Some vids about arranging on fingerstyle guitar say to figure out the chords first either by ear or through tabs, then to play the melody on top of them. But that seems like too much work. For one, even if I know the chord I'd have to start thinking of the different ways to play that chord all over the fretboard so it could be played next to the melody note I want.

Instead, I thought that chords could be inferred from the melody. For example first I find the melody by ear. Then for each melody note I’d like a chord on, I add a bass note that sounds good with it and then add middle notes and eventually the chord reveals itself from the shape I formed.

So I just keep adding supporting notes to the melody until I have my chord. And I keep doing this to find the other chords. Sometimes I don’t even bother to know what the name of the chord is (especially if it’s somewhere in the middle of the fretboard) as long as it just sounds “good” with my melody note. So I was wondering if this is right. Can chords be found from the melody alone when arranging by ear?

Edit: Please note that this question assumes that I have heard the song before. I mainly arrange for well known songs that I have in my head. Pop songs, traditional songs, etc. Mainly as an exercise for myself. I am aware of what these songs are "supposed" to sound like as I'm arranging them. But I'm trying to find the chords from the melody as the melody is what I usually start off with. Here's an example of the style I'm working towards. I wrote this question with the guitar in mind. Especially due to its complexity with chord-melody throughout the entire fretboard. But I guess this would work for any instrument that can play chords as well as melody.

  • Yes, it has to do with perception and memory. – jjmusicnotes Aug 6 '19 at 19:41
  • 8
    If you take a melody you are unfamiliar with and come up with “chords that work” with that melody the possibilities are endless based on the style, the taste of the arranger and a million other factors. You may come up with chords but you may also have a different song when you are done. – b3ko Aug 6 '19 at 20:15
  • I've noted that the original harmonization of "Rose of May" from Final Fantasy IX is not quite the harmonization I would have used--it's off by some chords. I've also often ended up minorly reharmonizing other video game themes I've arranged. – Dekkadeci Aug 7 '19 at 0:23
  • My understanding is that much of traditional Irish music began as melody only, and the guitar/piano/concertina accompaniment was inferred by later performers from the melody alone. – schadjo Aug 8 '19 at 14:42

No, not if by inferred you mean unambiguously deduced. For example, organists have developed the reharmonization of hymn tunes into a fine art.

Yes, if by inferred you mean finding chords that more or less fit. That's because harmonization is possible at all. Some sequences of chords will fit better (by various criteria) than others, of course.

Postscript: Béla Fleck and the Flecktones' The Star-Spangled Banner is reharmonized so imaginatively that, without a spoiler (as I've done here), few can guess the melody, albeit never disguised, until conventional harmony returns at "O say does that star-spangled..."

  • 2
    A song could be harmonized in multiple ways so there won't ever be a situation where it's "unambiguously deduced". Two arrangers might pick different chords based on their taste. – user34288 Aug 6 '19 at 20:43
  • 6
    Reharmonization is also ubiquitous in jazz music. – Your Uncle Bob Aug 6 '19 at 20:43
  • 4
    And in Brahms, and in probably any genre that has a concept of chord! I just picked a genre that makes a big deal of teaching how to reharmonize. – Camille Goudeseune Aug 6 '19 at 20:56
  • Also, check out Stravinsky's arrangement of The Star-Spangled Banner. – schadjo Aug 8 '19 at 14:43
  • @foreyez, your original question made me think you wanted to write out the song as originally composed and arranged. If you seek a technique that opens the door to reharmonizations, then there's no need to look up the chords, and using your ear is a better approach. If you seek a technique for transcribing the original chords of the song, then your described procedure could lead to incorrect transcriptions of the chords. Maybe the term "infer" is too vague. To me, this implies "infer the original chords." – jdjazz Aug 8 '19 at 18:22

It's very style dependent, but yes melody can imply harmony (chords.)

I think a crucial aspect is to not think one note at a time. Look for melodic segments that match common harmonic patterns.

Let's say you are in C major. The melody tone is G. As a single note it isn't clear how to harmonize it. It could be a I or V chord. Now imagine the melodic segment is G F E. It's much clearer the harmony could be V V7 I. Of course it could be harmonized other ways too. But the point is to look for common melodic patterns when thinking about the harmonic implications of a melody.

And of course there are melodies that outline chords. Like a melody of C E G. An obvious harmonization is to just harmonize that with a C major chord.

  • yep that's a good point, the melody notes often trace the outline of a chord. but often, just whatever "sounds good" seems good enough. I noticed academia tends to make harmonization formulaic. – user34288 Aug 6 '19 at 22:39
  • 2
    I wouldn't go too far with this being academia being formulaic. You will see patterns like this - formulas - in many styles. – Michael Curtis Aug 7 '19 at 12:40
  • 1
    @foreyez, I think it all depends on what you mean by "good enough." Are you looking for a way to transcribe the original chords, or are you asking about techniques for reharmonizing a song? Are you looking for chords that are "good enough" to make the song recognizable to a crowd? (This could be quite subjective.) Some clarification on this point in the original question might be helpful. – jdjazz Aug 8 '19 at 18:26
  • @jdjazz that's a good point. I'd say "good enough" to make the song recognizable to a crowd. doesn't have to be 100% the way the writer originally wrote it. but if you look at youtubers such as 'acoustic trench' he tends to arrange as well (youtube.com/user/AcousticTrench/videos) and he does it by ear too. I'm not one that looks at tabs or sheet music and I never will. but I hear these songs in my head very clearly so I know if a chord sounds off. – user34288 Aug 8 '19 at 19:05
  • @jdjazz not to mention most songs are just different combinations of I IV V vi. so it becomes repetitively easy. – user34288 Aug 8 '19 at 19:34

No. Chords cannot be "inferred" from melody alone. Creative contribution from an arranger is required.

