Music Hub › Guides › Getting started › Finding the chords of a song by ear

The bass line: the easiest voice to isolate by ear

Faced with a song whose key is already known but not its chart, the temptation is to dive straight into the full chord as it sounds at first pass, with all its voices mixed together, its melody, its harmonies, and sometimes several instruments playing at once. That's almost always the wrong way in. The right one is the bass line, for a reason that owes as much to the physics of sound as to how the ear works: on most recordings, the bass occupies a low register that's almost entirely its own, far from the crowded swirl of mid and high frequencies where the voice, guitar and piano jostle for space. An isolated low note, usually played one at a time, is mechanically easier to pick out by ear than a four- or five-voice chord stacked in a register where everything overlaps.

This isn't just a listening convenience: it reflects what the bass actually does in the vast majority of styles. The Music Hub guide on building a bass line from a chord chart lays it down as the very first universal reflex, whatever the style: play the root of the chord on the first beat of the bar. Seen from the transcriber's side rather than the composer's, that habit flips around exactly in reverse and becomes a detection method: the lowest note you hear at the moment the chord changes is, most of the time, that chord's own letter name. Spotting that note alone is enough to give its letter name, even before knowing whether the chord is major, minor, or carries a seventh.

That clue does have a limit worth knowing before trusting it blindly. The bass doesn't always play the root: a chord written as a slash chord puts a different note in the bass, the third, the fifth, or sometimes a seventh, precisely to smooth out a progression or set up a pedal bass that stays fixed while the harmony changes above it. A bass note that seems to contradict the chord you otherwise think you're hearing isn't necessarily a sign of a mistake: it may simply signal an inversion, a case the guide on building a bass line covers in its own right, notation and uses included. At the transcribing stage, it's enough to keep that possibility in mind without dwelling on it too long: if the rest of the chord, once played on keyboard, clearly sounds different from what the bass alone suggested, an inversion is the first thing to consider before concluding there was a listening error.

In practice, isolating the bass often takes nothing more than a bit of directed attention: deliberately focusing on the low register, even ignoring the melody or lyrics for a moment, is usually enough to bring it out on most recordings. Listening on headphones, or on a speaker that renders low frequencies properly, this simple shift of focus is enough to turn a confusing wall of sound into a sequence of clearly identifiable low notes, one at a time, right in step with the chord changes.

Testing the I, the IV and the V before anything else

Once the chord's root has been spotted in the bass, or even before getting there, a second clue narrows the field of possibilities even further: in the key already identified, three chords are statistically far more likely to be the one you're after than the other four. They're the first, fourth and fifth degrees, the tonic, subdominant and dominant, the three major chords that between them cover all seven notes of the scale, the tonic and the dominant each belonging to two of those three chords. That property is no accident: it explains why a very large share of popular, folk and rock repertoire never strays far from this trio, and why it's the first reasonable bet when facing a chord you haven't yet identified for certain.

Testing these three candidates takes no special equipment: just play them one after another, on keyboard or guitar, looping the passage you're trying to identify, and listen for which of the three creates no friction with what you hear on the recording. The method works by elimination as much as by confirmation: a chord that clashes noticeably with the melody or with the bass note already spotted gets ruled out immediately, often after a single try, leaving only one or two serious hypotheses to compare more closely.

This reflex also does something the full inventory of the seven diatonic degrees, covered in the next section, can't do on its own: it sets a priority order. Faced with an unknown chord, testing the I, then the V, then the IV, in that order, before considering the minor degrees keeps you from chasing every direction at once. That hierarchy isn't arbitrary: the V in particular is often recognisable even before being confirmed by comparison, because it carries the pull toward the tonic that the Music Hub guide on finding the key of a song covers in depth; a chord that seems to want to head straight back to the tonic right after it sounds is almost always that fifth degree, which lets you spot it by instinct before even playing it to check.

This trio obviously isn't enough to cover a whole song's chart, or else no song would ever use more than three chords: it only gives the first thing worth trying, the one most likely to be right on the first guess. As soon as none of the three fits, the right move isn't to keep forcing it but to switch to the full inventory of diatonic degrees, which covers the four remaining chords of the key.

