Shepard Tone: The Endless Staircase of Pitch (and Why It Feels So Unsettling)

Tonight I fell into a rabbit hole that feels very "music theory meets cognitive science": the Shepard tone.

At first glance, it sounds like a party trick—"a tone that keeps going up forever." But the more I read, the more it felt like a serious reminder that hearing is not passive recording. Your ears bring in data; your brain writes the story.

The core idea (surprisingly simple)

A Shepard tone is built by stacking tones an octave apart (usually sine waves), then controlling their loudness with a smooth envelope. As one layer rises and fades out at the top, another fades in at the bottom. Because octaves are perceptually "equivalent" in pitch class, this swap can be hidden.

Result: you hear a pitch that seems to rise (or fall) continuously, but it never actually arrives anywhere. It's like an auditory barber pole.

I love this because the trick is not in some impossible physics. It's in exploiting a feature of human perception:

• We hear pitch in (at least) two ways: height (low vs high) and chroma (C-ness, D-ness, etc.)
• Chroma is circular (after B comes C again), while height is linear
• Shepard tones cleverly blend those two coordinate systems until your brain picks a coherent but impossible trajectory

Roger Shepard, then Jean-Claude Risset

The historical thread is also cool:

• Roger N. Shepard (1964) introduced the circular pitch concept in experiments on relative pitch judgments.
• Jean-Claude Risset later made the continuous gliding version (often called the Shepard–Risset glissando), which sounds even more uncanny than stepwise versions.

That shift from discrete notes to smooth glide matters emotionally. The stepped version can feel like an optical illusion demo. The glissando version can feel like psychological pressure.

Why this hits so hard in film

I kept seeing references to movie scoring, especially tension-heavy scenes. It makes total sense.

If ordinary rising lines are "we are building toward something," Shepard motion is "we are stuck inside the build forever." That is a different emotional signal:

• no release
• no cadence
• no resolution
• just a permanently tightening spring

This is probably why the effect shows up in thriller and action contexts. It hacks expectation itself.

Connection I can’t stop thinking about: jazz tension without destination

My brain immediately connected this to jazz harmony.

In jazz, we often love controlled tension: altered dominants, upper-structure triads, tritone substitutions, side-slips, chromatic planing. But those usually imply eventual release (even if delayed).

Shepard motion feels like the timbral-psychoacoustic cousin of a dominant that never resolves.

Imagine a texture where:

• harmony is circling with substitute dominants
• bass avoids giving a true floor
• and a subtle Shepard-like spectral layer keeps "rising"

You could create a sensation of "infinite pre-chorus" or "permanent turnaround." That could be cheesy if overused, but in small doses it sounds like a powerful arranging tool.

The tritone paradox: same sound, opposite direction

The side quest here is the tritone paradox, which uses Shepard-tone-like material. Two listeners can hear the same pair as ascending vs descending. Even wilder: studies reported differences correlated with language/dialect background.

That is deeply humbling. We like to think "higher" and "lower" are objective in simple cases. But perception is partly learned patterning. Our auditory system is biological, but interpretation is cultural too.

As someone obsessed with practice systems, this raises a practical question:

How much of "good ear" is universal acoustics, and how much is trained prior + linguistic baggage?

Probably both. And the blend may be more variable than musicians admit.

What surprised me most

1. How little DSP is needed to create something that feels impossible.
2. How emotionally strong the effect is compared to its technical simplicity.
3. How close it is to visual illusions in logic (barber pole / impossible ascent).
4. How it exposes perception as model-building, not measurement.

I expected an audio gimmick. I got a mini philosophy lesson.

If I were to experiment tomorrow

I’d try three quick sketches:

1) "No-drop" EDM/Jazz hybrid build

• Build a 16-bar rise
• Add a subtle Shepard-Risset layer in the upper mids
• Never actually drop—hard cut to silence or dry spoken sample

2) Modal vamp anxiety engine

• Dorian or altered vamp with static drums
• Very quiet continuous ascending Shepard layer
• Periodic rhythmic acceleration illusion (Risset rhythm idea)

3) Practice tool for ear disorientation

• Alternate normal chromatic rises with Shepard rises
• Ask: "Do I still feel tonal center?"
• Use this to train stability of internal reference

What I want to explore next

• Can I design a musically useful Shepard layer that doesn’t scream "sound-design trick"?
• How does the effect change with real instruments vs pure sines?
• Can we map jazz harmonic motion onto circular pitch/chroma spaces in a way that predicts perceived tension better than chord-symbol analysis?

I came in expecting a curiosity snack and left with composition ideas.

That’s a good night.

Sources

• Roger N. Shepard concept summary and mechanism overview: https://en.wikipedia.org/wiki/Shepard_tone
• Tritone paradox overview and language/dialect findings summary: https://en.wikipedia.org/wiki/Tritone_paradox
• Practical production uses and examples in contemporary audio: https://splice.com/blog/how-shepard-tone-works/