Online materials for timbre and affect study

This study explored the role of timbre in the perception of affect dimensions in isolated musical sounds. The sounds used in the experiments are available here.

Eerola, T., Ferrer, R., & Alluri, V. (in press). Timbre and affect dimensions: Evidence from affect and similarity ratings and acoustic correlates of isolated instrument sounds. Music Perception. [link to the online edition]

 Experiment 1 stimuli

In experiment 1, participants evaluated perceived affects of 110 instrument sounds that were equal in duration, pitch and dynamics using a three-dimensional affect model (valence, energy arousal, and tension arousal), preference and emotional intensity. The sounds were taken from the McGill University Master Samples (MUMS) collection (Opolko & Wapnick, 2006). These sounds are contained in a compressed archive (MUMS excerpts.zip), in which the filenames are numbered from 1 to 110. The names, articulations and other details of the files are given in a separate document (Appendix 1.pdf). 

Experiment 2 stimuli

This experiment used 18 sounds from the Experiment 1 in an emotional dissimilarity task to reveal the underlying affect structure. The numbers, which refer to the indexing of the Experiment 1, is given in the table below. 

Nro Exp. 1
1 12 
3 82 
10  60 
11  64 
12  37 
13  22 
14  67 
15  25 
16  28 
17  63 
18  16 


Experiment 3 stimuli

 In experiment 3, a set of 105 instrument sounds were used. These were obtained from Vienna Symphonic Library (VSL). The library includes sounds played at different dynamic levels, as well as in different registers. The samples were then split into three subsets of 35 samples each.  

Subset 1

The Subset 1 was chosen from the forte dynamic level. These included 11 unique instruments (violin, cello, trombone, trumpet, bassoon, flute, oboe, marimba, clarinet, horn, vibraphone) with up to 7 articulations (Plain, Staccato, Vibrato, Legato, Sforzato, Marcato, and Pizzicato, see  Appendix 2 below).

Subset 2

The Subset 2 comprised of the same 35 sounds but were taken from a different initial dynamic level (mezzo-forte) in the sample library but equalized to the same level as the Subset 1 sounds.

Subset 3

Finally for Subset 3, we wanted to take the same 35 sounds from Subset 1, and make a straightforward alteration to the spectrum to see whether such a manipulation of ratio of high-frequency to low-frequency energy modified the affect ratings in the predicted direction. For this purpose, a two-pole IIR filter with a resonant frequency at 2000 Hz was applied to each sound (see the paper for details).



Opolko, F., & Wapnick, J. (2006). The McGill University Master Samples Collection on DVD (3 DVDs). Quebec, Canada: McGill University.