I’m looking for objective criteria for classifying musical works.

Formal metadata (like artist, title, length, publication date, language etc.) doesn’t help in saying something about the actual kind of music; categorization in genres can help, but I guess they are typically subjective.

Is there some genre-like categorization that is not subjective?

No matter if computers (via audio analysis) or humans do the work, the important point is that

  • there can be no discussion if or if not a musical works belongs to one or several of the categories, and
  • each category is "useful" to a human, for learning something about the music, for finding similar works, etc.

Use case: Think of a huge database of musical works which should contain only verifiable metadata. Formal metadata is great, but it doesn’t help for grouping/finding similar works. Genres don’t work because they often lead to discussions.

  • 5
    Nothing is going to be purely objective and discussion free for the simple fact that sound creation is not an injective function and, even if it were, we cannot create perfect recordings. Can you attempt to categorize music based on, say, tempo (one of the most easily analyzed aspects)? Yes. Will any human or computer algorithm for detecting tempo be perfectly accurate and exactly the same as every other algorithm? No. – Matthew Read Feb 24 '15 at 19:42
  • @MatthewRead That comment seems to be very close to a good answer – Zach Saucier Feb 24 '15 at 20:24
  • 3
    @MatthewRead But that potential of error in measurement is found in exact sciences too, and has nothing to do with the objectiveness of the definitions. – Anton dB Feb 24 '15 at 21:05
  • 1
    @JCPedroza OP specified "no discussion" which means no error. Besides though, I'm not just talking about measurement error. It's impossible to decompose a combined waveform into its constituent parts without knowing exactly what those constituent parts are. – Matthew Read Feb 24 '15 at 21:08
  • 4
    @MatthewRead 1) You were speaking about measurement error exclusively on that comment. 2) It's not impossible, it is already being done at some degree, and as DSP evolves more and more things are becoming possible. 3) Even if it was impossible for a computer, it would be possible for a human, based on a set of very well defined and standardized definitions, to group music based on those shared characteristics. – Anton dB Feb 24 '15 at 21:17

there's no objective categorisation possible with genres anything like the ones we have now. What genre a song is is a matter for judgement, not measurement, and there's no central authority to arbitrate on the matter.

A few reasons why it's difficult to label songs accurately by the genres we know and love:

Genre labels are usually behind the times. In most cases, the first instances of songs that retrospectively come to be regarded as part of a genre weren't described as such at the time of their release, as the genre name hadn't been coined yet. And usually, early songs are not archetypal examples of a genre, but crossover works showing some elements of extant genres, and some novel elements.

Genre labels get re-defined. Look at what r 'n' b means now, compared to 50 years ago.

Genre labels are hierarchical - some genres are definitely subsets of others; yet others arguably so (and people will argue - look at the talk page on any Wikipedia page about a genre). This gives yet another dimension for disagreement.

...and yet they are not truly hierarchical either, because some overlap without one being a subset of the other (maybe 'Punk' and 'New Wave', for example)

Genre labels describe completely different dimensions of musicology. Looking at https://en.wikipedia.org/wiki/List_of_popular_music_genres...

  • some relate where the music is from ("K-pop");
  • Some are descriptions of the techniques used ("Dub") and the technology ("Electronica");
  • some have cultural references ("Christian country music");
  • Some relate to how seriously the music is intended to be taken ("Novelty music");
  • Some are named after record labels ("2 Tone", "Motown").
  • Some relate to how the music is listened to ("Dancehall").
  • ...etc.

