Stairway to Abstraction: Selection, Arrangement and Dissection

A Layer Cake, or a Tower of Babel? Toward a General Theory and Test for Selection and Arrangement in Copyright Law

One of the most bedeviling copyright concepts is… well, it has lots of different names. In the recent “Stairway to Heaven” decision, it was called “selection and arrangement.” But in other contexts, especially software, it’s called “selection, sequence, organization” (SSO). (I talked about the Ninth Circuit’s “Stairway to Heaven” en banc decision in more detail last time.)

One model of the abstraction of a copyrighted work: ideas at the top, more expressive as you work your way down.

Morte de Everyman

To start, here’s an easy set of examples:

1. You are charged with putting together an anthology of select 15th Century English literature. There’s lots to choose from, and the decision of what to include is yours alone. You arrange them in chronological order.
2. You are charged with putting together a comprehensive anthology of 15th Century English literature. You must include it all (and will obviously consist of multiple volumes). This sort of thing has been done before, with the works arranged in chronological order, so you group them thematically according to ideas that were important to the 15th century English, based on your own knowledge and supposition.
3. Same as No. 2, but this hasn’t been done before, so you just arrangement in chronological order.

All of the underlying works are in the public domain, obviously. But you¹Or your employer or the commissioning party, depending on the circumstances. have a copyright under scenarios 1 and 2. Under scenario 1, your copyright is based on your selection of the works. Under scenario 2, your copyright is based on your arrangement of the works. However, under scenario 3, you have no copyright because you were unable to exercise any creative or editorial control over what to select (remember, you have to select it all), or over the selection (chronological order isn’t sufficiently creative).

The Detective League of the Floating Island

This is all normal and more or less spelled out in the Copyright Act. Where it complicated is when you apply this logic within a single work. Imagine if we did Alan Moore’s League of Extraordinary Gentlemen one better: we take several public-domain characters, but then put them in a public-domain story, say Arthur Conan Doyle’s A Study in Scarlet, but transported to a public-domain setting, say, Jonathan Swift’s Laputa. Regardless of how I use words to creatively put this mish-mash together into a palatable whole, do I have a copyright in the selection and arrangement of the characters, story and setting?

OK, that question was slightly misleading because, no matter what awful novel I end up making out of this mess, I will have copyright in the novel because I will be making numerous creative choices, even if they are terrible ones. So, let’s re-ask: at the level of character + story + setting, do I have a protectable interest?

We can sharpen the question this way. Suppose someone else, reading my awful novel, thinks to herself: This is a great “idea”²For a certain value of “idea.”, but the execution is miserable. She then proceeds to write her own novel, using the exact same characters, story and setting, but just about everything else is different. There are, after all, lots of ways to fit characters into A Study in Scarlet. Do we keep the Watson narration? Do we make one new characters the narrator, or perhaps go with an omniscient narrator? Are our new characters dropped into the story, or do they take over roles in the story?

Did she infringe my copyright in my novel? Did she just take the “idea” of placing certain characters into a certain story in a certain setting? Or was that too much?

(OK, as an aside, I think we can answer this question: No, she didn’t infringe. Here’s the test. If you summarize a plot with enough detail, you’ll infringe the work because, as I’ve just said, the plot, at a sufficient level of detail, is protectable. Think of the difference between IMDb’s plot summary (not detailed enough) and plot synopsis (probably detailed enough but nobody seems to care). So, if all you summarized was the characters, A Study in Scarlet and Laputa, that wouldn’t be detailed enough to infringe. She just took an “idea.” If, by contrast, she also copied the way I integrated the characters and the setting into the story, then maybe she infringed, but I think we’d need to know more.)

