Skip to main content
SONOLOGIUMATLAS
All sound essays

Editorial essay · 2026-06-08

The Epistemology of Data Sonification

Hear the mapping, and preserve the phenomenon’s separate identity

By Evelyna Bellamy · 618 words · about 4 min read

Estimated at 200 words per minute for the abstract and main prose; references and interactive studies take additional time. This editorial essay has not been externally peer reviewed.

Revised 2026-10-11. Complete bounded sound essay with identified sources and declared examples. Unsupported synthesis, universal resonance and unreported measurement claims withdrawn.

Sonification makes selected data relations audible through a specified transformation. NASA’s 2022 Perseus example involves gas pressure waves resynthesized into the audible range; a declared three-value mapping shows a different technique. Neither is an unchanged microphone recording of a black hole, and a compelling sound does not validate its own mapping.

A sound arrives through a decision

When a dataset becomes a sound, the transformation selects what will be heard and how. A spatial position might become stereo placement, a brightness value might become loudness, or an ordered sequence might become playback time. The listener encounters that representation through speakers or headphones. The scientific phenomenon retains its separate identity.

That distinction does not diminish sonification. It identifies the work that makes the listening possible. I want an account that lets the reader ask which data property changes the sound, what was discarded and how another reasonable mapping would differ.

Perseus has a medium

NASA’s release of May 4, 2022 describes the Perseus galaxy-cluster sonification using pressure-wave signals identified from Chandra X-ray data. The cluster’s gas supplies a medium in which those waves can occur; the claim is not sound traveling through empty vacuum. The signals are extracted radially and resynthesized 57 and 58 octaves upward into human hearing.

The factors are 2⁵⁷ and 2⁵⁸, approximately 1.44×10¹⁷ and 2.88×10¹⁷. They express a frequency rescaling, not an unaltered recording. The same release describes a separate M87 multiwavelength sonification. The two examples should not be merged into one generic account of “black-hole audio.” No independent extraction or comparison of their input data was performed here.

A different mapping, completely declared

For a teaching example, take three dimensionless values x=0, 0.5 and 1, already normalized to a chosen range. Define f(x)=220×2^(2x) Hz, use the same envelope and gain for each tone, and play them in supplied order for equal durations. The resulting frequencies are 220, 440 and 880 Hz. This is an authored mapping of invented values, not NASA’s processing pipeline.

Equal increments in x become equal octave intervals. A linear-frequency alternative, f(x)=220+660x, gives 220, 550 and 880 Hz for the same inputs. Both preserve increasing order; they emphasize different interval relationships. Choosing between them is part of the representational design, not a discovery of the dataset’s one inherent musical scale.

Normalizing a real variable would need its units, range, outlier policy and missing-value treatment. If zero means an absent observation rather than the smallest valid value, mapping it to 220 Hz would invent a measurement. The absence needs its own representation or a visible gap.

Reproducibility extends to playback

A sonification record should identify the input version, preprocessing, ordering, mapping equations, units, output duration, timbre, gain behavior and source credit. A scientific interpretation should also state which relationship the listener is expected to detect and how that interpretation was evaluated. A published sound can be reproducible while its usefulness for a particular task remains untested.

Playback adds another transformation. Digital gain is not a calibrated sound-pressure level at the ear. These monographs start no sound automatically; the argument and calculated outputs remain available as text. A separate listening study would need explicit playback conditions and appropriate evidence about its listeners.

A representation can be moving and accountable

The earlier source correction distinguished sonification from vacuum recording; this full revision develops the consequences through a real release and a checkable mapping. It does not promise that a musical contour proves the physical process or that accessibility is established merely by adding sound.

My judgment is that sonification is strongest when its choices remain audible in the explanation. The listener can appreciate the transformed result while following the numbers back to the data property. That route is what makes the sound useful as an argument, rather than asking its emotional force to supply evidence it cannot contain.

Sources and limits

  1. National Aeronautics and Space Administration. (2022). New NASA black hole sonifications with a remix. NASA. https://www.nasa.gov/universe/new-nasa-black-hole-sonifications-with-a-remix/

    Release dated May 4, 2022. Perseus gas pressure-wave signals are resynthesized 57 and 58 octaves upward; this is distinct from the release’s M87 multiwavelength mapping.

    Open primary source
  2. Hermann, T., Hunt, A., & Neuhoff, J. G. (Eds.). (2011). The sonification handbook. Logos Verlag. https://sonification.de/handbook/

    Editors’ open handbook record checked. Supplies the field context; the three-value mapping below is a fully declared editorial example, not an experiment attributed to the book.

    Open primary source

Corrections and source updates can be proposed through the archive’s provenance information.