about uptake by an institution.
- No instruction for machine-learning practice, and not held by the anthology.
Limitations
- Cases are selected to show the pattern. No sample of image cases is drawn, and counter-cases are mentioned (the Second and Ninth Circuits accepting whole-image fair use, pp. 753–754) without weighing.
- The psychology is borrowed. Verbal overshadowing is shown in faces, tastes, voices and maps, not in side-by-side comparison of two works, which is what factfinders do. The replication caveat in n. 234 is material to C4 and is confined to a footnote.
- The reform is a sketch. Where near-exact copying ends, and what the narrowed derivative-works right would cover, are left open (n. 258). The derivative-works right is admitted to be “hardly a model of clarity” (p. 739).
- The pictures could not be checked. The argument leans on reproduced images (Roth, Boisson, Franklin Mint, Ty, Spawn), which did not survive extraction.
Open questions
- Does verbal overshadowing operate when the receiver has both works in front of them, as factfinders usually do? The article assumes it.
- Would a differences-based test for near-exact copies be more predictable in practice, or move the dispute to the derivative-works right without reducing it?
- For the record: when a rendering into words changes how a receiver later perceives the source, is that loss in the rendering or a change in the receiver, and can a measure of what survives (CLAIM-220) tell the two apart?