My favourite of the answers supplied to date (over on Quora, where this question was posed anonymously) is Earl Dingman‘s. As he states, the most important effect of changing key signature, is the resulting change in difficulty for some instruments as opposed to others… their “performability”.
There are other, subjective factors (here’s a fascinating list showing how different keys were “felt” by European composers around the time Equal Temperament came to be a widely-held standard… and of course Spinal Tap’s & Schubert’s insistence on D-minor as the saddest key)… but none is as clear and demonstrable as “performability”; we can definitively compare the level of technical ability it takes to play a tune convincingly on different families of instruments:
- traditional string instruments are tuned such that it’s easiest to play them in keys that are natural or slightly sharp (particularly C, G, D)… this is why so many old fiddle tunes are found in those keys.
- wind & brass instruments tend to be easiest to play in slightly flat keys (particularly F, Bb, Eb)… this is why so much of Louis Armstrong & early/Hot Jazz repertoire was recorded in those keys.
- keyboards (piano, harpsichord, organ, etc) are highly versatile as a result of having been invented with the concept (of musical keys) in mind; the pattern (7 white keys + 5 black keys), was deliberately chosen to democratise the musical keys, ironing out preferences as far as possible. There is however a mild prevalence of keys with 1–3 sharps OR flats.
- different voices will of course be more or less comfortable, perhaps have to strain, playing in different keys… and at different times of day! If I ask you to sing the “Ba-dee-ya” of “September” by Earth, Wind & Fire, at 6AM or 6PM, you’re likely to give a very different performance each time.
- cheat instruments such as the fretted strings (guitar, bass, mandolin, etc) for which you can buy a $5 capo to artificially shorten the neck… are still affected tonally in that the higher the capo is placed, the less fullness of body you’ll get in the sound (especially when strumming with a pick).
It’s when we get into mixed instrumental groups, such as an orchestra or bands like Vulfpeck (these key-changes were applied in post-production rather than performed live, but you get the idea!), that we see “performability” becoming more complicated by the fact multiple instrument families are represented. Perhaps in this case there’s an argument for picking key based on something less obvious, like performance venue: should we be able to re-tune our streams so that the same music played on AirPods could “feel” as acoustically close to the experience we’d have had hearing it performed in the Carnegie Hall? The standing waves & resonances encountered in each space, will be very different… and how about electronic music, where there are no performers and/or instruments to consider?
As we start to look beyond the “performability” (ease/difficulty of performing the music on a given instrument), this widely-shared article by Ria Misra provides a fascinating deep-dive into the prevalence of different keys across different genres:
The data came from Spotify’s API, and it’s only with technologies of this kind that we’re becoming able to survey en masse the preferences of listeners, as part of the attempt to determine whether key signature has any effect on music other than the ease/difficulty of playing on particular instruments. Here’s to the future, and new technologies coming online to enable more nuanced analyses going forward!