70.5% of all chords in Billboard Year-End hits from 2002 to 2021 sit inside looping passages, compared with 53.7% in a random sample of Billboard Hot 100 singles from 1958 to 1991 (Chander and Quinn, ICMPC17-APSCOM7 2023). The basic chord vocabulary has barely moved: in songs from Rolling Stone’s “500 Greatest Songs” list, the I, IV and V chords supply 32.8%, 22.6% and 16.3% of all chord roots (de Clercq and Temperley, Popular Music 2011). What has changed is the color around them, with the dominant-seventh harmonic topic falling by about 75% in Hot 100 songs between 1960 and 2009 (Mauch et al., Royal Society Open Science 2015). We aggregated data from Royal Society Open Science, the Rolling Stone rock corpus, the Chordonomicon dataset of chord charts, Hooktheory’s TheoryTab database, a Yale comparison of Billboard chord loops, and two 2026 studies in Music Theory Online and the International Journal of Contents. Tempo and key get their own roundup in our song tempo and key statistics; this one looks at the chords themselves.
TL;DR
- 70.5% of chords in 2002-2021 Billboard Year-End hits sit in loops, up from 53.7% in 1958-1991 Hot 100 singles (Chander and Quinn, ICMPC17-APSCOM7 2023).
- In 1950s songs on Rolling Stone’s list, I, IV and V account for more than 95% of all harmonic content (de Clercq and Temperley, Popular Music 2011).
- After the two-chord move IV then I, the V chord follows 41% of the time in TheoryTab songs (Hooktheory Trends API documentation).
- 71.2% of 51,007,915 notated chords are major type and 25.7% are minor type (Kantarelis et al., Chordonomicon, arXiv 2024; our calculation).
- Seventh chords fell from 27.7% of chords in 1940s songs to 8.25% in the 2020s so far (Chris Dalla Riva, Can’t Get Much Higher 2025).
- The dominant-seventh harmonic topic declined by about 75% in Hot 100 songs from 1960 to 2009 (Mauch et al., Royal Society Open Science 2015).
- Only 12 of 178 Year-End Hot 100 hits from 2023 to 2025 fit a pure “Static-Loop” harmonic type (Kim, International Journal of Contents 2026).
- 41% of 100 anime opening songs contain at least one chromatic modulation (Joy Li, Music Theory Online 2026).
- Power chords are 2 to 21.5 times more common in punk than in any other genre (Chris Dalla Riva, Can’t Get Much Higher 2025).
- The clustering coefficient of the pitch-transition network fell from 0.65 to 0.45 across 464,411 recordings from 1955 to 2010 (Serra et al., Scientific Reports 2012).
1. The Core Chord Vocabulary: Which Chords Pop Actually Uses
Music theorists name chords by Roman numerals that count up from the home note of the key: I is the home chord (the tonic), IV and V are built on the fourth and fifth notes of the scale, and a lowercase numeral such as vi marks a minor chord. In the key of C, I is C major, IV is F major, V is G major and vi is A minor. Counted this way, one chord in three in the Rolling Stone rock corpus is the home chord itself, and I, IV and V together cover 71.7% of all chord roots (32.8% + 22.6% + 16.3%, our sum).
The deeper point is that harmonic time is even more lopsided than harmonic counts. Measured by bars instead of chord changes, the I chord takes up almost half of the 200-song Rolling Stone corpus (RS200). Pop does not just return home often; it lingers there. The datasets disagree on exact shares because they count different things (whole songs, song sections, or user-written guitar charts), but every one of them puts the same few major chords at the top.
| Metric | Value | Source |
|---|---|---|
| Share of all chord roots that are I, IV and V, Rolling Stone 100-song corpus (1950s to 1990s) | 32.8%, 22.6% and 16.3% | de Clercq and Temperley, Popular Music 2011 |
| Share of all chord roots that are bVII (the flat-seven chord, for example B-flat major in C) and VI | 8.1% and 7.2% | de Clercq and Temperley, Popular Music 2011 |
| Share of all bars (measures) spent on the I chord, 200-song Rolling Stone corpus | 48.5% | de Clercq, EuroMAC 2014 corpus presentation |
| Share of all chords in the TheoryTab database that are I and IV | 17.6% and 12% | Hooktheory Trends API documentation |
| Most popular four-chord progression in the TheoryTab database | I-V-vi-IV (C-G-Am-F in C) | Hooktheory blog, 1,300-song analysis |
| Share of about 52 million user-notated chords that are G major or C major | 24% | Chris Dalla Riva, Can’t Get Much Higher 2025 |
| The seven most common chords in the McGill Billboard sample (1958 to 1991) | All major: C, D, G, A, E, F and B | Burgoyne, Wild and Fujinaga, ISMIR 2011 |
Context note: the Rolling Stone corpus figures come from the de Clercq and Temperley rock corpus paper, which counts chord roots with major and minor chords combined. Most recent available data: de Clercq and Temperley, 2011 and 2014. The G and C dominance in user-written chord charts counts absolute chord names, not Roman numerals, so it mixes harmony with the keys songs are written or transcribed in (our interpretation); chord letter names also assume a shared tuning reference, a topic covered in our concert pitch statistics.
