One experiment, three conditions
In 2023 Souza and Leal Barbosa at the University of Porto tested music with and without lyrics on four tasks: remembering words, remembering visual patterns, reading comprehension and arithmetic. Each participant did every task in silence, with instrumental music and with music with lyrics. The instrumental tracks were lo-fi hip-hop. The lyrical ones were four Portuguese pop songs from the national chart, heard by Portuguese psychology students.
Music with lyrics hurt verbal memory, visual memory and reading comprehension, by a small amount, an effect size of about 0.3. Its effect on arithmetic was not reliable. The instrumental music did not reliably help or hurt anything.
The students’ own judgements were partly accurate. They noticed that the lyrics were costing them. Some felt the instrumental music was helping, which the scores did not show.
The older finding behind it
The result fits a long line of work on what psychologists call the irrelevant speech effect: speech you are not trying to listen to still interferes with reading and with holding words in mind. A 2026 meta-analysis by researchers at Tianjin Normal University combined 30 studies of background speech during reading, with 606 effect sizes in total. Both intelligible and unintelligible speech slowed reading. Intelligible speech, which the listener can understand, slowed it more and delayed the processing of individual words at every stage.
That has a practical implication the studies themselves did not test directly. Lyrics in a language you understand are likely to be more disruptive than lyrics in one you do not, and both more than no words at all.
Why speech interferes is still argued over. One account says the problem is meaning competing with meaning; another says it is the sound of speech itself. The meta-analysis found that neither fully explains the data, and its authors favour a third account, which they call the Process Interference Hypothesis, as the more complete explanation.
The office version
The same pattern appears outside the laboratory. A 2017 analysis by the Finnish Institute of Occupational Health combined surveys and acoustic measurements from 21 open-plan offices and 883 people. Distracting background speech largely explained how disturbed people felt by noise overall. A playlist with lyrics adds one more voice to a room that may already have several.
Limits
The Porto authors were careful about their own study. It was not large enough to detect very small effects, so it cannot rule out a slight cost from instrumental music. The lyrical and instrumental tracks differed in more than the presence of words. The reading task was easy for the sample, and most participants were young women studying psychology.
None of this shows that music is bad for work in general. The evidence is specific to tasks involving words and memory. For reading, writing and learning, quiet or wordless sound is the safer choice. For routine tasks, the studies here say little either way.