THERE exists something of a paradox in the realm of popular music. On the one hand, people have made a habit of commenting on the (seemingly) limited harmonic vocabulary of the genre [2], and on the other, this (seemingly) limited harmonic vocabulary goes unnoticed by many. At least two comedy acts have gone viral using this premise as the idea for a song. One of these two viral acts, the band, Axis of Awesome [3], performed a mashup of 38 songs that all use the harmonic progression I V vi IV, later to be coined “the Axis progression” (Richards, 2017). By the time the group introduces the fifth song that uses the Axis progression, the audience starts to laugh, suggesting that the idea that the songs all use the same four chords is new to many. This paradox that a genre of music can be both filled with the same musical progressions while going entirely unnoticed by many listeners allows for an investigation into the role of musical training, all while focusing on musical training’s relationship to musical schema theory, memory chunking, and statistical learning.
Part of why these harmonic progressions work as a looped unit can be traced back to early studies on key and pitch perception. These early studies show how certain chords are perceived as more stable than others, at least within the realm of western classical music (Krumhansl, 1980, 1990; Krumhansl & Kessler, 1982; Krumhansl & Shepard, 1979). These profiles are influenced by genre, which is shown to have an effect on how individuals rate tonal stability. For example, the music of North India is shown to have its own tonal hierarchies (Castellan et al. 1984) and within a western context, the tonal hierarchy in pop and rock music is shown to be more egalitarian than its classical counterpart (Vuvan & Hughes, 2019, 2021). As to be expected, much more than the tonal profile goes into music perception. Metrical placement is found to have an effect on how individuals perceive stability in music (Temperely, 2018; Shea et al., 2023), as does musical familiarity (Jimenez et al., 2022) and experience (Creel et al., 2004). While a lot has been shown to go into how individuals perceive harmony, there seems to be a gap on how harmony is perceived, if at all, as larger progressions.
One such area of music theoretical research that might prove useful is schema theory (Byros, 2012; Gjerdingen, 2007). Gjerdingen (2007, 10) borrows the term schema from psychology and writes that they resemble mental representations. While introduced as a means for exploring 18th-century Galant progressions, it pairs well with the loop-based nature of popular music, which Biamonte (2010) argues is a feature of rock music, and Hughes and Lavengood (2019) using pop schemata as a section in their open access Music Theory textbook. One difference in the 18th-century schemata that Gjerdingen talks about and their popular music counterparts is that in the 18th-century these were paradigms that focused on the musical outer voices and were typically not repeated. However, in most popular music settings, the schemata used are often looped for entire musical formal units, thereby leaning into the idea that popular music only uses the same four chords. Schemata, in this modern instance, function as chunks, which as Lorch (2022, p. 692) notes, have historically served as a “process for perceptual grouping” in a musical context. Chunking has been argued to work at the melodic level. Deutch (1980) finds that participants were able to dictate melodies at a more successful rate when the notes were closer together. In returning to the Axis of Awesome performance, the audience laughs when they hear the beginning of the fifth song in the performance, suggesting they know how the loop or schema will play out (with a bit of help from dialogue by the band members). In this way, by considering schemata as units in and of themselves, it also relieves music listeners (both with and without training) of having to hear every chord as their own unit, or even as stable or unstable.
Consider the other viral popular performance mentioned earlier, this time by Rob
Paravonian and his song “Pachelbel Rant.” [4]. Judging by the laughter of the audience
in this example, it would appear that those in the audience can pick up on the musical
loops quickly. Here he uses another common chord progression, I V vi III, which is a
modern variant of Gjerdingen’s Romanesca schema. It does not take long for the
audience to recognize the continued repetition of the harmonic progressions used. This
makes sense given that studies have found that the learning of musical systems seems to
transcend training. Participants have been found to learn entirely made-up musical
systems with ease (Loui et al., 2010; Loui, 2022), which Gjerdingen and Borne (2015)
suggest is due to music’s connection to construction grammar. One of their key arguments
is grounded in “user-based knowledge,” where they argue that people learn musical
systems due to the use of those systems. This also, generally speaking, places popular
music in a different categorical realm than other genres, given its frequent occurrence
in day-to-day life, adding further curiosity to the paradox of how popular music can use
so few chord progressions, while also having so few people recognize them. Individuals’
ability to learn musical systems is further supported by work in corpus analysis of
popular music harmonies. deClerq and Temperely (2011) in their analysis of 100 songs
from the Rolling Stones “500 Greatest Songs of All Time” found that the IV chord was the
most likely chord to come after the tonic chord. deClerq (2017), in an expanded version
of the previous dataset, finds that form in pop/rock music can correlate with the time
spent on the tonic chord, particularly in regard to the verse and chorus.
