Electronic Supplementary Material(Aeschlimann et al.)

Description of PCA selected stimuli

The 66 sounds included in the final battery were originally taken from the IADS (2s segments from 10 IADS sounds containing HV and from 17 sounds of objects considered as NV) and from the British Broadcasting Company (BBC) sound effects library (2s segments from 2 sounds containing HV and of 14 object sounds considered as NV). A last subset of stimuli was lab-recorded vocalizations from 7 different women and 4 men. These stimuli were artificially edited by cutting, copying, and pasting segments from the original laboratory recordings such that, for example, a sound that was originally 1s duration would have internal repetitions to render it 2s duration. So doing not only rendered the stimuli completely meaningless, but also minimized their prosodic content.

Sound Class / IADS / BBC / Lab-recorded
Positive / 10HV + 6NV / 1HV + 5NV
Neutral / 5NV / 6NV / 11HV
Negative / 10HV + 6 NV / 1HV + 5NV
  • The positive HV sounds included: a baby laughing (4 examples), a man laughing (1 example), and erotic exclamations (couple = 3 examples; female only = 3 examples).
  • Negative HV sounds included: a woman screaming (4 examples), fighting (4 examples), a man screaming with a gunshot (1 example), and a man vomiting (1 example).
  • The positive NV sounds included: applause (4 examples), opening of beer bottles (2 examples), a chord played on a guitar, lapping waves (3 examples), and a trickling river (1 example).
  • The negative NV sounds included: an alarm clock, a bicycle crashing, a bomb exploding, a buzzer, a car wreck, tires skidding, tuning of a violin, breaking glass, multiple gunshots, a siren, and a thunder clap.
  • Finally, the neutral NV sounds included: a tennis ball bouncing, a cartoon whistle, a touch-tone telephone, a busy office environment, wind blowing (2 examples), water flowing (2 examples), car horn, pen on paper while writing, and a freight train.

Analysis of acoustic parameters

RMS procedure

As it is impossible to achieve simultaneously both the physical amplitude and the perceived loudness of a set of sounds, we rather decided to fix their physical amplitude using a root-mean-square (RMS) procedure. We used Normalize, a software commonly used by sound engineers to achieve equal perceived loudness between different CD sound tracks. The mean physical amplitude of each soundis then transformed to -120.5dBFS, which correspond physically to 80dB, as can be seen in the table below (Mean I).

HV acoustic parameters

Acoustic parameters of each HV sub-class were calculated using Praat software (version 5.0.05, Boersma 2001): the mean fundamental frequency (F0) and its variability in percent (F0*[%]), as well as the mean intensity (I) and its variability in percent (I*[%]) are presented with their standard deviation are tabulated below.

Mean F0 / Variability / Mean I / Variability
[Hz] / of F0 *[%] / [dB] / of I *[%]
Positive HV / 302 55 / 26 13 / 80.2 1.0 / 14.7 ± 3.8
Neutral HV / 178 59 / 4 2 / 81.1 0.5 / 4.4 ± 1.9
Negative HV / 343 61 / 25 13 / 80.8 0.8 / 9.3 5.1
Comparison of acoustic parameters between valence classes

In order to compare acoustic parameters from each valence (positive, neutral and negative), we performed a time-frequency power analysis using Fast Fourier Transform (FFT), comparing these valence classes with a series of pair-wise non-parametric t-tests based on a bootstrapping procedure with 1000 permutations per time-frequency bin (e.g. Efron, 1982; Manly, 1991). Note that no correction for multiple contrasts or autocorrelation was performed. The spectrogram Matlab function was used for this calculation with no overlapping and zero padding parameters, giving a bin width of ~2ms and ~40Hz. These non-parametric t-tests were performed for each time-frequency bin, and their overall results are presented in the three plots, below, where black points representsignificant (p0.05) differences. Three contrasts were performed, independent of sound source category: positive vs. negative, positive vs. neutral, and negative vs. neutral. In no case was the proportion of time-frequency bins exhibiting a significant difference greater than 2%, though there is a tendency for neutral stimuli to differ from positive and negative stimuli within the lower frequency range.

Contrast / Ratio* / Proportion
Negative vs. Positive / 47 / 178358 / 0.03%
Negative vs. Neutral / 2817 / 178358 / 1.58%
Positive vs. Neutral / 2472 / 178358 / 1.39%

(*Ratio = # of bins with significant differences/total # of bins)

References

Boersma P, Weenink D (2001) PRAAT, a system for doing phonetics by computer.
Glot International5(9/10): 341-345.

Efron B (1982)The jackknife, the bootstrap, and other resampling plans. Society of Industrial and Applied Mathematics CBMS-NSF Monographs.

Manly BF (1991) Randomization and Monte Carlo Methods in Biology. Chapman & Hall, London, UK

Owren MJ, Bachorowski JA (2007) Measuring emotion-related vocal acoustics. Handbook of emotion elicitation and assessment. OxfordUniversity Press.