In Carol Barnhart Inc. v. Economy Cover Corp., 773 F.2d 411, 418-19 (2d Cir.1985), we held that mannequins of human torsos were not copyrightable sculptural works. The parties before us today dispute whether the Copyright Act protects not human forms, but rather forms designed to hold up fish skins. This case therefore presents a modest question: Is taxidermy different?1 We hold that it is.
The parties also dispute whether the district court acted prematurely in considering, as part of its copyrightability inquiry, whether there are so few ways of expressing the idea of a fish mannequin that the idea and its expression effectively “merge.” We hold that it did consider the merger doctrine too early, and therefore vacate and remand for further proceedings.
I. Background
The plaintiffs are designers of taxidermy mannequins, which are used to mount animal skins. The plaintiffs registered their creations with the United States Copyright Office, and claim that the defendant, Dan Chase Taxidermy Supply Co. (“Dan Chase”), is infringing their copyrights by selling exact duplicates of their products.
After an evidentiary hearing, the district court concluded that almost all of the animal mannequins at issue (bear, moose, sheep, deer, steer, and antelope) had enough artistic detail to be copyrightable. 884 F.Supp. 71, 77 (N.D.N.Y.1995). The only exceptions were fish mannequins. The fish, the court found, had “no meaningful detail ... that is not commanded by the idea of a realistic fish.” 884 F.Supp. at 76. Accordingly, it held that the fish “exemplif[ied] the merger of idea and expression” and were not copyrightable. Id. The court dismissed the plaintiffs’ claims with regard to the fish and directed that the case proceed to trial to determine whether the other animal mannequins had, in fact, been copied by Dan Chase.
The parties settled as to all the other animals, and no trial ensued. Because of the settlement, only the fish-carving plaintiffs remain to appeal, and they contest the district court’s ruling that their fish mannequins were not copyrightable.
II. Discussion
Since the plaintiffs registered their fish mannequins with the U.S. Copyright Office, their works enjoy a presumption of protection, 17 U.S.C. § 410(c), and Dan Chase bears the burden of proving that the registered copyrights are invalid, Barnhart, 773 F.2d at 414.
The Copyright Act covers “[plictorial, graphic, and sculptural works” which include “three-dimensional works of fine, graphic, and applied art, ... diagrams, [and] models ... includ[ing] works of artistic craftsmanship insofar as their form but not their mechanical or utilitarian aspects are concerned.” 17 U.S.C. § 101. The Act does not, however, protect “useful articles” — that is, articles “having an intrinsic utilitarian function that is not merely to portray the appearance of the article or to convey information.” Id. A useful article will “be considered a pictorial, graphic, or sculptural work only if, and only to the extent that, such design incorporates pictorial, graphic, or sculptural features that can be identified separately from, and are capable of existing independently of, the utilitarian aspects of the article.” Id.
Mannequins straddle the line between utilitarian and sculptural works, and so one is tempted to describe them as “useful sculptures.” On the one hand, the district court accepted testimony that each sculptor’s “intent and purpose in creating the animal mannequins at issue was to make lifelike forms and to capture his or her individual perception of the animals.” 884 F.Supp. at 74. Such testimony supports the view that the mannequins are artistic expressions that qualify as protectable “sculptural works.” Yet the products indisputably serve a very practical purpose: They act as a mount for animal skins. Id. This characteristic suggests that they may be no more than “useful articles,” and, as such, not copyrightable. In fact, however, many objects are both useful and works of artistic craftsmanship. Even *322useful articles, moreover, can gain copyright protection for any “physically or conceptually separable” artistic features. Barnhart, 773 F.2d at 418.
The district court in this case, however, did not decide whether the fish mannequins were useful articles, and if they were, whether they nevertheless had any “physically or conceptually separable” artistic features that were copyrightable. (It merely noted that “most” of the animal forms in dispute were not useful articles, 884 F.Supp. at 74, and that “many” of the forms also had conceptually separable artistic aspects, id. at 75.) Instead, it held that “even if the fish mannequins are sculptural works or have artistic separable parts, the fish mannequins ... are not copyrightable” because “there are [so] few ways in which to express the idea of a realistic fish body in a mannequin [that] the expression merges with the idea.” Id. at 77.
A. Merger of Idea and Expression
We have explained the merger doctrine in the following way:
The fundamental copyright principle that only the expression of an idea and not the idea itself is protectable has produced a corollary maxim that even expression is not protected in those instances where there is only one or so few ways of expressing an idea that protection of the expression would effectively accord protection to the idea itself.
Kregos v. Associated Press, 937 F.2d 700, 705 (2d Cir.1991) (citation omitted); see 17 U.S.C. § 102(b) (“In no case does copyright protection ... extend to any idea____”). But, having recognized that the concept of merger differs from the question of infringement, we nevertheless indicated our strong preference for considering, in the context of the infringement inquiry rather than in that of copyrightability, the issue of whether an idea inevitably “merges” with its expression. Kregos, 937 F.2d at 705. A court will normally have a “more detailed and realistic basis” for evaluating the identity between idea and expression if it has all the contested forms of expression before it. Id. For in essence, the merger inquiry asks whether all realistic fish mannequins, no matter how artistic they might be, will necessarily be “substantially similar.” And only if this is so, is there no unique expression to protect under the copyright laws. See Durham Indus. v. Tomy Corp., 630 F.2d 905, 916 (2d Cir.1980); Reyher v. Children’s Television Workshop, 533 F.2d 87, 91 (2d Cir.) (“ ‘[S]iimlarity of expression ... which necessarily results from the fact that the common idea is only capable of expression in more or less stereotyped form will preclude a finding of actionable similarity.’”) (quoting 1 Nimmer on Copyright § 143.11 (1975 ed.)), cert. denied, 429 U.S. 980, 97 S.Ct. 492, 50 L.Ed.2d 588 (1976).
