How Loud Is a School Classroom? Noise & Learning Impact

Three independent frameworks converge on almost the same number, which is unusual and worth noticing. ANSI/ASA S12.60 Part 1 sets a one-hour average background level of 35 dB(A) for core learning spaces up to 20,000 cubic feet, together with reverberation limits of 0.6 seconds — tightened to 0.3 seconds for rooms serving pupils with hearing loss.

The WHO's Guidelines for Community Noise recommend 35 dB LAeq in classrooms during teaching sessions. The UK's Building Bulletin 93 requires an indoor ambient noise level of 35 dB LAeq over 30 minutes for new primary classrooms and 40 dB for refurbishments, 30 dB for spaces serving pupils with very high sensitivity to noise, and caps discrete events at 60 dB LA1 over 30 minutes.

All of these are unoccupied criteria: they describe the room's inherent quietness, not the lesson. That is deliberate — you can design and verify a building, but you cannot design how loud a class of nine-year-olds is.

This is where the evidence thins out. We looked for peer-reviewed measurements of activity noise in occupied general classrooms with a stated method, and did not find figures we could verify. A large 2021 study in the Journal of the Acoustical Society of America surveyed 220 classrooms and reported that 74% met reverberation recommendations, but we could not extract its occupied-level values.

So rather than repeat the 60–70 dB figures that circulate without sourcing, we are stating the gap. What can be said with confidence: occupied levels are far above the 35 dB design criterion, they vary with activity and group size, and a room that fails its unoccupied criterion will be worse in every occupied condition too.

A peer-reviewed survey of 68 school gymnasiums in Portugal measured both states, and it is the clearest occupied-versus-unoccupied contrast we found in a school setting.

Unoccupied, the gymnasiums measured 30–59 dB LAeq with a median of 42 dB. Occupied during physical education, they measured 68–90 dB LAeq with a median of 80 dB — roughly 38 dB above the unoccupied median. Reverberation times ran 2.5 to 8.1 seconds (median 4.8), and speech transmission scores (RASTI) were 0.26–0.54 with a median of 0.34, which is poor.

Gymnasiums are not classrooms, and nobody should transfer the 80 dB median into a maths lesson. What the data does show is the scale of the occupied-versus-empty difference in a real school building, and how badly long reverberation damages intelligibility.

The reason standards care about a 5 dB difference in an empty room is speech intelligibility. What matters for a child is the signal-to-noise ratio between the teacher's voice and everything else, and reverberation smears speech over time as well as adding level.

The gymnasium data shows both effects at once: a median RASTI of 0.34 with reverberation approaching five seconds means speech is genuinely hard to follow, regardless of how loudly anyone talks. That is why BB93 and ANSI/ASA regulate reverberation alongside background noise, and why the standards tighten both for pupils with hearing loss.

School cafeterias come up constantly in this discussion, and the sourcing is weak. An Acoustical Society of America lay-language summary reports that some school cafeterias exceed 85 dBA, which is plausible and consistent with the gymnasium data, but it is a summary rather than a measurement report with a stated method.

We are including it flagged as low-weight evidence rather than as a figure to rely on.

There is a real limit here that is easy to miss: a 35 dB(A) criterion is close to the noise floor of consumer microphone hardware. Verifying an unoccupied classroom against BB93 or ANSI/ASA requires an instrument-grade meter with a low enough self-noise, and a phone or laptop is not that instrument.

Occupied levels, at 60 dB and above, are well within range of a calibrated consumer measurement. Uncalibrated, our meter reports A-weighted signal level in dBFS — a relative figure, useful for comparing group work against quiet reading in the same room on the same device. Calibrated once against a trusted meter, it reports an estimated dBA figure. Neither mode can be used for a compliance test.