The task is the biggest factor
Research on background sound and cognition points fairly consistently in one direction: music interferes most with tasks that involve language and sequence.
Reading comprehension, writing, and anything requiring you to hold a sequence in mind are reliably impaired by music with lyrics. The competition is direct — your language processing is already occupied, and the lyrics are asking for the same resource.
Repetitive, low-demand tasks are different. Data entry, tidying files, routine admin: here music often improves performance, largely by keeping arousal up and boredom down.
The practical rule: the more your task uses words, the fewer words your audio should have.
The irrelevant speech effect
There is a well-documented finding that background speech disrupts short-term memory for ordered material, whether or not you are paying attention to it. You cannot simply decide to ignore it.
This is why an open-plan office can feel so much harder than a noisy street. Traffic is loud but meaningless. Half a conversation is quiet and impossible to tune out. It is also why podcasts are a poor choice while writing, however interesting.
Silence is not always the winner
The intuitive assumption is that silence must be optimal. Two reasons it often is not.
First, complete silence is rare. What people call silence is usually an unpredictable soundscape — a door, a car, someone's footsteps. Unpredictable sound is more disruptive than steady sound, because each new event captures attention. Consistent background audio masks those interruptions.
Second, some people underperform in silence because arousal drops too low. This varies considerably between individuals, and it is a large part of why the research literature looks inconsistent — studies are averaging across people who respond in opposite directions.
What tends to work, by category
Instrumental music
The reliable middle ground. No lyrics means no language competition. Familiar tracks work better than new ones — novelty pulls attention.
Ambient and drone
Minimal melodic movement, nothing that resolves or surprises. Excellent for demanding work because there is almost nothing to attend to. Some people find it flat.
White, pink and brown noise
Pure masking with no musical content. Some evidence suggests moderate noise can aid performance for people who are more distractible, though findings are mixed. Brown noise — deeper, less hissy — is easier to tolerate for long stretches than white.
Binaural and frequency audio
Sits between music and noise. The evidence for a specific neurological effect is genuinely mixed, but as consistent, lyric-free background audio it does the masking job well. If the ritual helps you start, that is a real benefit even setting the entrainment question aside.
Lyrical music
Fine for routine tasks and exercise. Poor for reading, writing, or learning anything new.
Volume matters more than people expect
Studies on noise intensity and creative cognition have found that moderate levels can support performance while loud levels impair it. Around 50–60 dB — audible but not dominant — is the usual sweet spot. If you are noticing individual notes, it is probably too loud.
A useful ritual effect
Beyond the acoustics, using the same audio consistently for focused work builds an association. After a few weeks the track itself becomes a cue that a work session has started. This is ordinary classical conditioning, and it is one of the more reliable benefits available here — but only if you keep the audio consistent rather than shuffling.
Working out your own answer
Pick one task you do regularly and measurably. Try it for a week in silence, a week with instrumental audio, a week with noise. Record something countable — words written, problems solved, minutes before your attention broke.
A week each is enough to see a pattern, and the answer you get for yourself is worth more than any average across study participants.