Data Explorer: Hardness, Punch, and Perceived Heaviness
This view shows the relationship between acoustic models and listening judgements. Each point represents a short segment of audio taken from the producers' mixes and stems.
- The x-axis displays acoustic hardness, estimated from spectral shape and dynamic properties.
- The y-axis displays either perceived heaviness or perceived punch, based on listener ratings.
- Colour and shape can be used to distinguish producers, stems, or song sections.
Click on any point to hear the corresponding audio and to see its numerical values.
Why This Matters
Across hundreds of segments, acoustic hardness explains a large share of the variance in perceived heaviness. Segments with higher hardness are typically judged as heavier, even when overall loudness is held within a realistic range.
Punch behaves differently. Very punchy segments can be judged as heavy or not, depending on how transients interact with sustained spectral density. In dense, distorted mixes, additional punch may make events more articulated rather than simply heavier.
The patterns here are therefore not a recipe, but evidence that heaviness is constrained by acoustic tendencies while remaining open to aesthetic, stylistic, and contextual factors.
How to Use This View
- Start by selecting a subset of stimuli (for example, all choruses, or all guitar-only stems).
- Click on points along a diagonal band: do they sound similarly heavy, or are there perceptible differences?
- Find outliers: segments that are judged very heavy but have only moderate hardness, or vice versa. What production choices explain them?
- Compare producers: do some cluster in a harder or softer region? Do their stems show similar or different patterns?
Advanced Metrics
In addition to hardness, punch, and heaviness, you can inspect:
- Peak-to-Loudness Ratio (PLR) – a rough index of macro-dynamics
- Loudness Range (LRA) – variation of loudness over time
- Spectral centroid and low-frequency energy – balance between brightness and low-end weight
These measures do not replace listening, but they show how different aspects of dynamics and spectrum co-vary with heaviness judgements.
Hardness vs. Perceived Heaviness
Each dot is one producer + section stimulus (n = 18). The regression fits mean heaviness (survey) to the acoustic
hardness model.
r = 0.77, R² = 0.60
Tap a point to inspect producer metrics and audition the stimulus.
Hardness vs. Punch Trade-off
Each bubble is one producer + section stimulus (n = 18). Bubble size represents perceived heaviness (larger = heavier).
Producers face a trade-off: maximizing hardness often reduces transient punch, and vice versa. r = 0.39, R² = 0.15
Bubble size = Perceived heaviness rating
Tap a bubble to inspect producer metrics and audition the stimulus.