As a counterexample, here is a melody that should be clearly completely ambiguous about "its chords", and some silly chords for it. Counterexample

Here is another counterexample. First there's a two-bar melody and then many different chords for it. Assigning chords to melody notes is a matter of opinion and creative artistic choices, sometimes technical choices. It is limited by your technique, experience and imagination. Your questions are: What do you want to do with the melody? What can you do with the melody?

Another counterexample

  • Missing an Am at the end of the penultimate example? Or has my technique, experience and imagination run out? – Tim Aug 8 '19 at 15:30
  • @Tim if you feel you want an Am, the last example resolves the issue. :) Maybe I should've left out the double barlines. My point was that there are many ways to assign chords to a melody and these example melodies are designed to highlight this. IMO, being able to assign chord changes to melodies should be every musician's basic skill. How do you know what chords and notes taste like in different contexts, if you haven't toyed around with them? A musician who can't do that is like an actor who can't tell a story unless it's fully written out. I suppose all theater schools teach improvisation? – piiperi Reinstate Monica Aug 8 '19 at 16:03
  • I understand exactly what you're showing, and the point is well made - until we come to that example. Which shows that actually, any old chord can be played with any old note, so it's unconvincing! The other ideas prove the point far better. But yes, as an end of phrase, I feel it's unfortunate. And the double barline tells it probably is... – Tim Aug 8 '19 at 16:09
  • "Technique" being a limiting factor - I meant that very often chords are selected in order to be playable easily enough, and so they're a compromise between what the arranger might optimally want to hear or what would give a maximum musical impact in the arranger's imagination, versus what can actually be played in practice. I'd liken it to visual or technical design work - there might be some given design goals or restrictions, but the designer is still left with many degrees of artistic freedom. – piiperi Reinstate Monica Aug 8 '19 at 16:11
  • @Tim I fixed the example, thanks. I didn't think anyone would read it that carefully, and I just typed chords on top of copy-pasted bars. :) I hope nobody actually tries to analyze the silly first example. – piiperi Reinstate Monica Aug 8 '19 at 16:31

Simple answer to the lead question - NO. Not even when the melody happens to be several notes long. There is, obviously, one chord which fits the original melody - that's the one the writer decided fitted best.

Your method for finding the chords is one of several - but flawed from a time (and experience) point of view. You seem to be looking willy-nilly for notes that might fit together. That's making it a process that's time consuming, for no good reason. That time wasted should (could?) be better spent actually learning more chords, and how harmony works. It'll pay off in the future. Bit like trying to find the notes of a diatonic tune by testing each fret on the guitar - not understanding that nearly half of those won't be included anyway!

Knowing what notes constitute the snippet of melody being harmonised will help. If it's one long note - let's say C - then it stands a good chance that that C note will be part of the chord in that bar. That narrows it down somewhat. Ideas are: Cmaj., A♭, Am, Fmaj., Fm, D7, Co, Cm7♭5 - to name but a few. The key of the piece will give clues as to which will probably work (or not), as will the surrounding bars' harmonies. Maybe a better way - more intelligent - than randomly trying stuff out?

By approaching the chord stucture your way, you learn one song at a time, and cannot use much - if any - of the information gleaned to help any future songs. I play with a couple of guys who still don't realise that if a song's in C, the other two main chords will almost inevitably be F and G, and I'm convinced that when they try to figure out a new song, they'll try all sorts of chords that most likely wouldn't feature in that song. Time spent playing/learning is always better used productively. Unless you feel that what you're doing is really helping you improve rapidly.

Back to my first para. Just about any melody can be re-harmonised. It happens a lot with jazz players, and is fun and challenging to do. There is never just one chord that fits under a melody line. There is probably a 'most appropriate' one, and there's certainly 'spurious' ones, but as long as the melody notes gel with those chords - probably part of a coherent sequence - then the chords are at least legitimate.