2. Chord Transitions: IV Is the Favorite Route Home
The textbook says V leads back to I. Pop listeners hear something else: the most frequent chord before the tonic in the Rolling Stone corpus is IV, at 39.6% of pre-tonic chords against 26.9% for V. This “plagal” pull (IV moving to I, the sound of a church “amen”) is one of the clearest statistical differences between rock and pop harmony and the classical rules most theory courses teach.
The Hooktheory transition data shows how predictable the next step is once a progression starts. After IV, I is the single most likely chord, and after IV then I, the V chord follows 41% of the time, tracing the IV-I-V stretch of the vi-IV-I-V rotation of the familiar I-V-vi-IV loop. Recent Billboard hits push this further: in the largest cluster of 2023 to 2025 Year-End songs, more than seven in ten IV chords resolve straight to I.
| Metric | Value | Source |
|---|---|---|
| Share of chords immediately before the tonic (I) that are IV vs V, Rolling Stone corpus (distribution excludes the tonic chord itself) | 39.6% vs 26.9% | de Clercq and Temperley, Popular Music 2011 |
| Probability that the I chord follows the IV chord, TheoryTab database | 33% | Hooktheory Trends API documentation |
| Probability that the V chord follows the IV chord, TheoryTab database | 27% | Hooktheory Trends API documentation |
| Probability that V follows the two-chord sequence IV then I, TheoryTab database | 41% | Hooktheory Trends API documentation |
| Share of times the iii chord (E minor in C) was followed by vi (A minor) or IV (F major), 1,300-song TheoryTab sample | 93% | Hooktheory blog, 1,300-song analysis |
| Share of IV chords that move to I, “Dynamic-Moderate” type songs (72 of 178 Year-End Hot 100 hits, 2023 to 2025) | 71.4% | Kim, International Journal of Contents 2026 |
| Share of bVII chords that move to I, “Dynamic-Complex” type songs (47 of 178 hits) | 45.0% | Kim, International Journal of Contents 2026 |
Context note: the Hooktheory Trends API documentation gives these probabilities as worked examples and notes that the TheoryTab library has grown a great deal since the API was first published, so treat them as a snapshot. The Kim percentages are conditional probabilities inside each song type (the share of a chord’s outgoing moves that go to a given chord), not the share of songs containing that move. Most recent available data for the 93% figure: Hooktheory’s early 1,300-song blog analysis (undated).
3. Chord Quality: Major Triads Dominate, Extensions Stay Rare
Chord “quality” describes the type of chord built on a root: major, minor, suspended (a chord with the third replaced by a second or fourth), power chord (root and fifth only, the distorted-guitar staple) and so on. In the Chordonomicon, a dataset of user-written chord charts scraped from the web, nearly nine out of ten chords are plain three-note chords (triads), and major-type chords outnumber minor-type chords by close to three to one.
This matters for anyone modeling or teaching pop harmony. Extended chords (four notes or more) are the exception, not the background texture, and inversions (chords voiced with a note other than the root in the bass) are rarer still in guitar-chart notation. A chart written by a fan simplifies harmony, so these shares describe how songs are learned and played as much as how the records were produced.