Literally Not Literal

This is what Prof. Nimmer calls “non-literal infringement” (if infringement it is). If you copied a section of my novel word-for-word, that would be “literal infringement” (again, if it’s infringement at all). But you can also infringe the structure of a work—a novel in this hypothetical. Now, I deliberately chose structural elements that were in public domain (and thus aren’t protectable) to make a broader point, but consider a more ordinary fictional novel. The plot is definitely protectable. Certain characters, if described with enough particularity, are protectable. Aspects of the setting might be protectable. Even items, if imaginative enough, are protectable (think of certain well-known items from Harry Potter, for example). Some things aren’t going to be protectable, such themes, ideologies, theses, goals, basic plot elements, and so forth, either because they’re too abstract or too common.

We can thus think of a work as layers of structure. At the top are the concepts, themes and goals—too abstract to be protectable. As you get less abstract, you pass through a fuzzy barrier into the realm of protectable content. At some point your novel about an eccentric detective who solves murders that the local police can’t (which is too abstract and common to be protected), which involves a false clue planted by the killer (also too abstract and common), turns into something definite enough to be protected. Perhaps something involving a wedding ring, a forced marriage in a faraway land, and taking a job as a cabbie—would that be enough to pass from mere concept to protectable plot? (Eh, probably not. You’d need to take more than that, I think.)

Regardless, it is a truism in copyright law that a sufficiently creative, original and definitive selection and arrangement of elements is protectable, even if those elements are, by themselves, non-protectable.

Dissecting a Descending Chromatic Arpeggio

That brings us to the “Stairway to Heaven” case, believe it or not. That’s because the plaintiff argued—vociferously, on appeal—that “Taurus” and “Stairway to Heaven” shared five elements in common, and that, even if each of those elements was by itself not protectable, in combination they were and were infringed. Each had descending chromatic arpeggios that:
1. Skipped note E and returning to tonic pitch A
2. Used notes of same duration
3. Had three identical two-note sequences
4. Had same eighth-note rhythms
5. Used same “pitch collection”

In short, plaintiff had dissected the one thing it claimed the two songs had in common into five musicological elements. Each of the elements, standing by itself, wouldn’t be protectable—hilariously so, in several cases.

When plaintiff lost the case, it belatedly argued that the jury had been mis-instructed on how to assess combinations of non-protectable elements. I guess the idea is that the jury might’ve dissected the descending chromatic arpeggios³Bearing in mind it was the plaintiff that did the dissecting., found them each to be non-protectable, then found there to be no similarity between protectable elements.⁴I say “belatedly” because plaintiff failed to make this argument at trial or to insist on its preferred jury instruction at the time. The jury should’ve been instructed, plaintiff argued, about how a combination of unprotectable elements can add up to something protectable and infringe-able.

In other words, plaintiff wanted to take advantage of the truism above (about multiple non-protectable elements adding up to a protectable whole), but on the other end of the expression scale. Instead of combining high-level structures to make a protectable expressive structure, plaintiff wants to combine very tiny elements to create a larger (but still fairly small) expressive element.

Combo Moves

The main mistake is that “combination” doesn’t cut it. Many works will coincidentally share multiple such (“tiny” or common) elements just by being in the same genre, and no one would say they infringe each other. It’s the selection and arrangement of the non-protectable elements that gives rise to protectable expression, not mere combination. Selection and arrangement implies a human creative mind was involved. They imply that the elements relate to each other in a meaningful or aesthetic way.

If you go back to those five elements in terms of selection and arrangement, you just end up with the very same descending chromatic arpeggios that the jury found were insufficiently similar. Why did the songwriter choose those elements? Because, together, they sounded good. Don’t overthink it.

There’s another mistake here, which goes back to what I talked about last time. When you dissect expression like this, you’re trying to prove copying, not substantial similarity. What you’re arguing is: it’s not a coincidence that the descending chromatic arpeggios have these elements in common. (And, even then, it’s not great evidence of that, since these elements are so common.) But, as I discussed last time, this sort of similarity serves a different purpose than what the law calls “substantial similarity.” It only tells you about a possible connection between the accused infringer and the copyrighted work. It doesn’t tell you whether the un-dissected segments are similar in an aesthetically meaningful way.