So not every word that is a the name of a music genre is even talking about the same quality of the music - or even talking about a quality of music at all.

| improve this answer | |
  • 1
    But what you describe are flaws with the definitions of the genres, and not a counter-argument for objective categorization. You might not be able to use those definitions to objectively categorize, but you can still build your own definitions based on objective criteria, and categorize music through them. – Anton dB Feb 24 '15 at 21:46
  • You could objectively measure a great many characteristics of musical pieces, and place each piece at a point (or perhaps covering a range) on the spectrum of each characteristic. If you could somehow create some kind of broad agreement on what characteristics to measure, then you might be able to define some spaces within that n-dimensional matrix, and maybe agree on some labels. So yes, I agree that you could categorise music objectively. But not a single 'genre' we have is objective like that; so objective categorisation would have to be very different to genre-like categorisation. – user16 Feb 24 '15 at 22:00
  • 1
    I don't know if not one genre in the history of music has been defined objectively (as you claim), but even if that was the case nothing is stoping you from objectively defining one. Genre and subjectiveness are not inherently tied, and subjectivity is not a requirement for a genre definition. – Anton dB Feb 24 '15 at 22:19
  • Actually the label-based ones like 'motown' could be close to being objective, though people would still lump in similar music. @JCPedroza I do totally agree with you that objective categorisation is possible, so it's purely a question of how we're interpreting the word 'genre-like' and how we're weighting that word's importance in the question. Not really a substantive disagreement..? – user16 Feb 24 '15 at 22:27
  • I've softened the language a bit to make it clearer what I mean - hopefully! – user16 Feb 24 '15 at 22:38

As MatthewRead already mentioned a bit, music is art and art is the product of a free and creative human spirit. This can not fully be categorized or reduced to a few groups of genres.
We humans are able to create that much more different styles, that it is just impossible to find objective rules.
You can maybe distinguish a classical song from a rap, but even between those entirely different genres, there exist cross-over songs. You can mix everything and create a new genre out of it. All our modern western (pop, rock, disco, electro, jazz, rap, reggae) music is derived from the Blues (which has its roots in the spirituals and worksongs of the African slaves in North America in the 19th century) anyway, so they ARE very similar.

| improve this answer | |

Yes, you can categorize music objectively based on objective characteristics and patterns of a song. Two songs that share something can be grouped together.

Objective characteristics and patterns of a song include the tempo, rhythm, key, progression, instrumentation, timbre, harmony, melody, among others. If a genre has an objective definition, you can objectively determine wether a song belongs to it.

One problem is that genres are not always defined objectively, there's often a lot of subjectivity going on. Here there are no definitive answers, no one is necessarily wrong or right. But that doesn't mean that you can't categorize music, it just means that some definitions are not very useful, that you have to change or ignore them.

So, if you want to implement an objective categorization, you'll have to find purely objective genre definitions and/or build your own. It can be done.

| improve this answer | |
  • 1
    But genres are mixing, developing and newly forming. Most songs can't be just assigned to a specific genre, as they cross-over many genres. And some genres are defined very broad, like pop music. Most songs would somehow fit there... – Byte Commander Feb 24 '15 at 21:25
  • 2
    @ByteCommander But that doesn't mean that it's impossible to objectively categorize music. It might be not possible to do through some specific definitions, like your Pop example, but nothing is stoping us from implementing an objective categorization (maybe a different one, maybe one based on subjectively-defined genres). You don't have to use Pop as a genre definition, you can change it, you can ignore it: you can build a catalog of objectively defined patterns. – Anton dB Feb 24 '15 at 21:36
  • 2
    I don't think it is possible to create good objective categories. Two songs with equal speed, instrumentation etc can still sound completely different. For example they can induce entirely different emotions (aggressive, relaxed, happy, sad) for different listeners too, so that important attribute would also not be a good objective pattern. You can categorize music in defined and measurable categories, but in my opinion, they will not tell you much on how people will react on the song. – Byte Commander Feb 24 '15 at 21:47
  • 2
    Music transports emotion. This is a fact! There are sad songs, happy songs, scary songs, aggressive songs, ... Whether a song is happy or sad is much more important for the listener than the speed or rhythm. – Byte Commander Feb 24 '15 at 21:52
  • 1
    Now you are starting to confuse me. My opinion is that the art of music is not coverable by a set of categories. Music is more than we can objectively measure and therefore we can not really describe it objectively. It is possible to analyze a bunch of songs and compare some structure data to predict which ones of them might be more successful (mainstream), but we can not just create some categories and distinguish songs by them. They will be not important for the actual listeners! Only for statistics maybe, but nothing more. My opinion. – Byte Commander Feb 24 '15 at 22:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.