Studies
often look at the role that musical training plays on participants, and results are
often mixed. Brown et al. (2012) found a slight impact of musicianship on the perception
of the sound of certain chords. Yet Hatmann et al. (2016) found no effects of
musicianship on hearing musical phrases. Most recently, Eitel et al. (2024) gave
participants two similar four-chord sequences and asked participants to indicate when
chords differed. They found that voice-leading distance, spectral pitch-class distance,
and chord transitional probability were all factors that could be modeled with their
paradigm, and all three of these factors also correlated with the Goldsmiths Musical
Sophistication Index (Müllensiefen et al., 2014). Results such as these suggest that
musical training could potentially have an effect on the perception of loop-based
harmonic progressions.
The current studies contribute to the above research by testing the role of musical training in the perception of popular-music harmonic schemata. We paired schemata from Hughes and Lavengood’s (2019) study with simple melodies as our stimuli. Our goal in building these experiments was to use simple versions of these schemata to see if those without musical training can pick up on these progressions in a simple context. While musical training is often explored through comprehensive approaches such as the Goldsmiths Musical Sophistication Index (Müllensiefen et al., 2014) and the Ollen Musical Sophistication Index (Ollen, 2006), Zhang and Shubert (2019) found that the best single item measure to ask participants is “what title best describes you” from the Ollen Musical Sophistication Index. Zhang and Schubert argue that this is because of its correlation with sophistication and ability to be demarcated into three musical identity levels. This is the approach that we take in our study. We use this, along with years of musical training in our experiment due to the length of time it would take to perform an exhaustive musical sophistication test, and the resources we had available.
All of our experiments included a trial portion. To the degree that we could, we wanted to remove the need for musical knowledge and vocabulary to maintain accessibility within our study. We made sure to explain key concepts such as “chord,” “melody,” and “chord progression.” Additionally, all answers to our experiments were audio samples to remove the need for musical knowledge and vocabulary. We began with two pilot experiments with smaller sample sizes to calibrate the difficulty of our procedure. These experiments were run with 33 participants each and used all the same stimuli and prompting that is used for the present two experiments. These pilot experiments produced a celling effect in that all participants did quite well with the paradigm. We used this to introduce a brief gap between our opening trial and the response option, and further prevented participants from returning to our opening trial so that they could relisten. In Experiment 1, we hypothesized that those with musical training would be able to match looped harmonic progressions with a melody to the same progression without a melody at a more successful rate than those without musical training. For Experiment 2, we hypothesized that those with musical training would be able to match looped harmonic progressions without a melody to the same progression with a melody at a better rate than those without musical training. Both experiments were run concurrently as different ways of asking the same research question.
Participants were recruited from Prolific. There were 107 participants. There were 43 males, 61 females, 2 nonbinary participants, and 1 agender participant. Their ages ranged from 18 to 67 (M = 32.47, SD = 11.05). Fifty-five participants were classified as musicians if they reported any amount of formal or informal musical training, and 52 were classified as nonmusicians. We excluded the scores of a single participant because they indicated they had no years of musical training but have been playing music since they were 13. This participant was not in the 107 mentioned above.
Testing took place using an online survey hosted on Qualtrics. We tested 9 chord progressions from Open Music Theory, “Introduction to Harmonic Schemas in Pop Music” (see Table 1; Hughes & Lavengood 2019). Our conditions for using progressions were that they needed to be able to be repeated in an eight measure phrase and have more than two chords so that the progressions are presented in two chunks. We excluded several progressions that did not fit the stimuli hypermeter, were already represented in the included progressions, or were too simple. Progressions that were three chords had their tonic repeated to fit a four-chord model.
