When Nikolai Sergeev fled Russia in 1918, he carried with him something more valuable than jewels or currency: a trunk filled with notebooks covered in cryptic symbols. These were the Stepanov notations of the Imperial Ballet's repertoire—works by Marius Petipa that would otherwise have vanished with the old regime. But there was a problem. After Sergeev died in 1951, no one could read them. The code had been lost.

This is the central paradox of dance notation: we create systems to preserve movement across generations, yet the systems themselves can become obsolete, leaving us with records we cannot decipher. It wasn't until 1988 that researchers finally cracked Nijinsky's modified Stepanov notation of "L'Après-midi d'un faune," allowing the 1912 ballet to be revived exactly as he'd choreographed it. Seventy-six years is a long time to wait for a translation.

The Problem Dance Shares With Music

Unlike music, which settled on a universal notation system centuries ago, dance has spawned dozens of competing methods. Each promised to be the definitive solution. None achieved universal adoption.

The first serious attempt came from the court of Louis XIV in the 1680s. The Beauchamp-Feuillet system—named after either Pierre Beauchamp or Raoul-Auger Feuillet, depending on whose 1705 complaint you believe—impressed Voltaire enough that he ranked it among the great achievements of his era. Denis Diderot devoted ten pages to it in his Encyclopédie. Yet by the 1780s, it had faded from use.

The pattern repeated throughout the 20th century. Vladimir Stepanov published his method in 1892, encoding dance movements like musical signs. Margaret Morris took an anatomical approach in 1928, providing separate symbols for each body part. Rudolf Laban published "Schrifttanz" the same year, using a vertical three-line staff that would become the most widely adopted system. Rudolf and Joan Benesh created their own method in the 1950s, using a five-line horizontal staff showing the dancer from behind. Eugene Loring invented "Kineseography." Alwin Nikolais developed "Choroscript." The list goes on.

Why Ballet Gets Recorded and Modern Dance Doesn't

Ballet is the easiest form to notate, which partly explains why it dominates historical preservation. The vocabulary is standardized: fifth position croisé, tendu devant, pas de bourée. A notator can use abbreviations and trust that dancers worldwide will understand. Contemporary choreographers teaching their work often resort to makeshift systems—SR/SL for stage right and left, formation diagrams, single letters representing individual dancers, counts scribbled next to movement descriptions.

Modern and contemporary dance resist notation because the movements often lack names. How do you write down something that's never been done before, that exists only in the body of the dancer who just invented it? Labanotation can theoretically capture any human movement—it's been applied to gymnastics, sports, remedial exercises, even zoological studies. But the more complex and unique the choreography, the more time-consuming the notation becomes.

This creates an uncomfortable truth: we preserve what's easy to preserve, not necessarily what's most worth preserving.

The Video Temptation

Modern choreographers face a choice their predecessors didn't have: why bother with notation when you can just film everything? Video captures nuance that symbols on a page cannot—the quality of a gesture, the relationship between dancers, the actual faces and bodies of the original cast.

But video has its own problems. Equipment fails. File formats become obsolete. (Try opening a QuickTime file from 1995.) Cameras must be set up; they require specific angles and lighting. A notated score can be consulted during rehearsal without plugging anything in. More importantly, notation forces the choreographer to analyze and understand the structure of their own work in ways that simply recording it does not.

The debate intensified with motion capture technology. Markerless systems like Move AI now use machine learning to recognize body points in 2-D video and generate 3-D renderings without physical markers or suits. Rashaad Newsome trained an AI humanoid on Black feminist theory and used performance capture of vogue dancers for his 2022 work "Being." SAG-AFTRA negotiated new contracts addressing AI security and compensation for motion-capture performances.

These technologies raise a question: is this notation or documentation? The distinction matters. Notation abstracts movement into symbols that can be interpreted. Documentation preserves a specific performance. One is a recipe; the other is a photograph of a meal.

When Dead Languages Come Back to Life

The Harvard Theatre Collection houses what's known as the "Nikolai Sergeev Dance Notations and Music Scores for Ballets"—those notebooks that escaped Russia in Sergeev's trunk. Mona Inglesby sold them to Harvard between 1967 and 1969. For years they sat largely unused, a dead language waiting for a Rosetta Stone.

When researchers finally learned to read Stepanov notation, they discovered something unexpected. The notations revealed that Petipa's original choreography differed significantly from what companies had been performing for decades. Each generation of dancers had subtly altered the steps, believing they were preserving tradition while actually transforming it. The written record exposed the myth of unbroken transmission.

This is what notation does at its best: it reveals what memory obscures. Alexander Gorsky, who took over the Imperial Ballet notation project after Stepanov died in 1896, understood this. He perfected the system and prepared detailed tables of signs, creating a record that would outlast everyone who danced those roles.

The Notation No One Uses

The irony is that we now have better notation systems than ever before, yet fewer choreographers use them. Labanotation and Benesh notation are sophisticated, comprehensive, and teachable. The Benesh Institute was established in London in 1962 specifically to train notators for ballet companies. Universities offer courses in both systems.

But learning to read and write dance notation takes nearly as long as learning to read and write music. Most choreographers would rather spend that time making dances. Professional notators exist, but hiring one is expensive, and funding for dance is scarce enough without adding archival costs.

Meanwhile, Professor Ben Baker at Colby College used machine learning to analyze dance videos and identify attributes of ten hip-hop genres with 76% accuracy—double what human observers achieved. The computer doesn't need a notation system. It finds patterns in raw movement data.

Perhaps the future of dance preservation isn't a better notation system but the abandonment of notation altogether. Or perhaps we'll realize too late that something essential is lost when we stop translating movement into symbols—the same way something is lost when we stop reading maps and trust GPS instead. We arrive at the destination either way, but we no longer understand the territory.