Loudness, Fletcher-Munson, and Your Ears

Here is a thing that happens to everyone. You spend an evening on a mix. It sounds enormous — the low end is solid, the vocal is present, everything sits. The next morning you play it quietly while making coffee and it is a completely different record: thin, mid-heavy, no weight at the bottom, and somehow smaller than you remember.

Nothing changed. The file is identical. What changed is the volume, and your ears are not the same instrument at two different volumes. This lesson is about that, and it is probably the most immediately useful thing in the whole of the fundamentals.

Loudness is not intensity

Two words that get used interchangeably and mean different things.

Intensity is physical — how much acoustic energy is actually arriving, measurable with a meter, no human required.

Loudness is perceptual — how loud it seems. It depends on intensity, but also on frequency, on duration, on what else is playing, and on what you have been listening to for the last hour.

The relationship between the two is not proportional, and it is not close to proportional. A useful rule of thumb: you need roughly ten times the power to sound about twice as loud. That is a +10 dB change. Doubling the power — which is +3 dB, and which costs you a second identical amplifier — is a clearly audible change but nowhere near twice as loud.

This is why a 200-watt amplifier is not twice as loud as a 100-watt one, why adding a second PA cabinet is a smaller upgrade than people expect, and why the loudness war produced such diminishing returns for such enormous cost in dynamics.

The range being compressed

The quietest sound a healthy young ear can detect is a pressure variation of about 0.00002 pascals — twenty millionths of a pascal, a displacement of air smaller than the width of an atom. The loudest it can take before pain sets in is on the order of 20 pascals and above.

That is a ratio of about a million to one in pressure, and it is the entire reason the decibel exists. Nobody wants to write mix notes in pascals. So we compress that range logarithmically:

• 0 dB SPL — threshold of hearing. Not silence; the quietest detectable sound.
• 120 dB SPL — threshold of feeling. You stop only hearing it and start sensing it physically.
• 130 dB SPL — threshold of pain, though this varies between people, and tolerance is not the same thing as safety.

Whole numbers instead of long strings of decimals, and a scale that happens to line up reasonably well with how perception actually behaves. That is the whole trick of the decibel, and it gets a course of its own later.

Fletcher and Munson, 1933

In 1933, two researchers at Bell Labs, Harvey Fletcher and Wilden Munson, ran an experiment that should be on the wall of every studio. They played listeners a reference tone at 1 kHz, then tones at other frequencies, and asked them to adjust each until it sounded equally loud as the reference. Then they wrote down the actual physical levels required.

The result was a set of curves — one for each reference loudness — showing what a tone at any frequency has to measure in order to seem as loud as the 1 kHz reference. These are the equal-loudness contours, still universally known as the Fletcher-Munson curves.

They are not flat. They are not close to flat. Reading them tells you three things:

1. We are far less sensitive at the frequency extremes. Low bass and extreme treble have to be physically much louder to seem equally loud. At the bottom of the spectrum the difference is dramatic — tens of decibels.
2. We are most sensitive between roughly 1 and 5 kHz, with the peak around 4 kHz. This is exactly the ear canal resonance from the previous lesson. The anatomy and the psychoacoustics are the same fact seen from two directions. And it is where speech consonants live, which is unlikely to be a coincidence.
3. The curves change shape with level. This is the crucial one, and the reason this lesson exists. At high listening levels the curves flatten out considerably. At low listening levels they become far more exaggerated, particularly at the bottom end.

That third point, restated in the way that matters to you: the frequency response of your hearing changes depending on how loud you are listening. You do not have one set of ears. You have a different set at every monitoring level.

Phons, sones and weighting

Three pieces of vocabulary that come straight out of those curves.

The phon is a unit of perceived loudness. A sound has a loudness of 40 phons if it seems as loud as a 1 kHz tone measured at 40 dB SPL. It ties subjective loudness back to a physical reference, which decibels alone cannot do — 40 dB SPL at 60 Hz and 40 dB SPL at 3 kHz are the same measurement and nothing like the same experience.