Narrowing the field with the diatonic degrees of an already-known key

Having already found the song's key, using the method covered in the Music Hub guide devoted to it, completely changes the nature of the problem posed by a chord you can't identify on the first try. Without a known key, an isolated chord could theoretically be any one of dozens of possible symbols across the twelve notes of the chromatic scale. With the key in hand, that number collapses to seven serious candidates, one per degree of the scale, each with a quality fixed in advance by that scale's intervals: three major chords on degrees I, IV and V, three minor chords on degrees ii, iii and vi, and one diminished chord on the seventh degree. This table of the seven diatonic degrees, which the Music Hub's key-finding guide covers in its own right, becomes a different tool here: no longer a way of guessing the key, but a list of suspects to check one by one against what you're hearing, chord after chord, all the way through the song.

This narrowing of the field directly changes how you listen. Rather than asking abstractly which chord you're hearing, the question becomes much narrower: which of these seven chords, all already known in advance, best matches what you hear. It's then enough to compare the colour of the chord you've spotted, major or minor, against the quality expected on each of the seven degrees: a chord that sounds minor immediately rules out degrees I, IV and V, leaving only three candidates, ii, iii and vi, to be told apart by their position in the bar or by the root already spotted in the bass. A diminished chord, rare but recognisable by its marked instability, points almost always to the seventh degree, the only one in the key to carry that colour.

The diatonic inventory also does a quiet but valuable job: it lets you check a hypothesis after the fact. Once a provisional chart has been written out from start to finish, reading it back against the seven expected degrees quickly reveals where something's off, a chord that matches none of the seven, or a progression that seems unlikely given each degree's usual role. That isn't always a sign of a mishearing: it can just as easily signal a genuine chord borrowed from another key, a case covered later in this guide. But in the great majority of cases, a chart that doesn't fit the diatonic table of the announced key at all should first raise doubts about the key itself before raising doubts about the theory: it's then worth going back to double-check that key before continuing the chord-by-chord transcription.

Recognising major, minor and seventh by colour, by ear

Once a chord's root has been spotted in the bass and the field narrowed to the key's seven diatonic degrees, what's left is deciding its colour: major, minor, or one of the seventh variants that the Music Hub guide on seventh chords builds note by note. This step doesn't require counting semitone intervals on the spot: it's played by ear, on an immediate sensation that most listeners pick up on well before they can name it.

That perception is neither trivial nor purely subjective: it was studied scientifically as early as 1935, when the researcher Kate Hevner asked listeners to describe, using adjectives, musical excerpts that were identical in every respect except mode, and showed that the major mode was consistently linked to adjectives like happy or serene while the minor mode drew words like sad or dark. More recent work, notably a study published in 2001 by Simone Dalla Bella, Isabelle Peretz and colleagues, pinned down the age at which this association sets in during childhood: sensitivity to mode colour alone, independent of tempo, becomes measurable between the ages of six and eight, whereas at five only tempo still shapes the judgement, and it keeps strengthening well into later childhood to reach the near-automatic reliability it has in adulthood. In other words, the ability to recognise a chord's colour by ear isn't some gift reserved for the lucky few with perfect pitch, an ability that's actually far rarer than commonly assumed, even among trained musicians: it's present in almost every listener from childhood on and keeps sharpening in adulthood through sheer repeated practice.

Beyond the plain happy-or-sad feeling, a few more technical cues help decide faster during a transcription. A major chord sounds stable and complete, as if it could stand on its own without asking for anything more. A minor chord keeps that same stability but with a darker tint, with no tension pushing it to resolve elsewhere. A dominant seventh chord, by contrast, carries a distinctive tension, almost an itch, that makes the chord feel like it absolutely wants to resolve onto another chord right after: it's precisely the tritone sitting between its third and its seventh, described in detail in the guide on seventh chords, that produces this effect. A major seventh, conversely, adds richness without adding tension: it sounds settled, almost dreamy, closer to the rest of a major chord than to the pull of a dominant 7.

The best exercise for building up this recognition isn't replaying the same song over and over, but comparing directly, on a keyboard or guitar, a major chord and its minor version on the same root, then that same major chord and its dominant-seventh version: hearing the contrast directly, a single note changing or being added each time, anchors the difference far faster than a theoretical definition read without the sound to go with it. The piano built into every Music Hub song page serves exactly this purpose, giving the instant sound of any chord to compare against what you just heard on the recording.