| Metric | Value | Source |
|---|---|---|
| Share of 51,007,915 notated chords classified as major type | 71.2% (36,331,633 chords; our calculation) | Kantarelis et al., Chordonomicon, arXiv 2024 |
| Share of notated chords classified as minor type | 25.7% (13,104,882 chords; our calculation) | Kantarelis et al., Chordonomicon, arXiv 2024 |
| Share of notated chords that are suspended vs power chords | 1.4% vs 1.4% (721,086 and 717,252 chords; our calculation) | Kantarelis et al., Chordonomicon, arXiv 2024 |
| Share of notated chords with three notes vs four notes | 89.2% vs 8.7% (45,477,125 and 4,457,563 chords; our calculation) | Kantarelis et al., Chordonomicon, arXiv 2024 |
| Share of notated chords written as inversions (a non-root note in the bass) | 2.8% (1,437,712 chords; our calculation) | Kantarelis et al., Chordonomicon, arXiv 2024 |
Context note: all five percentages divide the counts in Tables 1 to 3 of the Chordonomicon paper by its total of 51,007,915 chords; the paper prints the counts, not the percentages. The paper’s seven chord types (major, minor, suspended, power, half-diminished, augmented, diminished) add up exactly to that total, so “major type” includes chords whose base triad is major, whatever extensions they carry.
4. Harmonic Change by Decade: From Three Chords to Borrowed Color
The 1950s were the most harmonically compact decade in the Rolling Stone corpus: I, IV and V supplied more than 95% of the harmonic content of 1950s songs on Rolling Stone’s list. From the 1960s on, chords borrowed from the parallel minor key, above all bVII, entered the vocabulary and peaked in the 1980s, while IV gave up ground.
The seventh-chord story runs the other way. Audio analysis of the Hot 100 and a much larger chord-chart dataset agree on direction: dominant-seventh harmony, the sound of blues and jazz, faded steadily, while minor-seventh color surged during the funk, disco and soul years. The chord-chart analysis by Chris Dalla Riva uses the Chordonomicon dataset (about 680,000 songs taken from Ultimate Guitar) and puts numbers on the fall. Annotation-based work on 21,320 Hot 100 tracks from 1958 to 2022 places the highest harmony novelty peak (the point where harmony changed most sharply) in 2016.
| Metric | Value | Source |
|---|---|---|
| Share of harmonic content supplied by I, IV and V in 1950s songs, Rolling Stone corpus | More than 95% | de Clercq and Temperley, Popular Music 2011 |
| IV share of chord roots, 1950s songs vs 1990s songs, Rolling Stone corpus | 32.1% vs 18.7% | de Clercq and Temperley, Popular Music 2011 |
| bVII share of chord roots in the 1950s, at its 1980s peak, and in the 1990s, Rolling Stone corpus | 0.7%, 10.8% and 6.4% | de Clercq and Temperley, Popular Music 2011 |
| Change in mean frequency of the dominant-seventh harmonic topic (H1), Billboard Hot 100, 1960 to 2009 | Declined by about 75% | Mauch et al., Royal Society Open Science 2015 |
| Change in mean frequency of the minor-seventh harmonic topic (H3), 1967 to 1977 | More than doubled | Mauch et al., Royal Society Open Science 2015 |
| Share of all chords that are seventh chords, 1940s songs vs 2020s songs so far, user-written chord charts | 27.7% vs 8.25% | Chris Dalla Riva, Can’t Get Much Higher 2025 |
| Clustering coefficient of the pitch-transition network (lower values mean fewer interlinked pitch moves), 464,411 recordings, 1955 to 2010 | Fell from 0.65 to 0.45 | Serra et al., Scientific Reports 2012 |
| Years of major stylistic revolutions on the Billboard Hot 100 across all 58 annotated attributes, and year of the highest harmony novelty peak | 1964, 1983 and 2016; harmony peak in 2016 | Oramas et al., TISMIR 2025 |
Context note: the Mauch team describes H1 as the topic that “captures the use of dominant-seventh chords” and reads its decline as the fading of jazz and blues from the chart; see the open-access Royal Society Open Science paper. The Serra et al. study found that the ranking of the most-used pitch combinations barely moved (mean Spearman correlation 0.97 across years), so the restriction is in how pitches connect, not which ones are used. Most recent available data for the decade rows: Mauch et al. 2015 (data to 2010) and Serra et al. 2012 (data to 2010).
5. Loops, Repetition and Static Harmony
The four-chord loop is the defining harmonic shape of 21st-century pop, and the data now measures it directly. In the Yale-Geerdes corpus of 2002 to 2021 Year-End hits, 70.5% of chords sit inside looping passages, and the loops themselves repeat more within each song. Yet the loops are not interchangeable: in both eras roughly four in five distinct loop types (chord patterns counted regardless of key, which the authors call loopclasses) appear in only one song.
