Addressing Nocturnal Disturbances: A Plea for a Conducive Study Environment — Epoche C1
To the Residence Hall Management Office, I am writing about noise reaching Room 412 of Building C between 22:00 and 03:00, which over the past two months has made both concentrated study and adequate sleep unreliable. I have written to you once before on this matter in more general terms, and I am writing again because that letter asked for the wrong things. Having looked into what is actually known about noise, partitions and sleep, I can now say which of my earlier proposals is likely to work, which is close to useless, and which one I dismissed as a minor measure although the physics makes it the most effective item on the list. I would also like to propose that we replace the whole dispute about whether the noise is excessive with a measurement. The nature of the problem, stated more usefully The disturbances recur nightly and fall into four kinds: conversation and laughter from the corridor, common areas and adjacent rooms; amplified music, of which the low frequencies dominate what reaches me; door closures and foot traffic; and late activity in the communal kitchen. That list has not changed. What has changed is my understanding of which items on it matter and why, and it turns out that they matter for two entirely different reasons, requiring two different remedies. The first reason concerns study, and it is not loudness. Why loudness is the wrong measure for the study problem Pierre Salame and Alan Baddeley (1982) established the finding that governs this. They had participants memorise short sequences of visually presented digits and recall them in order, while irrelevant sound was played that participants were instructed to ignore. Unattended speech substantially impaired recall. Crucially, it did so even when the speech was meaningless to the listener, and the degree of disruption varied with how phonologically similar the irrelevant material was to the items being remembered — while continuous noise at a comparable level produced far less disruption or none. Their interpretation was that speech gains obligatory access to the same short-term store that holds the material being rehearsed, and corrupts it, whereas noise does not. The practical consequence is that the noise which damages study is not the loud noise. It is the intelligible noise. A conversation at conversational volume through a wall interferes with reading and revision more than considerably louder sound of a non-speech kind, and this is why I can work beside a fan or a boiler and cannot work beside two people talking three metres away through a door. Simon Banbury and Dianne Berry (2005) found the same ordering in field conditions rather than the laboratory, surveying office workers about what disrupted them and testing habituation directly: conversation was reported as the most disruptive component of office noise, and the disruption did not disappear with familiarity — habituation to the effect on performance was limited even where the reported annoyance settled down. This matters for policy, and it is the first place my earlier letter went wrong. A quiet-hours rule enforced by asking people to turn things down is a rule about sound level. It will not touch the noise that is doing the most damage to my work, because that noise is already quiet. If the aim is a residence in which students can revise in their rooms, the target is speech reaching those rooms in an intelligible form, and the only two ways to reduce that are to attenuate it or to mask it. Why the low frequencies get through, and why my headphones are not the answer My earlier letter said that bass penetrates walls and floors so that concentration is impossible even with noise-cancelling headphones. The first half is correct and has a straightforward explanation, which I set out because it bears directly on which building measures are worth funding. The second half is the reverse of the truth and I should withdraw it. Sound pressure level is reported in decibels as $L_p = 20\log_{10}(p/p_0)$, where $p$ is the sound pressure and $p_0 = 20\,\mu\mathrm{Pa}$ is the reference. The quantity that describes a partition is its transmission loss: the difference in level between the two sides. For a simple limp panel, the governing relation is the mass law, given in standard treatments such as Marshall Long's (2014) text on architectural acoustics in the normal-incidence form $$\mathrm{TL} \approx 20\log_{10}(m f) - 47\ \mathrm{dB},$$ where $m$ is the surface mass of the panel in kilograms per square metre and $f$ is the frequency in hertz. The structure of that expression is what matters here. Because $20\log_{10} 2 \approx 6$, transmission loss rises by about $6\,\mathrm{dB}$ for each doubling of surface mass, and by about $6\,\mathrm{dB}$ for each doubling of frequency. Apply the second of those. Speech occupies roughly the range from 200 to 4000 hertz; the fundamental of a bass line sits near 60 hertz. From 1000 hertz down to 62.5 hertz is four doublings, so the same partition delivers about $4 \times 6 = 24\,\mathrm{dB}$ less attenuation at the bass frequency than at the speech frequency. A wall that gives a serviceable $40\,\mathrm{dB}$ against conversation gives something in the region of $16\,\mathrm{dB}$ against the bass, and lightweight stud partitions of the kind used in most residence buildings perform worse than the mass law predicts at low frequencies because the cavity resonates. The bass is not getting through because it is loud. It is getting through because the wall is, in effect, four times thinner to it. Now the correction. Active noise-cancelling headphones work by measuring the incoming sound and generating an inverted waveform to cancel it, and this process is most effective at low frequencies — below roughly a kilohertz — because long wavelengths make the phase relationship stable across the space around the ear. They are poor at exactly the sounds this letter is most concerned with: speech consonants, which are high-frequency, and imp