Checking the hypothesis on keyboard or guitar

A hypothesis about colour or root, however solid it seems by ear alone, should always be checked by replaying it against the recording rather than comparing it from memory once the music has stopped. The most reliable method is to play the candidate chord while the song keeps running, on keyboard or guitar, simply turning your own instrument down relative to the recording: a correct chord blends into the mix or creates exactly the expected tension, while a wrong chord produces a beating or clashing sound almost immediately, noticeable even to an untrained ear, far more clearly than when comparing two sounds heard one after the other.

When two neighbouring hypotheses both survive a first test, such as a major chord and its parallel minor, built on the same root, which share two notes out of three, the most efficient move is to isolate the one note that tells them apart, the third, and play it alone against the recording rather than the full chord. That single note is almost always enough to settle it: it either sounds right or it doesn't, without the ambiguity a full chord can create when two correct notes out of three partly mask the third that isn't.

Keeping a tonic reference sounding, or at least staying mentally present while testing a chord, also helps you avoid drifting as you listen. It's the same principle as the movable-doh reference covered in the Music Hub guide on finding the key of a song, applied here not to the whole song but to each chord tested in isolation: with the ear anchored on the reference tonic, you hear much faster whether the candidate you're testing departs from it in the right direction or the wrong one.

One trap is worth flagging here: relying solely on the melody sung over the top to guess the underlying chord. A single melodic note very often belongs to several different chords at once, as the root for one, the third for another, a passing note for a third: checking that a melody note fits inside a chord never proves that chord is the right one, only that it's among the compatible candidates. Only comparing the full chord, played against the entire mix rather than against the vocal line alone, allows you to conclude with real confidence. The piano built into every Music Hub song page, which gives the instant sound of any chord without needing an instrument on hand, makes this test especially quick to repeat as many times as needed.

Slowing down without distorting: the tools that train the ear

A fast chart or a dense harmonic passage can simply go by too quickly for even a trained ear to identify each chord before the next one has already arrived. Slowing playback down without changing its pitch is an old need, one that far predates today's software, and the way it's been solved has changed a great deal over the decades.

Before the digital era, the only way to slow down a recording was mechanical, and it had a major drawback: on a turntable as on a reel-to-reel tape recorder, playback speed and pitch are physically locked together. Slowing the tape or record down inevitably drops the pitch along with the tempo, and speeding it up just as inevitably raises both together. That constraint didn't stop generations of jazz musicians from learning fast passages by ear by deliberately slowing a record down, even if it meant mentally shifting every note up an octave afterward when the record was spinning at half speed, to compensate for the pitch drop that slowing it down causes. This slowed-down transcription technique has nothing to do with the half-speed mastering used separately for cutting vinyl records, which involves running both the source tape and the cutting lathe at half their normal speed to better capture high frequencies: the two half-speed uses share only the name, not the purpose.

Analogue recording studios also had, on certain reel-to-reel tape machines, a variable-speed control that deliberately exploited this same coupling between speed and pitch, for creative rather than educational ends: slightly slowing the tape down during recording, then playing it back at normal speed, makes an instrument or a voice sound higher-pitched and faster than it actually was at the moment of the take, a technique knowingly used in a number of historic productions.

Digital technology changed the game by separating what had stayed mechanically welded together on tape or on disc: signal-processing algorithms now let you slow a recording down without touching its pitch, or conversely change its pitch without touching its tempo, two operations that were simply impossible to separate before these techniques arrived. Music transcription software, and a good many mainstream audio players, now build in this kind of feature under various names, often close to speed or tempo, without touching the note actually played. For a passage that resists the ear at normal speed, dropping the playback speed to half or three-quarters, with no pitch change, is very often enough to make chords that seemed to blur into one another stand out distinctly.

This tool never replaces the ear work detailed earlier in this guide: all it does is give more time to an ear that already knows what to listen for, the root in the bass, the chord's colour, its place among the key's diatonic degrees. Slowing down a passage you don't yet know how to listen to correctly just slows the confusion down along with it.

Handling chords that resist the diatonic pattern

Even once the key is settled and the reflex of checking the seven diatonic degrees is well established, a chord in the chart sometimes refuses to match any of the usual seven suspects. That's almost never a sign of a mishearing: a good share of the repertoire, including the simplest-looking songs, occasionally borrows a colour from outside its home key without ever really changing key. The Music Hub guide on chord substitutions and reharmonization covers the full construction of these borrowings note by note, secondary dominant, passing diminished chord, borrowing from the parallel minor; the concern here is different, and more modest: recognising at first hearing that you're probably dealing with one of these cases, so as not to wear yourself out trying to force it into one of the seven expected degrees.

