How Many Floors Can Your Ears Tell Apart?

weirdly accurate.

    How Many Floors Can Your Ears Tell Apart?

    How many pitches can you tell apart?

    About five, for most people, when the tones come one at a time with nothing to compare them to. In Irwin Pollack’s 1952 experiment, listeners heard single tones spaced evenly in pitch and said which one each was; every tone carried about 2.3 bits of information, equivalent to telling only five tones apart without error. George Miller made the limit famous in 1956 in “The magical number seven, plus or minus two”.

    This test turns that experiment into a tower. Every floor has its own pure-tone ding, low at the bottom and high at the top, spread evenly in pitch over 250–2,500 Hz. Each building starts with a lift ride through every floor; then a ding sounds with no reference, you tap its floor and you are told the right one. The buildings grow from 2 floors to 3, 4, 6, 8, 12 and 16, always the same height, so the tones get closer. Two right in a row clears a building; the first miss means riding the same building again, the second ends the climb, followed by 16 more “rush hour” dings in that building. All your answers are fitted with the classic noisy-ear model of absolute identification, which gives the bits you transmit, the floors they are worth and a likely range.

    There is no percentile. Studies of this task tested a handful to about 120 people each, all in their own way, and none publishes how scores spread across many people, so your result sits next to landmarks, not a ranking. It is not a hearing test and says nothing about your health. A phone speaker, the room and the volume all move the number, and a few runs tell you more than one. It is not a perfect pitch test either: you need no note names or music training.

    What the test measures

    One thing: how many tones you can keep apart from memory alone, with no reference tone to compare them with.

    The task

    • Absolute identification, the task of Pollack (1952) and Miller (1956): hear one tone, say which of the set it was, with feedback after every answer
    • Seven buildings of 2, 3, 4, 6, 8, 12 and 16 floors; each floor a pure tone, 0.5 s long, the floors spaced evenly in pitch over 250–2,500 Hz
    • Every floor is matched for loudness (ISO 226) and then made randomly 0–6 dB quieter, so only pitch can give a floor away
    • A lift ride through every floor before each building; two right in a row clears it; two misses in the whole climb, the first a retake
    • Then 16 rush-hour dings in the building where you stopped, the one where floors start to blur

    Your result

    • The information each ding gives you, in bits, and the number of floors it is worth (up to 16), with a likely range
    • The tallest building you cleared, and a grid of where each rush-hour ding landed
    • Landmarks from the research: students in their first hour (Stewart, Brown & Chater 2005), Pollack’s listeners (about 5 tones), listeners before and after ten hours of practice (Dodds et al. 2011), five listeners after about 1,200 practice dings (Fulgosi & Žaja 1975)
    • In everyday terms: the tower spans about 40 piano keys, and how many keys wide each of your floors is
    • Patterns in your answers, such as leaning high or low, or being surest at the top and bottom floors
    • A shareable card and a link that challenges a friend

    Before you start

    • No note names or music training needed
    • Sound on; on an iPhone, switch off silent mode first
    • The lowest floor is 250 Hz, so a phone speaker can play every floor; a quiet room helps
    • About two minutes

    Pitch memory FAQ

    Is this test free? Do I need to sign up to see my result?

    Yes, it’s free, and there’s nothing to sign up for: no account, no email, no payment. Your result appears on this page the moment you finish, and you can share it or play again straight away.

    How many pitches can the average person tell apart?

    About five or six, when tones are heard one at a time with no reference. Pollack’s 1952 listeners transmitted about 2.3 bits per tone, equivalent to five tones identified perfectly, and Miller’s 1956 review put the ceiling at about 2.5 bits, roughly six equally likely tones. Adding more tones does not raise the number: in Pollack’s data, confusions became frequent from five tones on.

    What is the magical number seven, plus or minus two?

    “The magical number seven, plus or minus two” is the title of George Miller’s 1956 paper on how few categories people can keep apart when judging one thing, such as the pitch of a lone tone. For pitch, Miller read the limit from Pollack’s experiment: about 2.5 bits, or six tones. This pitch memory test runs the same kind of task, absolute identification, as a tower whose floors get closer together.

    How is the pitch memory test scored?

    The pitch memory test fits every answer you give with the classic noisy-ear model of absolute identification, then reports how much information each ding carries, in bits, and the number of perfectly distinguishable floors that is worth: 2 to the power of your bits, up to 16. The likely range comes from refitting 120 resamples of your answers. There is no percentile, because no study publishes a spread of scores across many people.

    Can you train your ears to tell more pitches apart?

    A little. In Dodds et al.’s 2011 study, six listeners practised with 36 tones a semitone apart for ten one-hour sessions and rose on average from 2.21 to 2.59 bits, about 4.6 to 6 tones. Fulgosi and Žaja’s five listeners reached a median of 3.11 bits, about 8.6 tones, after some 1,200 practice trials, and Dodds et al.’s best listener, who had music training, reached about 14.4 tones.

    Is this a perfect pitch test?

    No. People with perfect pitch can name 50 or 60 different pitches (Miller, 1956), using note names learned over years. This pitch memory test needs no note names: you learn a building’s floors on a single lift ride and then identify them from memory, which is the ordinary limit almost everyone has. If you know note names, the perfect pitch test on this site measures the other skill.

    Is this a hearing test?

    No. The pitch memory test measures how many tones you keep apart in memory, not how well you hear, and nothing in the result says anything about your health. The tones run from 250 to 2,500 Hz. A score that seems low is more often about the phone speaker, the room or the volume than about your ears; if you are worried about your hearing, a hearing professional can test it properly.

    Where the numbers come from

    The landmarks are taken from each study as published; floors are 2 to the power of the reported bits. All were lab studies on headphones with feedback, each with its own tones and amount of practice, so they are reference points, not a ranking. The site does not publish its own players’ data.

    How the score works

    Transmitted information

    How much one ding tells you about which floor it came from, in bits (Garner & Hake 1951; Miller 1956). Keeping 2 floors apart takes 1 bit, 4 floors 2 bits, 8 floors 3 bits; your equivalent floors are 2 to the power of your bits.

    The noisy-ear model

    Each ding is treated as heard at its true height plus random noise that is a fixed fraction of the whole range, with a possible constant lean up or down and a 2% chance of a random tap. The noise and lean are fitted to every answer you gave, in every building, by maximum likelihood.

    Why a model, not a raw count

    Counting bits straight from a short grid of answers reads high: the studies used hundreds to thousands of trials per listener, and a two-minute run has a few dozen. In simulations of this exact procedure the model came within about one floor. The raw count with Miller’s (1955) small-sample correction is still shown in the fine print.