The interpretation flips a common complaint. Recent pop is not built from a handful of shared stock progressions; it repeats a song-specific loop harder. Chander and Quinn note that the most schematic (most widely shared) loop in the older corpus is a two-chord major shuttle, while in the newer corpus it is a form of the “Axis” progression, the family of loops built on vi, IV, I and V. Static harmony was already the norm decades ago: in the Hot 100 audio study, “major chords without changes” was the most common harmonic topic in 43 of 50 years.
| Metric | Value | Source |
|---|---|---|
| Share of chords inside looping passages: McGill Billboard corpus (1958 to 1991) vs Yale-Geerdes Billboard corpus (2002 to 2021) | 53.7% vs 70.5% | Chander and Quinn, ICMPC17-APSCOM7 2023 |
| Median loop repetitiveness score (TF, a term-frequency measure; higher means more repetitive), McGill vs Yale-Geerdes | 0.26 vs 0.41 (U = 372,890.5, p < .0001) | Chander and Quinn, ICMPC17-APSCOM7 2023 |
| Share of distinct loop types (loopclasses) heard in only one song, McGill vs Yale-Geerdes | 79.6% vs 77.0% | Chander and Quinn, ICMPC17-APSCOM7 2023 |
| Mean weight of the “major chords without changes” topic (H8) in Hot 100 songs, 1960 to 2010 | 0.236 (mean ± 95% CI: 0.236 ± 0.003), the highest of 8 harmonic topics | Mauch et al., Royal Society Open Science 2015 |
| Years in which H8 was the most common harmonic topic | 43 of 50 | Mauch et al., Royal Society Open Science 2015 |
| Share of Hot 100 songs with a substantial (>12.5%) H8 weight | Nearly two-thirds | Mauch et al., Royal Society Open Science 2015 |
| Year-End Hot 100 hits (2023 to 2025) classified as “Static-Loop” type (self-repeating, minimal chord vocabulary) | 12 of 178 (6.7%; our calculation) | Kim, International Journal of Contents 2026 |
| Mean number of distinct Roman-numeral chords per song: “Static-Loop” type vs “Dynamic-Complex” type | 4.083 vs 13.362 | Kim, International Journal of Contents 2026 |
Context note: the Chander and Quinn preprint compares 735 McGill songs (69,452 chords) with 537 Yale-Geerdes songs (62,675 chords). The Kim 2026 study built its 178-song sample from Ultimate Guitar chord charts, and its largest type (“Dynamic-Moderate”, 72 songs) makes up 40.4% of the sample, so pure loop songs remain a minority even in 2023 to 2025.
6. Harmonic Fingerprints by Genre and Region
Genre is where chord vocabulary splits most sharply. Seventh chords are 2.5 to 13 times more common in jazz than in any other genre in the Chordonomicon chord charts, and power chords are 2 to 21.5 times more common in punk. Country sits at the opposite pole from jazz: its five most-used major chords cover most of what is played.
Outside the Anglophone charts, harmony works differently again. A 2026 Music Theory Online study of Japanese anime opening songs finds key changes common, and most of them are not the stepwise “pump-up” (raising the key by one step, which Li notes is common in Western popular styles): over half move by a minor third, and the most popular place to change key is the jump into the chorus. Older audio-based chord extraction across a commercial music library also shows rock’n’roll as the genre most concentrated on its stock four-chord patterns.
| Metric | Value | Source |
|---|---|---|
| How much more common seventh chords are in jazz than in any other genre, user-written chord charts | 2.5 to 13 times | Chris Dalla Riva, Can’t Get Much Higher 2025 |
| How much more common power chords are in punk than in any other genre | 2 to 21.5 times | Chris Dalla Riva, Can’t Get Much Higher 2025 |
| Share of all chords played that come from the five most-used major chords: country vs jazz | 61% vs 39% | Chris Dalla Riva, Can’t Get Much Higher 2025 |
| Anime opening songs with at least one chromatic modulation (a change of key), 100-song corpus | 41% (21% with two or more) | Joy Li, Music Theory Online 2026 |
| Share of the corpus’s 72 modulations that move up or down by a minor third | 38 of 72 (52.8%; our calculation) | Joy Li, Music Theory Online 2026 |
| Share of modulations placed at the prechorus-to-chorus boundary vs stepwise “pump-up” modulations | 45.83% (33 of 72) vs 19.4% (14 of 72) | Joy Li, Music Theory Online 2026 |
| Share of pieces containing their genre’s prevailing four-chord patterns (audio chord extraction): rock’n’roll vs blues vs jazz | 68% vs 44% vs 43% | Barthet et al., CIM14 2014 |
Context note: the genre ratios come from Chris Dalla Riva’s analysis of about 680,000 songs, a newsletter analysis rather than a peer-reviewed study. The Joy Li article surveyed 108 opening songs and kept the 100 set in verse-prechorus-chorus form. The Barthet et al. Big Chord Data study ran automatic chord recognition on 997,580 pieces; most recent available data: 2014.