We hold that the district court invoked the merger doctrine prematurely. It is true that Kregos did not categorically forbid district courts from reaching merger questions before hearing evidence about substantial similarity. There may be highly unusual cases in which virtually all of an idea’s possible expressions are before a district court at the copyrightability stage. In such rare cases it may perhaps be possible to determine the merger issue while deciding whether a given expression is copyrightable. But we agree with Kregos that this is very unlikely and therefore adhere to our strong preference that the question be decided only after all the evidence of substantial similarity is before the court.
In the current case, eight of the plaintiffs’ fish mannequins were admitted into the record as exhibits and are before us. See 884 F.Supp. at 73, 76 (determining eopyrightability based upon eight of the plaintiffs’ fish, selected by stipulation of the parties). But we have not seen the defendants’ mannequins. And at the moment, we cannot say that all realistic fish are necessarily so similar to each other that to copyright any of them would be tantamount to copyrighting an idea rather than an artistic manifestation.
More important, we do not know how the district court could comfortably reach that conclusion at this stage of the proceedings. Certainly, the court was able to compare the eight fish forms that were before it, some of which were by different craftsmen. But had the court considered the defendants’ purported copies alongside the plaintiffs’ original fish *323mannequins, it would have had a somewhat broader evidentiary basis for determining whether the universe of possible fish mannequins is so narrow as to preclude copyright protection. It could, for instance, compare and contrast purported copies and admitted originals by different carvers. After seeing all such additional evidence, the court may very well adhere to its view that all realistic fish mannequins must look the same. But such a sweeping conclusion would then be more reliably grounded and there would exist a more “detailed and realistic basis,” Kregos, 937 F.2d at 705, for determining whether the particulars of all fish mannequins are dictated by the very notion of a lifelike fish.
Because we believe that the district court’s determination should be made only after hearing evidence about substantial similarity, we vacate the judgment and remand for that hearing.
B. Useful Articles
Dan Chase urges us to affirm anyway, arguing that fish mannequins are “useful articles” and hence are not entitled to copyright protection regardless of merger. It relies primarily on our decision in Barn-hart, in which we found that styrene human torsos used to display blouses and sweaters were not copyrightable. In that case, the parties had stipulated that the torsos were “useful articles,” 773 F.2d at 414, that were designed “to display clothes,” id. at 418. Taking this concession as our starting point, we concluded that the torsos did not possess separable aesthetic or artistic features, and that nothing about them was therefore copyrightable. Dan Chase argues that the fish mannequins serve the same purpose as the torsos did in Barnhart — they act as a mount for skin or clothing. In both cases, Dan Chase says, the “display” function dictates the shape of the mannequin.
We do not agree that Barnhart mandates a finding that fish mannequins are “useful articles” undeserving of copyright protection. First, in Barnhart, we had no occasion to question the parties’ stipulation that the human torsos were “useful articles.” But, even apart from the stipulation, we do not believe that the Barnhart torsos can be analogized to the fish in this ease. In Barnhart, the headless, armless, backless styrene torsos were little more than glorified coat-racks used to display clothing in stores. The torsos were designed to present the clothing, not their own forms. In taxidermy, by contrast, people look for more than a fish skin; they wish to see a complete “fish.” The superficial characteristics of the fish, such as its color and texture, are admittedly conveyed by the skin, but the shape, volume, and movement of the animal are depicted by the underlying mannequin. Whether the fish is shown as resting, jumping, wiggling its tail, or preparing to munch on some plankton, is dictated by the mannequin and by its particular form, not by the skin. 884 F.Supp. at 74 (crediting expert testimony that “a taxidermist buys a mannequin which expresses the pose, attitude and appearance of the animal that will satisfy his or her customer”).
In short, the fish mannequin is designed to be looked at. That the fish mannequin is meant to be viewed clothed by a fish skin, rather than naked and on its own, makes no difference. The function of the fish form is to portray its own appearance, and that fact is enough to bring it within the scope of the Copyright Act. 17 U.S.C. § 101; accord Superior Form Builders v. Dan Chase Taxidermy Supply Co., 74 F.3d 488 (4th Cir.1996) (distinguishing Barnhart and holding that mammal taxidermy mannequins are “sculptural works” rather than “useful articles” because their utilitarian aspects serve “merely to portray the appearance” of the animal), pet. for cert. filed, 64 U.S.L.W. 3765 (Apr. 29, 1996).
We conclude that fish mannequins, even if considered “useful articles,” are useful insofar as they “portray the[ir] appearance.” 17 U.S.C. § 101. That makes them copyrightable.
III. Conclusion
The district court may be right: If you’ve seen one fish mannequin, you’ve seen them all. But it may also be wrong. In the end, a determination can only be properly made after all the evidence of “substantial similarity” has been heard.
*324Accordingly, we vacate the judgment and remand the case.
. Cf. United. States v. Forma, 42 F.3d 759, 760 (2d Cir.1994).