Figuring out chord progressions by hand is slower than using a tool
Roman Numeral Analysis Music Theory is just a system for labeling chords based on their relationship to a key rather than their absolute pitch names. It replaces C, E, G with I in C major. This labeling makes it faster to spot patterns across transpositions, which matters if you're analyzing a Bach chorale or a pop song that keeps shifting keys halfway through. Start with the key. You need to know whether you're in a major or minor mode before anything else makes sense. Look at the key signature and the final cadence. A piece ending on a major tonic with a sharp seventh in the melody is almost certainly major. If the sixth and seventh degrees are flattened, you're likely in a minor key. This step alone takes most beginners wrong about a third of the time because they assume the key signature equals the key, which is wrong in natural minor where the key signature is shared with the relative major. Once you've locked in the key, label each harmony by scale degree. The tonic is I, the supertonic is ii, the dominant is V. In minor keys, the supertonic becomes ii° because the lowered sixth and seventh create a diminished triad. The submediant is VI, not vi. Capitalization tells you the quality. Uppercase means major or augmented. Lowercase means minor or diminished. An augmented sixth gets the plus sign: III+. A German augmented sixth might be labeled Ger+6, which I usually collapse into bVI+ because the functional label gets messy and nobody agrees on it anyway.
Here is a concrete example in C major. C to F to G7 to C. That is I to IV to V7 to I. Now look at Am to Dm to E7 to Am in the same key. That is vi to ii to V7/vi to vi. The vi chord shows up as a temporary home base, and the E7 is a secondary dominant pointing at it. This is where most people mess up. They label E7 as V because they forgot you have to check whether the target chord actually appears. If the Dm never shows up, calling E7 a V is wrong. It would be a French augmented sixth or just a stray dominant depending on how the voices move, but not a secondary dominant if the resolution is absent. When you run into non-diatonic chords, which happens constantly in Romantic music and pop songs with modal interchange, stop trying to force every chord into the home key. Borrowed chords from the parallel minor belong in a major key context. Eb major in C major is not some exotic alteration. It is bIII, borrowed from C minor. Label it bIII and move on. Do not try to rekey the passage just because a chord sits outside the diatonic palette.
Specific problems I have hit and how I worked around them
I spent two days last year labeling an Chopin nocturne where the right hand played chords that clearly belonged to D-flat major while the left hand was grounding everything in C-sharp minor. The key changed so frequently that a single Roman numeral grid looked like a crime scene. I ended up labeling in short two-measure spans instead of committing to one master key, then mapping the shifts at the cadential boundaries. It is not perfect, but it is faster than restarting the whole exercise from scratch every time you encounter a new modulation. The workaround cut the time from about four hours down to roughly forty minutes for a piece that took me longer than I would like to admit the first pass. Another issue that comes up constantly is figured bass versus Roman numeral confusion. When you see a bass note with numbers above it, those numbers are interval abbreviations, not Roman numerals. A 6/4 chord over the bass note G in C major is not G6/4 in some new notation system. It is I6/4, a second inversion tonic. Beginners routinely label that as V6/4 when it functions as a cadential six-four, and while the Roman numeral stays the same, the function label matters for voice-leading analysis. Call it V4/2 in functional terms if your instructor requires that level of detail. Just do not call it V6/4 without explaining the cadential resolution.
Get the Full Details
Tools that save time
Manual analysis works fine for simple diatonic progressions. It becomes painful quickly. I use a script that parses MIDI input and outputs candidate Roman numerals based on pitch-class sets and a target key signature. The script handles common cases reliably, which usually cuts the initial pass from two hours down to about fifteen minutes depending on how tangled the voicings are. You can find open-source implementations on GitHub under repositories like roman-numpy or by searching for midi-to-roman-numeral scripts. The output is not final. You still have to audit every borrowed chord, every non-chord tone, and every ambiguous inversion. But starting from a machine-generated grid saves you from making the most basic errors on the first pass. Roman numeral analysis breaks down in chromatic harmony. If you are looking at late Scriabin or certain Stravinsky passages where the tonal center is deliberately obscured, the labels become meaningless. You will end up with something like #iv/ii°7, which is technically correct but completely useless for understanding what the music is doing. In those cases, use pitch-class set theory or Schenkerian reduction instead. Do not pretend Roman numerals explain everything. They are a simplification tool, not a universal framework. There is also the issue of non-functional harmony. Jazz charts and neo-soul progressions often stack extensions that change the color without changing the Roman numeral. A Maj7#11 sounds very different from a plain Maj7, but both are I. If you need to capture that difference, use chord symbols alongside the numerals or switch to a labeled function system like Hooked on Chords. Mixing Roman numerals with extended chord notation usually ends up cluttered and hard to read, which is why I tend to keep them separate and use one or the other depending on the analytical goal.
The system also struggles with modal music and music that refuses to settle on a single tonic. Folk tunes that drift between Dorian and Mixolydian modes, ambient pieces that treat the tonic as a hovering center rather than a pull destination, and minimalism that relies on additive processes rather than harmonic progression. Labeling those passages with Roman numerals produces results that look precise but convey almost nothing about the actual sound. In those contexts, pitch-class analysis or interval-class vectors give you more honest information, even if they are harder to work with on paper.