7. The Datasets Behind the Numbers
Every chord statistic is only as good as the corpus beneath it, and the corpora differ by orders of magnitude. Chordonomicon, which its authors call the largest dataset of its kind, holds 51,007,915 chords from 667,858 tracks, while expert-transcribed corpora such as the McGill Billboard sample or the Rolling Stone corpus hold between a few hundred and about a thousand songs. Big crowd-sourced sets give statistical power; small expert sets give accuracy and exact chart sampling.
The practical consequence is that headline numbers should be read with their method attached. A share computed from fan-made guitar charts, from expert listening, or from audio signal processing can differ even for the same songs. These same datasets also feed next-chord prediction models, a link explored in our procedural and generative game music statistics.
| Metric | Value | Source |
|---|---|---|
| Size of the Chordonomicon chord-chart dataset | 667,858 tracks; 51,007,915 chords | Kantarelis et al., Chordonomicon, arXiv 2024 |
| Distinct chord symbols in the Chordonomicon dataset | 736 (3,526 counting inverted chords) | Kantarelis et al., Chordonomicon, arXiv 2024 |
| Unique singles transcribed in the McGill Billboard sample, 1958 to 1991 | 1,100 singles by 533 artists, from 1,365 chart slots with audio | Burgoyne, Wild and Fujinaga, ISMIR 2011 |
| Distinct chords used per song, McGill Billboard sample | 11.8 on average (range 1 to 84) | Burgoyne, Wild and Fujinaga, ISMIR 2011 |
| Distinct chords and chord classes in the full McGill Billboard data, over 414,059 labeled beats | 638 chords; 99 chord classes | Burgoyne, Wild and Fujinaga, ISMIR 2011 |
| Billboard Hot 100 songs analyzed from audio, 1960 to 2010 | 17,094 songs (86% of the Hot 100) | Mauch et al., Royal Society Open Science 2015 |
| Songs and chord changes in the CoCoPops meta-corpus (Billboard plus Rolling Stone subsets) | 398 songs; 37,010 chord changes | Arthur and Condit-Schultz, ISMIR 2023 |
Context note: the McGill Billboard paper sampled 2,000 chart slots and found audio for 1,365 of them (68 percent). The CoCoPops paper notes the McGill and Rolling Stone corpora share only fifteen songs, so they are close to independent samples.
Summary: Pop Chord Progression Frequency by the Numbers
| Metric | Value | Source |
|---|---|---|
| Share of chords inside looping passages, 1958-1991 vs 2002-2021 Billboard corpora | 53.7% vs 70.5% | Chander and Quinn, ICMPC17-APSCOM7 2023 |
| Median loop repetitiveness score, 1958-1991 vs 2002-2021 | 0.26 vs 0.41 | Chander and Quinn, ICMPC17-APSCOM7 2023 |
| Share of chord roots that are I, IV and V, Rolling Stone corpus | 32.8%, 22.6% and 16.3% | de Clercq and Temperley, Popular Music 2011 |
| Share of chords before the tonic that are IV vs V | 39.6% vs 26.9% | de Clercq and Temperley, Popular Music 2011 |
| Probability that V follows IV then I, TheoryTab database | 41% | Hooktheory Trends API documentation |
| Most popular four-chord progression, TheoryTab database | I-V-vi-IV | Hooktheory blog, 1,300-song analysis |
| Share of notated chords that are major type vs minor type | 71.2% vs 25.7% | Kantarelis et al., Chordonomicon, arXiv 2024 |
| Seventh chords as a share of all chords, 1940s vs 2020s | 27.7% vs 8.25% | Chris Dalla Riva, Can’t Get Much Higher 2025 |
| Power chords in punk vs any other genre | 2 to 21.5 times more common | Chris Dalla Riva, Can’t Get Much Higher 2025 |
| Dominant-seventh harmonic topic, 1960 to 2009 | Declined by about 75% | Mauch et al., Royal Society Open Science 2015 |
| Pitch-transition network clustering coefficient, 1955 to 2010 | 0.65 to 0.45 | Serra et al., Scientific Reports 2012 |
| Year of the highest harmony novelty peak on the Hot 100 | 2016 | Oramas et al., TISMIR 2025 |
| ”Static-Loop” type among 2023-2025 Year-End hits | 12 of 178 | Kim, International Journal of Contents 2026 |
| Share of IV chords moving to I in the largest 2023-2025 hit cluster | 71.4% | Kim, International Journal of Contents 2026 |
| Anime opening songs with at least one key change | 41% | Joy Li, Music Theory Online 2026 |
| Modulations at the prechorus-to-chorus boundary | 33 of 72 | Joy Li, Music Theory Online 2026 |
| Distinct chords per song, McGill Billboard sample | 11.8 on average | Burgoyne, Wild and Fujinaga, ISMIR 2011 |
| Rock’n’roll pieces containing the genre’s prevailing four-chord patterns | 68% | Barthet et al., CIM14 2014 |
Methodology and Sources
- Mauch, MacCallum, Levy and Leroi, The evolution of popular music: USA 1960-2010, Royal Society Open Science 2015: audio analysis of 17,094 Billboard Hot 100 songs with eight harmonic topics.
