concurring, with whom Judges McKEE and SMITH join.
I join in Parts I and II of the majority opinion. For the reasons set forth in Part II hereof, I do not join in Part III of the majority opinion, dealing with short phrases. I write separately to set forth additional grounds for affirmance, relying on the doctrine of scenes a faire, which I believe undergirds and complements the majority’s explanation of why the Southco part numbering system does not meet the originality requirement.
I.
A.
Scenes a faire has been most commonly employed in the literary or dramatic context to describe those otherwise expressive elements of a work that are “standard, stock, or common to a particular topic or that necessarily follow from a common theme or setting.” Dun & Bradstreet Software Servs. v. Grace Consulting, Inc., 307 F.3d 197, 214 (3d Cir.2002) (quoting Gates Rubber Co. v. Bando Chem. Indus., 9 F.3d 823, 838 (10th Cir.1992)). Industry standards often trigger the doctrine of scenes a, faire. As the Tenth Circuit has explained, scenes a faire has been utilized
to exclude from protection against infringement those elements of a work that necessarily result from external factors inherent in the subject matter of the work. For computer-related applications, these external factors include hardware standards and mechanical specifications, software standards, and compatibility requirements, computer manufacturer design standards, industry programming practices, and practices and demands of the industry being serviced.
Mitel, Inc. v. Iqtel, Inc., 124 F.3d 1366, 1375 (10th Cir.1997).
To develop its part numbers, Southco first chose certain “characteristics” of the captive screw, such as screw length, thread size, or finish, to incorporate into its system. Next, Southco determined which “values” to assign to each characteristic. For example, the values for the “materials and finishes” characteristic include “natural finish, aluminum” or “black finish, steel.” Next, Southco set Arabic numerals to represent each characteristic and its attendant values. Thus, the symbol “10” was chosen to represent the value “natural finish, aluminum” in the material and finish characteristic. Finally, unique part numbers were generated for individual parts using the parameters provided by the first three steps.
The majority focuses its originality analysis on this final stage, when the part numbers are assigned mechanically to identify particular types of captive screws. I agree with the majority that at this stage, once all of the decisions about possible characteristics, values, and numerals *288have been made, “all of the products in the class could be numbered without the slightest element of creativity.” I acknowledge that at least superficially, the earlier stages of Southco’s part numbering system are the product of a number of choices about which characteristics to represent, choices about which values to permit, and choices about which symbols to use. I would not ignore this part of the process as the majority does. To draw an analogy to Burrow-Giles: Sarony’s photograph was protected even though all of his choices evincing originality were made before the copyrightable “work[ ] of authorship [was] fixed in a[ ] tangible medium of expression,” 17 U.S.C. § 102; and so too we should not pass by any of Southco’s choices evincing originality simply because they were made before any part numbers were actually computed and written down. See Majority Op., supra, at 285-87.
But even this cannot protect Southco’s part numbers: As the description of Southco’s part numbering system makes clear, the company selected characteristics for its system based on customer demand (an external constraint), thereby precluding copyright protection by scenes a faire. Likewise, once the characteristics were chosen, the values — such as screw thread sizes, screw lengths, or ferrule types — • were determined by industry standards rather than through any exercise of originality by Southco. Thus, for the characteristic “finish,” the values are limited to the types of finishes for screws manufactured by Southco and therefore are, determined solely by the part identity, rather than through some exercise of creative expression. Similarly, for the characteristic “thread size,” the values are simply the standard industry sizes.
B.
As I read Judge Roth’s dissent, the linchpin of her position lies in her statement that
Southco uses the first two digits of each nine-digit part number to indicate product line but it could use three digits (perhaps to easily accommodate more than ninety-nine product lines), or letter instead of numbers, or a combination of letters and numbers, or even simple abbreviations in lieu of coded letters or numbers. All these possible variations apply to each set of digits in the part number, to the order of the sets, and to the identification of which product attributes should be grouped together in the same set. Of course, there is nothing pre-determined about the length of a part number. For instance, Southco could choose to use more than nine digits to accommodate products with too many values to be easily expressed in only nine. These seem like relatively mundane choices, but, as the Supreme Court indicated in Feist, “the requisite level of creativity is extremely low; even a slight amount will suffice.”
While this argument is not without force, there appears to be a continuum involved, and I disagree with the conclusion that Judge Roth draws, for the choices at issue seem to me to fall clearly on the unoriginality spectrum. Arbitrary choices such as these do not satisfy the originality requirement. As far as the purpose of the Copyright Clause is concerned, there is no reason to give an incentive to churning out arbitrary symbols, for purely arbitrary decisions do not advance “science.”
Judge Roth cites American Dental Association v. Delta Dental Plans Ass’n, 126 F.3d 977 (7th Cir.1997) (ADA) as support for the notion that picking digits can be original. But the subject matter of that case was different; at issue was the coding system used for various dental procedures. The holding of that case is driven by the *289originality in the editorial selections that had to be made in putting the code together. The examples from ADA quoted in footnote 6 of Judge Roth’s dissent refer to editorial choices made in ordering the procedure code taxonomy in a way that was (1) expandable in certain respects, which expresses predictions about how dental science will develop; and (2) categorized in a certain way (i.e., the heading under which a procedure appears) that reflects an original way of expressing the similar relationships among different dental procedures. In ADA there was plainly a record that described the editorial choices that were made in developing the taxonomy. All the illustrations in the portion quoted by the dissent are the product of editorial choices. In contrast, the Bisbing declaration here flatly explains how baldly unoriginal was the numbering system’s architecture: the screws exist in various sizes, so it needs to use (arbitrary) symbols to represent those sizes; and so on.
