GEMScapes: linking acoustic features to aesthetic emotions in ambient and cinematic music using the GEMS-9 scale
Year: 2026
Type: preprint
Abstract: Ambient and cinematic music are widely used in behavioral and immersive multimedia paradigms, yet standardized stimulus sets with computational acoustic norms remain scarce. We introduce GEMScapes, an open-access dataset of 38 original, high-fidelity instrumental tracks designed to elicit aesthetic emotions free from familiarity or semantic bias. We combined computational acoustic profiling across temporal, spectral, cepstral, tonal, harmonic–percussive, rhythmic, and stereo spatial domains with 816 trial-level ratings of 36 of the tracks, provided by 65 participants in 68 listening sessions on the Geneva Emotional Music Scale (GEMS-9). To account for stimulus- and participant-level dependencies, we fitted crossed linear mixed-effects models (LMMs) and validated them with ordinal models across 171 candidate pairs, identifying nine predictor–emotion associations (Benjamini–Hochberg false discovery rate [FDR]-adjusted p < 0.05) involving six acoustic and spatial descriptors across six emotional dimensions. Temporal variability and soundstage spatialization were the main predictors: variability of spectral tilt predicted Tension, variability of percussive energy predicted Joyful Activation, Transcendence was associated with wider stereo width and smoother, non-percussive onsets, Peacefulness and Nostalgia with spectral stability, and Sadness with lower spectral brightness and reduced timbral modulation. Variance partitioning revealed a continuum from strongly stimulus-driven states (Tension: 46% stimulus vs. 9% listener variance) to predominantly listener-dependent ones (Wonder: 8% vs. 32%). GEMScapes provides an acoustically characterized resource for music perception, affective computing, and neurocognitive research.
Source: PsyArXiv (OSF Preprints)
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7233468232