The sone goes further and tries to make the scale proportional, so that two sones genuinely sounds twice as loud as one — useful in research, rarely encountered in a studio.

Weighting curves are where this reaches you in practice. Sound level meters offer A, B and C weighting, and these were derived as approximate inverses of the equal-loudness contours at roughly 40, 70 and 100 phons respectively. The meter deliberately mis-measures in exactly the way your ear does, so the reading corresponds to what a person experiences.

• dB(A) rolls off the low end heavily. It models quiet listening, and it is what noise and hearing-safety regulations are written in.
• dB(C) is much closer to flat. It models loud listening, and it is the right choice for measuring anything with real low-frequency content — a club, a PA, a cinema.

Which is why the same sound system can measure 95 dB(C) and 85 dB(A). Neither is wrong. Quote the wrong one and you can understate a bass-heavy system's actual output by ten decibels, so always say which weighting you used.

The part that costs people money

Here is the whole practical consequence, and it is worth reading twice.

Because the curves flatten at high levels and exaggerate at low ones:

• A mix balanced loud — above about 85 dB SPL — will sound bass-light and dull when played back quietly. At high level your ears gave you the low end for free, so you did not add enough. Take that free gift away and the bottom falls out.
• A mix balanced quiet will sound bass-heavy and harsh when played back loud. At low level your ears hid the extremes from you, so you overcompensated. Turn it up and the compensation is suddenly audible and excessive.

Neither mix is wrong in the room it was made in. Both are wrong everywhere else, which is the problem, because you do not get to choose where anyone listens.

There is a further trap on top of that: everything sounds better loud. Not more accurate — better. The frequency balance seems fuller and more exciting because the contours have flattened, so more of the spectrum is arriving at comparable perceived level. This is entirely real, and it is why the demo room at a hi-fi shop is louder than your living room, and why the mix you fell in love with at midnight was partly just loud.

So what do you actually do?

Three things, and they are not complicated.

1. Pick one moderate reference level and mark it. Somewhere in the region where you can hold a normal conversation over the music. It does not need to match any standard; it needs to be the same every single day, so that your ears' response is a constant rather than a variable.
2. Check at more than one level, always. Loud for a moment to confirm the low end holds together and nothing is unpleasant. Quiet for a long moment to check the balance — quiet listening is brutally honest about arrangement, because anything that vanishes was never carrying its weight.
3. Treat very loud listening as a specific diagnostic tool, not as a working state. There are things you can only hear loud. Go there, listen, and come back down.

One more piece of history that makes sense in this light: the loudness button on old hi-fi amplifiers. It is not a volume control — it boosts the lows and highs specifically to compensate for the equal-loudness contours at quiet listening levels. It exists because the manufacturers understood exactly this problem in the 1960s. A volume control is frequency-neutral. A loudness switch is Fletcher-Munson compensation with a marketing name.

Worked example: the same mix, twice

Take one arrangement — kick, bass, guitars, vocal, cymbals — and balance it twice.

Mix A, balanced at 90 dB SPL. At that level the contours are relatively flat, so the low end arrives generously without help. You set the kick and bass where they sound right, which turns out to be a fairly modest amount. The 3 to 5 kHz region feels aggressive at that volume, so you pull the vocal presence back a little, and you gently tame the cymbals.

Now play Mix A at 65 dB SPL. Your low-frequency sensitivity has collapsed relative to the mids. The kick and bass, which were only modestly present, now sit well below the point where you can register them properly — the track sounds thin and mid-forward. The vocal presence you pulled back is now missing, so the words feel indistinct. The cymbals you tamed have vanished. The mix sounds small.

Mix B, balanced at 65 dB SPL. You push the kick and bass up until you can feel them at that quiet level, and you add presence to the vocal so it stays intelligible. Everything sits nicely.

Now play Mix B at 90 dB SPL. The bass is overwhelming, because you added a large amount to compensate for a deficiency that no longer exists. The vocal presence boost is now sitting on top of a band your ears are already emphasising, and it is harsh. The mix sounds bloated and fatiguing.