I also note that the originality in ADA comes from the system’s ability to state something original about the relationship of one dental procedure to another. The same is true in Southco’s system, but the relationships expressed are totally unoriginal (Screw X is the same material as Screw Y, but a different material from Screw Z; Screw A is the same amount longer than Screw B as Screw C is longer than Screw D). Dental procedures do not come pre-categorized in the way that raw physical characteristics do. The originality in ADA stems from the tricky-to-sort-out nature of the myriad medical procedures under consideration.
In sum, not only are there a limited number of relevant characteristics, but the characteristics chosen by Southco were dictated by industry standards, customer preferences, or the objective characteristics of the captive screw itself. The scenes a faire doctrine, therefore, dispels the notion that there was the requisite originality in Southco’s selection of characteristics and values to merit copyright protection.
II.
The majority relies upon the Copyright Office’s “long-standing practice” of denying copyright protection to words and short phrases because names, titles, and short phrases “typically lack[ ] any creativity whatsoever.” The majority fails to demonstrate, however, that the Southco part number system is a “typical” case. In order for any test that purports to distinguish between short phrases and copyrightable compositions to be viable it would have to identify the point at which a title or short phrase becomes a descriptive narrative. Presumably the length of the writing in question informs this determination, but what else? The majority does not specify the test, and this is not a situation, I respectfully submit, where we “know it when [we] see it.” The Copyright Office says that short phrases “lack ... creativity” (emphasis added), but Bleistein v. Donaldson Lithographing Co., 188 U.S. 239, 23 S.Ct. 298, 47 L.Ed. 460 (1903), suggests that courts generally should refrain from subjective assessments of creative merit, and I agree. Whatever the test, I think the inquiry would inevitably draw us back to the constitutional requirement of originality-the presence of the “creative spark” from Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340, 345, 111 S.Ct. 1282, 113 L.Ed.2d 358 (1991). But to do that is to render the short phrases notion nothing more than an unhelpful way of restating the problem.
Put differently, the problem in this case is whether the Southco part numbers are words, short phrases, names or titles, or whether they are instead a compilation of data, a system of classification, or some*290thing else. Indeed, the part numbers seem to fall into the gray area between a short phrase and a more extensive work. This ambiguity is highlighted by the fact that the cases cited by the majority as approving of the Copyright Office practice are either distinguishable from the instant case or rest their decision to withhold copyright protection on other grounds. See, e.g., Kitchens of Sara Lee, Inc. v. Nifty Foods Corp., 266 F.2d 541, 545 (2d Cir.1959) (approving the Copyright Office practice, but ultimately analyzing the case under theories of originality and merger); CMM Cable Rep., Inc. v. Ocean Coast Props., Inc., 97 F.3d 1504, 1519-21 (1st Cir.1996) (finding that some aspects of the claim were foreclosed because they were merely slogans or short phrases, but resting primarily on lack of originality grounds); Murray Hill Publ’ns, Inc. v. ABC Communications, Inc., 264 F.3d 622, 633 (6th Cir.2001) (holding a line from a movie was not copyrightable where it was “nothing more than a short phrase or slogan, dictated to some degree by the functional considerations inherent in conveying the desired information”).
There are still greater problems with the majority’s approach. It appears to rely entirely on Skidmore deference. See Skidmore v. Swift & Co., 323 U.S. 134, 140, 65 S.Ct. 161, 89 L.Ed. 124 (1944). But deference to a practice of the Copyright Office, employed to assist in its humongous administrative task of deciding whether or not to grant a copyright registration, does not seem to me to be sufficient to decide an infringement case in federal court. The Copyright Office may employ rules of thumb; we may not, especially when we are dealing with a constitutional provision, for Feist holds that originality is a constitutional requirement, see 499 U.S. at 346, 111 S.Ct. 1282.
Moreover, mere homage to deference cannot prevail without at least analyzing the countervailing arguments. Southco argues forcefully that the government-ami-cus has not justified its argument that part numbers are short phrases. Southco properly points out that the term “phrase” used by the Copyright Office is a grammatical term peculiarly adapted to copyrightable subject matter expressed in words. Concomitantly, it is difficult to see what the basis is for treating numbers or numerical symbols as “short phrases.”
Southco also forcefully challenges the position that the Copyright Office in fact considers numbers as short phrases. Its brief persuasively argues, “There is nothing in the record of this case indicating that the Copyright Office ever considered that part numbers were uncopyrightable as ‘short phrases.’ No regulation of the Copyright Office refers to part numbers or any other numbers. No case is cited in which the Copyright office was a party, holding that part numbers or any other numbers are uncopyrightable.” Without a regulation or policy clearly addressing numbers, deference seems inappropriate.
For all these reasons, I would not rest this decision, even in part or in the alternative, on the short phrases argument. I add only that, as the foregoing discussion makes clear, I am in agreement with what Judge Roth has written in Part II of her dissenting opinion on this point.