Lokki, T. (2014). Tasting music like wine: Sensory evaluation of concert halls. Physics Today, 67(1), 27.
A few months ago, I read this article called “Tasting Music like Wine: Sensory evaluation of concert halls” by Tapio Lokki and was fascinated by two things:
– The lightness of the article and how it introduced such a complex topic as concert-hall acoustics with an anecdotal situation;
– The methodological intrincacy with all the 3D sounds recorded in such an enginious way (all orchestra musicians were recorded solo, placed 24 columns in a stage, each column playing only one instrument, and recorded the full “orchestra” in several places of the venue. Very simply put.)
– After all, there were three interesting things: I loved the use of Wine tasting know-how for the evaluation of the subjective experience of concert-halls.
So the situation is the author and his wife are listening to a concert while drinking some wine. While the wife enjoys the concert but not so much the wine, the author felt totally the opposite. Both perceived wine and music differently. After some thought, the author concluded that both wine and music have a lot in common, because each can be characterized by a multidimensional array of perceptual attributes.
Both are a matter of personal taste, and each person may concentrate on different aspects of the taste or sound. The thing is, winemakers have a solution for this, and have since long developed techniques to determine what makes good or bad wine.
Like the aroma wheel.
The first question than is: could these methods be tailored for the perceptual evaluation of concert-halls?
The wine tasting methods like sensory profiling demand comparison of samples, that is, imagine you have a table with a line of glasses, all with wines different from each other, and you may and must drink a sip from one and the other as many times as you, as an assessor, find necessary. Could this be made with sound?
The answer is yes, and please read the original article to find out how.
In winetasting two methods are used to gather attributes of wines: consensus vocalubary profiling, when a number of assessors reach a number of consensual adjectives for each wine; and individual vocabulary profiling – the one used in this work – where a number of assessors (usually 15 or more) salient which charactersitics can be found in the wine.
The first experiment had 20 listeners, and all heard 3 recording positions out of 3 Finnish concert-hals. Together, they’ve suggested 102 attributes. After clustering the data, one cluster (overall volume and perceived distance) managed to explain more than 50% of the variance.
The second experiment had only one distance – 12m from the stage -, 9 halls and 17 assessors. They’ve suggested 60 attributes clustered in 7 groups.
After more analysis (hierarchical multiple-factor analysis), it was possible to distinguish two groups out of this last evaluation (after ordering by preference also): one group preferred intimate sound in which they could easily distinguish individual instruments and lines, and another group which preferred louder and more reverberant sound with good envelopment and strong bass.
Very impressive how it was possible to understand this information. Would portuguese listeners make the same evaluation?