Helena Jahncke tested 113 employees from five offices before and after they moved into an activity-based workplace. Beforehand, those in shared and open-plan offices scored 14 per cent worse on a concentration-demanding task than those in cell offices, and an acoustician recorded the cell offices as 15 decibels quieter. After the move, the comparison ran within each person. Moving from the active zone, where noise was unrestricted, into the quiet zone raised performance by 16.9 per cent. Moving into an individual room raised it by 21.9 per cent.
Valtteri Hongisto built a model that predicts the performance loss from the speech transmission index, drawing on published decrements that run from 4 to 45 per cent depending on the task. Performance starts to fall once the index passes 0.2, and the worst of it is reached above 0.60. Anything that reduces intelligibility without reducing level therefore works. The paper reports no sample of its own, because it pools other people’s numbers rather than measuring anyone.
Ruifeng Yu ran 32 people through simulated X-ray screening at three levels of speech intelligibility: none, an index of 0.51 and one of 0.69. Performance at 0.51 was worse than at 0.69 and worse than in silence. At that level a listener keeps trying to resolve what is being said.
Eric Sundstrom surveyed 2,391 employees across 58 sites before and after office renovations. Fifty-four per cent said they were often bothered by noise, mostly talking and telephones. That disturbance correlated with dissatisfaction with the environment and with the job, and not with performance as rated by the workers themselves or by their supervisors. Elizabeth Sander reached the same shape in a simulated office in 2021: with 40 participants, open-plan noise did not reduce cognitive task performance against a quieter soundscape at all. What it did reduce was mood, measured by self-report, by facial expression and by heart rate and skin conductance.
Jahncke and Sundstrom worked in real offices with 113 and 2,391 people. Yu and Sander worked in controlled or simulated spaces with 32 and 40, and Sander’s participants had a mean age of 22.65 and were mostly students rather than office workers. Hongisto measured nobody and pools everyone.