The clearest signal remains a false local tonic: an unexpected major chord, immediately followed by a chord that seems to be its natural resolution, without that resolving chord being the song's real tonic. That pair almost always gives away a secondary dominant, borrowed to briefly tonicise a neighbouring degree of the key, a mechanism that in practice stays remarkably hard to spot for an untrained ear while being almost perfectly regular once you've caught it the first time.

A second signal is a sudden, momentary darkening that doesn't, however, establish a genuine minor-key feeling over a whole section: that's the signature of a borrowing from the parallel minor, often on the fourth or seventh degree, which colours one or two chords before an immediate return to the original major colour. The third signal, finally, is purely rhythmic rather than harmonic: a chord held for less than a full beat, slipped between two more stable diatonic chords, which often doesn't need to be identified with absolute precision to play the song well, only spotted and placed in the right spot.

Faced with any of these three signals, the best discipline is to resist the temptation to force the chord you're hearing into the nearest diatonic degree just because that's the easiest option: a slightly wrong chord, forced in on the assumption that it must be one of the seven expected ones, introduces a quiet error that can throw off the reading of an entire passage. Noting the chord down provisionally as you hear it, even if it means checking its exact construction later with the help of the guide on substitutions, is always better than dressing it up as a diatonic degree it doesn't really resemble.

When the version you're hearing isn't the one you think it is

A chart patiently worked out by ear can perfectly well be correct and yet match no tab found online for the same title, simply because the version you're listening to isn't the one that tab describes. This mismatch isn't rare, and it's worth knowing about before wrongly doubting yourself.

The most frequent case comes down to how a song is actually played on guitar: many guitarists use a capo to replay a song with easy shapes while still getting a different real pitch, which means the same tab can describe two keys at once, the shape played under the fingers and the pitch actually heard. That's exactly the distinction every Music Hub song page maintains, displaying the key of the shapes played and separately noting the recommended capo for recovering the recording's real pitch: without that detail, a chart correctly worked out by ear, in the record's real pitch, can seem to never match a tab that instead describes open positions without specifying the capo the original guitarist used. An alternate tuning, less common but far from rare, produces exactly the same confusion for a different reason: a hand shape identical to a familiar open chord can sound at a completely different pitch in that tuning, with no relation at all to what the same shape would give in standard tuning.

A second case comes down to the recording's own history rather than the instrument. Today's universal pitch reference, A440, wasn't recommended until 1939, at an international conference in London, and was only confirmed in 1955 by the International Organization for Standardization; before that date, reference pitch varied noticeably from one orchestra, hall or country to another, sometimes by nearly a full semitone. An old recording can therefore sound slightly higher or lower than its key as stated on a period score would suggest, with no transcription error involved at all. Closer to our own time, the tape recorders of analogue recording studios could introduce pitch drift well after the modern standard was adopted, each machine running at a speed very slightly off normal, particularly on old tapes rewound and replayed dozens of times over the decades.

A third case, simpler but just as common, comes from the fact that the same title often exists in several distinct official versions: a studio version, a live version replayed in a different key to spare the singer's voice on stage, a cover by another artist transposed for their own vocal range, a remix or remaster that slightly shifts the playback speed. Before concluding that a chart worked out by ear is wrong because it doesn't match a source found elsewhere, the first question to ask is whether the two actually describe the very same recording, not just the same title.

What chord-detection software does, and doesn't do

Ear training isn't the only way to get a song's chart: automatic chord-recognition software has existed since the late 1990s and now ships inside mainstream consumer apps. Understanding how it works directly illuminates the method described in this guide, because these algorithms stumble, almost point for point, on the same difficulties as a human ear.

The starting point of this research field is a paper presented in 1999 at the International Computer Music Conference in Beijing by the Japanese researcher Takuya Fujishima, describing a real-time chord-recognition system tested on around 800 electric-guitar chord sounds. Its central idea, the pitch class profile, condenses the sound energy detected through a Fourier transform into a twelve-slot vector, one per note of the chromatic scale, regardless of the octave that note is played in. That vector, compared against a theoretical template for each possible chord type, is still today the basic principle behind nearly every automatic harmonic-recognition system: rather like the key profile already met in the Music Hub's key-finding guide, but refreshed here much more often, several times a second, to track every chord change rather than settle on a single colour for the whole song.