This is why we try not to dissect works when looking at substantial similarity. You sometimes hear the phrase “total look and feel,” in this context.

But what if there are unprotectable elements? If we can’t dissect the work, how can we remove those elements? The answer isn’t easy, but here it is. While a claim for infringement can’t be built purely of unprotectable elements, it can be built on the selection and arrangement of those elements.

If you want to get philosophical, you can (especially in music) dissect anything to the point where it’s just a selection and arrangement of unprotectable elements. Every note, every word, every brushstroke, every stock character, every melisma, every classic portraiture pose—these are just unprotectable elements. What matters is what the songwriter, writer or artist does with those elements: selection and arrangement.

The reason I say this is hard is: how do you explain this to a jury? And in a way that’s consistent not only within music, but across all creative disciplines?

Softening a Software Test

Software cases point us in perhaps a useful direction. Because computer programs are functional (not protectable!), because software engineers share techniques among themselves (also not protectable!), and because a great deal of programming is already in the public domain (obviously not protectable!), much of a computer program is unprotectable. When there is a claim of non-literal infringement of a computer program, courts employ a painstaking, tedious and exacting procedure: abstract, filter, compare (AFC).

What you do is find the highest level of potentially protectable structure in the computer programs (the allegedly infringed and allegedly infringing ones). If you use the metaphor of levels, this would be the level after you’ve crossed from just ideas to expression, which I warned you was fuzzy. This process of finding levels of expression is what the courts call “abstracting,” because the further up the ladder you go, the more idea-like (abstract) the work gets. In the case of software, it’s (I think) when we stop talking in terms of what the software will do and how it will do it. To me, that’s the level of a detailed flow-chart. It’s not a disaster to start too high, though, because you’ll end up filtering everything out.

Once you’ve gotten your level of abstraction, you filter out elements at that level that aren’t protectable. Then you’re supposed to compare what’s left over and decide whether the works are substantially similar at that level. When you’re done, you drop down a level of abstraction and continue, until you’ve reached the level of the actual code.

This process is… not great. It takes forever. It takes a lot of evidence, which is expensive to gather. It requires educating the jury about software development. Worse, we don’t really have a theory of “creativity” for software. Most developers equate creativity with elegance or efficiency, but copyright law won’t protect the most elegant or efficient way to code something (under the merger doctrine). So. This makes it hard to talk in terms of “selection and arrangement” because coders will talk about selection and arrangement in terms of (non-protectable) functionality and elegance.⁵Again, software needs its own regime of protection. Copyright and patent don’t get it done.

But for other works, this system, appropriately relaxed, could work. Start at the highest level of abstraction, look at the elements, determine which (if any) are protectable, look for the creativity (if any) behind the selection and arrangement of those elements, compare anything that’s left over, and repeat the next level of abstraction. There are a few examples of courts using the AFC system in non-software contexts, but I’m concerned that it’s too exhausting (and expensive) for most litigants. So I’d like to see the development of a short-form or relaxed version of this.

But if Nothing Else….

If nothing else, the Ninth Circuit emphasizes that there’s no copyright in mere combination of otherwise non-protectable elements, but in the selection and arrangement of those elements. If we can keep that principle in mind, and in our jury instructions, we have a chance

Next time, I want to talk about “that footnote” and the “thickness” of copyright, which relates directly to selection and arrangement.

Thanks for reading!

Rick Sanders

Rick is an intellectual-property litigator. He handles lawsuits, arbitrations, emergency injunctions and temporary restraining orders, opposition and cancellation proceedings, uniform dispute resolution proceedings (UDRPs), pre-litigation counseling, litigation avoidance, and other disputes, relating to copyrights, trademarks, trade secrets, domain names, technology and intellectual-property licenses, and various privacy rights. He has taught Copyright Law at Vanderbilt University Law School. He co-founded Aaron | Sanders with Tara Aaron-Stelluto in 2011.