Rating of eight different violins based on different attributes (rating criteria) using a paired comparison (psychoacoustic hearing test).
In the next stage of the project, eight violins with different quality levels were played by two different violinists. We chose pieces of music which were intended to make it easy to hear the different rating criteria. Based on three hours of music, short sequences were then chosen for the subsequent hearing test. The attributes we studied were as follows:
- Bright - Dark
- Nasal - Not nasal
- Pleasant - Unpleasant
- Even - Uneven
- Good response - Poor response
- Full of colors - Monotonous
- Passionate – Boring
The hearing trial was carried out using a comparison of pairs. For each of the pairs, the test subjects were asked to decide which sound best matched the attribute in question. There was an amazingly high level of consistency, i.e. the order of violins which matched each attribute in a better or worse way was consistent for the most part for all eight violins. Circular groups of three were used to determine the degree of consistency, i.e. the response combination A > B > C > A ("circular group of three") is considered to be "inconsistent".
On the other hand, it also turned out that in some cases individual test subjects treated the queried attributes consistently while their individual definition of the attributes was different. In other words, different people rate the instruments very consistently, i.e. they are consistent in terms of their own individual ranking. However, the rankings of the individual test subjects did vary considerably (depending on the attribute). It would be useful to carry out further hearing trials with a larger number of test subjects.
As a provisional result of the psychoacoustic hearing test, we will now have a look at the ratings of the eight instruments based on a few of the attributes:
BTL scale values of seven attributes for all eight violins. Computation based on the consistent responses by the test subjects. The network diagram for the scale values shows, for example, that Violin #1 (Stradivarius) was not considered to have a particularly bright sound (0.67), but the test persons did believe it sounded very pleasant (2.96) and had a good response (2.96). On the other hand, Violin #7 was believed to have a very bright sound (2.34) but it was not very pleasant (0.69) and did not have a good response (0.44). One can also see that there were no significant divergences when it came to assessing the "nasal" attribute.
The BTL scale values are from G. Gediga (1996). Script: Osnabrück University.
The recordings and hearing tests were carried out as part of our EU-funded research project with support from the company Schalltechnisches Beratungsbüro Müller-BBM. Special thanks to Dr. Karsten Zerbs, Dr. Otto Martner and Dipl.-Phys. Paul Geißler.