Four years later, in 2003, the researchers Alexander Sheh and Daniel Ellis added an ingredient to this principle borrowed directly from automatic speech recognition: the hidden Markov model, which treats each chord as a probable state and each transition from one chord to another as a statistically more or less frequent move, learned from a large number of already-known charts. This addition fixes a flaw in Fujishima's plain template: without it, an algorithm can detect a stray chord for a fraction of a second because of a single passing note or a signal artefact, before immediately reverting to the correct chord; the hidden Markov model smooths out that instability by favouring the most likely progressions rather than detections too brief to be a genuine chord change.

This combination, pitch profile plus probabilistic model, became the basis for a benchmarking exercise held every year as part of a major international conference on music information retrieval, where several teams pit their algorithms against the same hand-annotated reference corpus. The detection rate for plain major and minor chords alone climbed steadily over successive campaigns, reaching around 83 percent by 2012, driven by ever finer chroma representations. More recent architectures, built on deep learning rather than hand-written templates, have since pushed that figure even higher on simple chords; recognition of enriched chords, from sevenths to the heaviest extensions, along with recognition of inversions, remains markedly less reliable by comparison.

This research eventually found its way into services available to any musician: a Dutch company, Chordify, founded in 2013 out of a hackathon project that applied this academic research directly, has since offered a tool that transcribes the chords of an online video or an imported audio file live, rendered as a chart that scrolls along with the music. These tools genuinely help with roughing out an unfamiliar song, but they inherit the same blind spots as a beginning musician: a seventh poorly told apart from a plain triad, a passing chord counted as a full-fledged change, or a bass inversion that throws off root detection, exactly the kind of case this guide details for the human ear. So no software, however refined, fully replaces the verification described earlier in this guide, on keyboard or guitar, against the real recording.

When to stop: the chart that's good enough beats the perfect one

Working out a chart by ear can, in theory, go on forever: refining a voicing, checking whether a given ninth was really played or only implied by the melody, hesitating between two inversions that sound almost identical on the recording. At some point, though, digging further stops adding anything useful, and knowing how to recognise that moment is as much a part of the method as the steps that come before it.

The first marker is functional rather than theoretical: a chart is ready as soon as it lets you play the song from start to finish without any noticeable snag, in step with the bass, with no note clashing against the melody or the other musicians. That chart may well simplify a thirteenth chord down to a plain seventh, or ignore an extension barely audible in the mix, without that hurting the use you're actually going to make of it, playing and accompanying a song rather than publishing a scholarly transcription meant for analysis. The guide on reading a chord chart makes the same point in its own way: a chart exists first to be played, not to prove absolute exactness.

The second marker has to do with the very nature of the chord still in question. A chord held for a full bar, carrying most of a passage's harmonic colour, is worth lingering over until you're sure of its root and its quality. A passing chord held for a fraction of a beat, slipped between two more stable chords, doesn't deserve the same relentlessness: what matters is sensing that it's there and placing it in the right rhythmic spot, not pinning an absolutely precise theoretical label on it, one that even some of the musicians who recorded it might not be able to name with certainty.

The third marker is the law of diminishing returns applied to the ear: the first few minutes spent on an unfamiliar song pay off enormously, the root in the bass, the chord's colour, its place among the diatonic degrees narrow the field of possibilities at every step taken. The minutes after that, spent hesitating between two barely different voicings of a chord already broadly identified, pay off less and less relative to the time invested. Recognising that turning point, and settling for a reasonable hypothesis rather than chasing an unlikely total certainty, is what separates an efficient transcription from one that never ends.

Finally, a chart worked out by ear doesn't need to be perfect on the first listen to be useful: replaying it a second or third time, a few days later, with a rested ear, often corrects on its own the two or three chords that stayed uncertain, with no need to start over from scratch. The method described in this guide, from the bass through to checking it on keyboard, stays the same on every pass: it applies just as well to a first rough transcription as to its gradual refinement over the listens that follow.

Common pitfalls to avoid

Explore in the Music Hub

More guides