Same engineer, same room, same monitors, same arrangement. The only variable was the monitoring level, and it produced two mixes with genuinely different frequency balances — neither of which travels. That is the entire argument for a fixed reference level, and it is why professional rooms have one.

What loud listening actually costs

The previous lesson described the outer hair cells as an active system doing real mechanical work to sharpen your frequency resolution. Active systems tire, and they can be destroyed. Here is what that looks like.

Temporary threshold shift

After a period of loud exposure, your hearing threshold rises — quiet sounds you could detect before are now below your floor. Everything sounds slightly dull and slightly distant. This usually recovers within about 24 hours, though after a heavy exposure it can take a week.

The mechanism is worth knowing because it explains the specific damage. Loud sound triggers constriction of the blood vessels feeding the organ of Corti, reducing the blood supply to the hair cells. The outer hair cells — the active, sharpening ones that handle low-level detail — are the most easily saturated and the most starved. The inner cells, which need higher intensity to fire anyway, carry on.

So what you lose first is not loud sound. It is quiet detail — which is precisely the thing you are paid to hear. A temporary threshold shift does not make you deaf; it makes you a worse engineer for a day, and you will not notice, because everything including your reference tracks is shifted equally.

Frequencies in the 2 to 6 kHz range induce the most temporary shift — the same band your ear canal is already amplifying by 20 dB. The anatomy that makes you sensitive there is the anatomy that gets damaged there first.

Chronic and permanent

If exposure repeats without enough quiet time to recover, the shift becomes chronic — a partial recovery that never quite completes, with each exposure starting from a worse baseline. Continue, and outer hair cells that never get the chance to recover gradually lose function and die.

They do not grow back. That is a permanent threshold shift, and there is no treatment. Hearing aids amplify; they do not restore resolution, which is the thing that was actually lost.

Susceptibility varies genuinely between individuals, so the fact that a colleague has mixed loud for twenty years without obvious damage tells you nothing about your own risk. Tinnitus — ringing or whistling after exposure — is the clearest warning sign there is, and it should be read as an instruction rather than an inconvenience.

The numbers

Occupational standards are built around 85 dB(A) for an eight-hour day, with exposure time roughly halving for every 3 dB above that. Applied honestly, that is confronting:

• 85 dB(A) — 8 hours
• 91 dB(A) — 2 hours
• 94 dB(A) — 1 hour
• 100 dB(A) — 15 minutes
• 103 dB(A) — under 8 minutes

A loud club runs comfortably above 100 dB(A). A rehearsal room with a live drummer sits around 100 to 110. A monitor wedge at head height can exceed 110. Under these figures, a single unprotected night out uses up considerably more than a week's safe exposure.

None of which is an argument for never being in a loud room — it is an argument for owning a decent pair of filtered earplugs, the kind that attenuate roughly evenly across the spectrum rather than just removing the top. They cost very little, they let you still hear the music properly, and they are the cheapest career insurance available to anyone in this field.

The short version

Your ears' frequency response changes with level, so the monitoring level you choose silently determines the frequency balance of everything you make. Pick one, mark it, keep it, and check your work quiet as well as loud.

And protect them. Everything else in this curriculum — every technique, every plugin, every room — is in service of a pair of sensors with about 16,000 irreplaceable cells in them, which respond to abuse by quietly taking away the fine detail first.

What to practice

• Set a monitor level with an SPL meter (a phone app is close enough) and mark it on the knob. Use only that mark for a week and notice how much more consistent your mixes get.
• Balance a mix at your normal level, then drop to near-whisper and list every element that disappears. Those are the elements you have been mixing with your ears' help rather than with the fader.
• Play the same track at 70 dB and at 90 dB SPL and write down how the bass and the top end change. Nothing about the file changed.
• Measure the SPL of the last venue, club or rehearsal room you were in, then look up the safe exposure time for that level. Most people are startled.
• At the end of a long session, play a reference track. If it sounds dull, do not fix your mix — stop working.