Why RSA events feel like networking but produce no network
Feber · RSA Connect · 2026
The Lombard effect: ambient noise causes speakers to raise their voices, which raises ambient noise further. Apparent networking is acoustic proximity, not network flow.
The Lombard EffectÉtienne Lombard (1911) · acoustic analogy for DVAP
Walk into an RSA event. The room is full of Fellows. Conversations overlap. There is energy, presence, a sense of connection. The ambient noise says: things are happening here.
But what you are hearing is the Lombard effect. As each speaker raises their voice to be heard above the noise, the noise rises further, triggering further compensation. The feedback loop produces the sensation of intense connection while delivering almost none. Acoustic proximity is not network flow. A room full of voices is not a room full of signal.
Strip away the acoustic illusion and the structure becomes visible. Three clusters of Fellows: people who already know each other, drawn together by disciplinary overlap, geography, shared project history. Dense within clusters; empty between them.
The cross-cluster connections that would make this a network, the bridges that would carry signal from one cluster to another, do not exist. The room has no routing. It is a crowd with the acoustic appearance of connection. The network test fails.
ΔS = k × ΔN
ΔS — change in speech level (dB)
ΔN — change in noise level (dB)
k — Lombard coefficient, fixed at 0.4
Δ (delta) stands for "change in" a value. The decibel scale is logarithmic: a 10 dB rise = a tenfold increase in sound intensity.
As more people enter the room, sound pressure rises. Each speaker compensates by raising their voice, which raises the ambient noise further, triggering further compensation. The attempt to communicate accelerates the breakdown of communication.
People in the room10
Room volume (m³)150
Live readout
ΔS = 0.4 × 10.0 dB
10 people → noise +10.0 dB → speech +4.0 dB
Below roughly 12 to 15 people, the room works. Signal reaches its intended recipient. Knowledge flows. Above that threshold, the Lombard feedback loop takes hold and intelligibility collapses — even though more people are present, more expertise is in the room, and more conversation is happening. The knowledge does not disappear. It is still carried by every Fellow present. The noise contains knowledge we simply cannot extract.
This is a complexity problem. Human cognitive and perceptual apparatus works brilliantly up to certain sound pressure levels. Beyond those levels, the same capacity that worked perfectly begins producing noise instead of knowledge. The transition is gradual, and that is precisely the point of the Lombard effect: speakers do not notice that they are compensating. Voice volume rises involuntarily, which raises the ambient noise, which triggers further compensation. By the time anyone registers that the room has become difficult, the feedback loop is already well established. The knowledge does not disappear. It is still carried by every Fellow present. The noise contains knowledge we simply cannot extract.
The volume of the room is a direct measure of the absence of network architecture. If Fellows could find the right person to talk to — if there were a system routing signal reliably to its intended recipient — they would not need to shout.
The threshold depends on room volume: a larger space absorbs more sound energy, allowing more people to converse before the Lombard feedback loop dominates. Move the room volume slider to see the cognitive limit shift.
How the Lombard Effect worksÉtienne Lombard · Annales des Maladies de l'Oreille et du Larynx · 1911
The coefficient (k = 0.4).
The Lombard coefficient measures the ratio of speech-level increase to noise-level increase. A coefficient of 0.4 means that for every 10 dB rise in ambient noise, a speaker involuntarily raises their voice by 4 dB. This is not a choice. It is a reflexive neurological response: the auditory feedback loop that monitors one's own voice detects that it is being masked, and the vocal system compensates automatically. The speaker does not notice. The compensation is below conscious awareness, which is precisely why the effect is so difficult to resist.
Human hearing and the decibel scale.
The decibel scale is logarithmic, not linear. A 10 dB increase represents a tenfold increase in sound intensity. Normal conversation sits at roughly 60 dB. A crowded room reaches 75 to 85 dB. At 85 dB, speech intelligibility drops sharply: the listener must be within approximately one metre to follow a conversation, and even then requires sustained concentration. The human auditory system is exquisitely sensitive (capable of detecting pressure variations of less than a billionth of atmospheric pressure) but it has hard limits. Above certain sound pressure levels, the same perceptual apparatus that works brilliantly in quiet conditions begins to fail, and the Lombard reflex makes it worse rather than better.
The diagnostic: noise as the absence of network.
The volume of a room is a direct measure of the absence of network architecture. In a well-connected network, signal reaches its intended recipient through designed routing: the right person finds the right conversation through the system rather than through the accident of physical proximity. When Fellows must shout to be heard, it means the system has no routing. Every Fellow present carries knowledge, expertise, and potential connections, but the information is trapped in the noise. The noise contains knowledge we cannot extract. This is not a social failing; it is a structural one. The RSA has 31,000 Fellows and no mechanism for directing signal. The Lombard effect is the audible symptom of that structural absence.