- de Clercq and Temperley, A corpus analysis of rock harmony, Popular Music 2011, and de Clercq, Typical Chords in Typical Song Sections, EuroMAC 2014 presentation: expert harmonic analyses of songs from Rolling Stone’s “500 Greatest Songs of All Time” list.
- Hooktheory, Trends API documentation and I analyzed the chords of 1300 popular songs for patterns: user-contributed TheoryTab analyses.
- Kantarelis et al., CHORDONOMICON: A Dataset of 666,000 Songs and their Chord Progressions, arXiv 2024: chord charts scraped from user-generated sites.
- Chris Dalla Riva, I Analyzed Chord Progressions in 680k Songs, Can’t Get Much Higher, April 17, 2025: newsletter analysis of the Chordonomicon data.
- Chander and Quinn, The Decline of Harmonic Schemata in Popular Music Chord Loops, ICMPC17-APSCOM7 preprint, May 30, 2023: McGill vs Yale-Geerdes Billboard corpora.
- Sungbum Kim, Data-driven Classification of Harmonic Progression Types in Popular Music, International Journal of Contents, Vol. 22, No. 3, September 2026.
- Joy Li, Modulation and Chromatic Mediants in Japanese Anime Songs, Music Theory Online 32.1, March 2026.
- Serra, Corral, Boguna, Haro and Arcos, Measuring the evolution of contemporary western popular music, Scientific Reports 2012: 464,411 recordings from the Million Song Dataset.
- Oramas et al., MGPHot: A Dataset of Musicological Annotations for Popular Music (1958-2022), Transactions of ISMIR 2025.
- Burgoyne, Wild and Fujinaga, An Expert Ground-Truth Set for Audio Chord Recognition and Music Analysis, ISMIR 2011.
- Arthur and Condit-Schultz, The Coordinated Corpus of Popular Musics (CoCoPops), ISMIR 2023.
- Barthet et al., Big Chord Data Extraction and Mining, CIM14 2014.
- Data watch: dataset sizes conflict slightly. The Chordonomicon abstract says “over 666,000 songs”, the paper’s final count is 667,858 tracks, its Hugging Face release lists 679,807 rows, and the Dalla Riva analysis describes “nearly 680,000 songs” and “nearly 52 million chords” against the paper’s 51,007,915. For the 100-song Rolling Stone corpus, the 2011 paper prints V at 16.3% and bVII at 8.1% of chord roots while the 2014 presentation lists 16.2% and 8.0% for the same corpus; we cite the paper. Hooktheory’s API page gives example probabilities from an earlier, smaller TheoryTab library, and the TheoryTab and Rolling Stone shares for the I chord (17.6% vs 32.8%) differ partly because they count different material (TheoryTab analyses of individual song sections vs every chord root in whole songs, with major and minor roots combined). A 2016 Empirical Musicology Review replication by Hubert Leveille Gauvin reported converging evidence for the falling use of dominant-seventh chords and rising use of minor-minor seventh chords, using a contrasting method and an independent musical sample; we could not open its full text, so no numbers from it appear here. Several decade-trend datasets end in 2009 or 2010, so newer chart harmony data may change the picture; the 2023 to 2025 figures here rest on a single 178-song study.
Last updated: November 8, 2026. We update this roundup quarterly, and the next refresh is expected when the ISMIR 2026 conference proceedings are published.