  • so you're saying instead to know the diatonic chords. and then to know each of the diatonic chords all over the neck. so say the diatonic chord is C that works with my melody note. and let's say there's 5 forms of C chords over the neck (CAGED). and let's say that the second form (cAged) is the closest to my melody note. so I'd use that.. is that what you're recommending? – user34288 Aug 7 '19 at 11:09
  • Not exactly sure what you need, but by being able to play the 5 (4's enough) shapes, you'll be able to find at least one which gives you the required note on top. For instance, C melody note, C chord - E shape and G shape both have a C note on top. Is that what you mean? Bear in mind that not every note in a melody will have a chord attached to it that contains that note. There are loads of passing notes in tunes. – Tim Aug 7 '19 at 11:24
  • I need the chord to be physically located next to the melody note. (this is fingerstyle/chord-melody guitar). but yes I think you understood me. also the chord name (C) doesn't have to be the same as the melody note (the melody note usually is one of the notes inside the chord, so melody could've been E or G for example). and the passing notes wouldn't get a chord so I'm not worried about those. – user34288 Aug 7 '19 at 11:31
  • Of course it's done by logic/formula and not by ear. The 'by ear' part has been done a million times before. We know it works. Why re-invent the wheel when someone (lots) has done it all for you. Even 'doing it by ear' as you say is a darned sight easier when you have the tools - i.e. know what it actually is 'inside your head' then going straight to it. Your way will work, no doubt, if you live long enough!! Your stubborn-ness will make you spend/waste an awful lot of time that could be used more productively, but who am I to tell you. I've only done it for 50+ yrs, teaching/playing, and ... – Tim Aug 7 '19 at 12:55
  • 1
    You promise..?! – Tim Aug 7 '19 at 13:02

You have to distinguish two cases.

Case 1. Deduce the original chords in the composer's arrangement, or some other arrangement from the melody.

Case 2. Deduce what chords would work to support the melody or harmonize in using the melody alone.

For case 1 the answer is no, there is no way to know from a melody what specific chords and inversions went along with it. The answer for case 2 is absolutely, this is exactly how we are taught to develop the chords for a song. The classical approach to multi voice harmony uses the fact that the I, IV and V (V7) chords cover the entire major scale. Thus the old joke that all you need to know to play rock music is 3 chords actually applies to classical music as well. The fact is there is a wealth of chords available to use in harmonizing a song but may are related by substitution theory. A simple example is a 6th chord and its relative minor 7th. These are in fact the same chord (example: C6 = {C, E, G, A} and A-7 = {A, C, E, G}). There are more elaborate subs but basically from a functional point of view all you really "need" are the I, IV, and V.

So, starting from the melody alone you can come up with a nice rhythm section to support the flow of the song but that may differ from what the original composer intended.

  • In the Case for 1. I'd like to add (even though I didn't write this in the original question), that these are songs that I generally know and have listened to. So take for example 'Somewhere over the Rainbow' even though we know the melody, we'd also know (just be hearing) if one of the chords was off. So take into account that this question also has the factor of having the memory of a song and harmonizing it by ear. – user34288 Aug 7 '19 at 16:27
  • 1
    I would bet that there a cases where a sub would work and you memory might not trigger an alert. But I agree that a mixture of theory and memory can get you very close. – ggcg Aug 7 '19 at 16:29

I don't think you can "infer" them (as you're not the original artist), but you could try something like this:


Lookup the scale that your melody is using for composition, then go into the guitar chords section and use those chords.

Sadly, All-guitar-chords.com used to have a tool called "reverse chord finder," which would let you put in the notes of a scale (melodies are usually based on scales), and then it would give you the key. That tool appears to be broken, but you could figure out the key manually from the notes they're using in the melody, and then use music theory to put chords in artistically.


If you're doing a cover, it's going to be somewhat different from the original anyway. Find the key that the piece is written in, then use your own creative judgment to approximate chords on important beats. This could be on the upbeat or the downbeat (or both?), depending on the genre and piece you're working with.

All-guitar-chords.com also has lots of chord progressions for different keys, so it's a great tool for writing music.

Keep in mind that some pieces will change keys, so you would have to adapt to this if you're working with such music.

EDIT: One other thing, you will probably want to use the root note whenever you are adding a chord. Try adding them on the I IV and V, although this will vary, stylistically.


The simplest way to show that this idea doesn't even begin to work is to look at some actual music.

Take a harmonically simple song like the Beatles' "yesterday" for example. In C, the tune starts D C C.

The D is on the first beat of the bar, so the first chord has to fit with the D, right? Wrong - it's C major. Not C major 9 - just plain C major!

And after a rest, the tune starts again with E and E sharp. So you want a chord that fits with E? Wrong again - it's actually B7, which you wouldn't be likely to think of at all if were locked in to the mindset of "this song is in C major". Oh, and the chord change is in the middle of the rest, where there are no notes at all to "help" you find the right chord.

At a more extreme level, I once wrote a set of 32 variations on the bass line for Pachelbel's canon (D A B F# G D G A and repeat endlessly) where every variation (except the first and last) had a completely different chord sequence - and most of the variations didn't have any obvious connection to "D major" at all. But the notes of the bass line were still there, unaltered, in every variation.

  • 1
    There's an E# in 'Yesterday' when it's in key C? Really? Where? – Tim Aug 7 '19 at 11:27
  • 2
    Well, there is a note enharmonic to E# in the key of C, namely F! What Tim means is that we don't have an opportunity or justification to call F by the name of E# in either the key of C, or its relative A minor. We would have to be in a key like F#, where we have these sharps: F# C# G# D# A# E#. – Kaz Aug 7 '19 at 21:01