The results[2] are published in an article behind a paywall, but they give details of the two tasks presented to the subjects, neither of which involved actually writing code. One was a 15-minute task of locating syntax errors, another was a 20-minute task of locating logic errors.
[1]: https://www.gwern.net/docs/music-distraction/2005-lesiuk.pdf [2]: https://www.tandfonline.com/doi/abs/10.1080/08874417.2000.11...
If I need a mood boost or something to get my body "moving" in my seat and keep me from mentally wandering, I'll throw on stuff I really know and love. Usually I'm doing menial tasks or simple coding.
If I'm having a hard time focusing and need a distraction for some parts of my brain but not all of it, I have a "deathcore" (really more like melodic (tech) death metal) playlist where most of the songs sound very similar and you can't understand a thing the vox are saying.
Typically if I'm super tired AND need to block out the outside world but even hard to understand screaming won't cut it, I have to resort to either instrumental metal or LoFi hip hop beats to study to.
If I'm currently listening to music, I really only do a full stop/pause to focus if I notice I've run a breakpoint three times and have no idea what I've been stepping thru, or if I'm trying to refactor or write a fresh, new logic flow and just can't get my brain focused enough to "check it" for all possible paths. That's when I cut the feed, double-time my focus, and once I'm convinced it's "safe" I can hit play again.
Brains are weird and finicky I tell ya.
On the other hand, there are several other factors, like how tired I am, how hard the foreground task is etc., so it is hard to pin-point it specifically to the lyrics language.
If I've just started both are probably going to be good results. If I'm 4 hours in the music one will probably do better because I've not wandered away from my desk distracted by a shiny pen.
[ETA]
According to [1] it's described in "Peopleware"[2]. Apparently needs corroboration (see sibling reply by 'civilized)
[1] https://www.quora.com/Does-listening-to-music-while-programm...
[2] https://en.wikipedia.org/wiki/Peopleware:_Productive_Project...
The effect of music depends on the person, the type of music (lyrics or instrumental - even if there is applause or not), how well the person knows the music and, of course, the task itself and its complexity.
Until there's a way to test all of those permutations, I think the best we are going to do is rely on all our anecdotal evidence, as it affects ourselves.
I personally feel like there are different components of my brain and some of them just want to do something, and if I don't "distract" them then they do something arbitrary and maybe in turn distract me... but if I need them, and they are busy deciding some music I am listening to, I lose out on the maybe-even-often subconscious processing they do that gives me valuable insight. I thereby find some (mostly useless) tasks easier if I am seriously even playing a video game with a targeted skill while doing them (people hate me in meetings sometimes ;P), or maybe listening to music, but if I am doing the hard coding work I "should be doing" I just can't :(.