Table 1. Included Progressions from “Introduction to Harmonic Schemas in Pop Music” (Hughes & Lavengood 2019).
| Included Progressions | Roman Numerals of the Progression |
|---|---|
| Double plagal | ♭VII–IV–I |
| Doo-wop | I–vi–IV–V |
| Singer/Songwriter | vi–IV–I–V |
| Hopscotch | IV-V–vi–I |
| Lament | i–♭VII–♭VI–V |
| Circle-of-fifths | i–iv–VII–III |
| Puff | I–iii–IV |
| Aeolian shuttle | i–♭VII–♭VI–♭VII |
| Lydian cadence | II♯–IV–I |
Using Musescore and the nine progressions, we created eighteen stimuli. Each progression was repeated twice using one chord per measure, totaling 8 measures, at 120 bpm. The chords were played in a blocked style. Each progression has two melodies, one that was ascending in nature, and one that was descending. The melodies were in a statement/restatement (aa’) style (see Figures 1a-c). In an attempt to be stylistic, the melodies were either stepwise or arpeggiated. The melodies were almost entirely quarter notes except in m. 7 on b. 3 where there is a half note. This was done to emphasize the aa’ form of the melody and end the phrase. All stimuli were in the key of C major and had a piano timbre. For a full list of the progressions and melodies, see Appendix A.
Fig. 1a. bVII IV I - Base Chords Example
Fig. 1b. I vi IV V - Ascending Example
Fig. 1c. IV V vi I - Descending Example
This survey began with an explanation of the essential terms “melody,” “chords,” and “chord progression” to make sure those without musical training understood what would be asked of them (see Appendix B). After completing a practice trial, participants moved onto the experimental trials. Participants completed nine experimental trials. In each trial, participants were presented with an audio example of a chord progression without a melody and told, “You will hear a chord progression repeated twice.” After clicking play and listening to the progression, participants would click next and could not replay the initial progression once they moved on from it. Then participants were presented with four audio examples and asked, “Looking below, which example has the matching chord progression?” Each audio example had a different chord progression with a melody. The questions and answers were randomized. The key difference from the pilot studies is that participants could not replay the initial progression once they moved on from it.
After the nine trials, participants were asked demographic information. They were asked if they had any formal and/or informal music training, and how years of music training they have (if any). They were also asked at what age they begin their musical training (if any). Participants finally self-reported how they view themselves as musicians (professional musician, semi-professional musician, serious amateur musician, amateur musician, music-loving nonmusician, nonmusician) using the Ollen Musical Sophistication Index (Ollen 2006), which Zhang and Schubert (2019) found to be the best predictive question for musical ability.
We found that the results for Experiment 1 were consistent with our hypotheses. Using a Welch Two Sample t-Test, we compared musicians (n = 55, M = 5.74, SD = 2.65) and nonmusicians (n = 52, M = 4.73, SD = 2.28) and found significant results, albeit barely, t(104) = 2.12, p = .036, d = 0.40.. The results can be seen in Figure 2. Due to the small sample size, we grouped participants into three groups based on their self-identity: pro musician (group 1) comprises professional musicians and semiprofessional musicians, amateur (group 2) comprises of serious amateur musicians and amateur musicians, and nonmusicians (group 3) comprises music loving nonmusicians and nonmusicians. We used an ANOVA to measure the difference between these three groups, shown in Figure 3, and found more notable results F(2,104) = 5.99, p = <.01, Cohen’s f = .30.
Fig. 2. Participants scores for Experiment 1 shown in box and whisker plots.
Fig. 3. Participants scores for Experiment 1 shown in box and whisker plots, grouped into those who identify as any form of professional, amateur, and nonmusician
Participants were recruited from Prolific. There were 90 participants. There were 41 males, 45 females, and 4 nonbinary participants. Their ages ranged from 19 to 64 (M = 32.2, SD = 11.78). Forty-seven participants were classified as musicians if they reported any amount of formal or informal musical training, and 43 were classified as nonmusicians. We excluded the scores of a single participant because they indicated they had musical training but never stated any training. This participant was not in the 90 mentioned above, nor do they seem to be the same individual excluded from Experiment 1.
These are the same as above in Experiment 1.
This survey began like Experiment 1, with an explanation of the essential terms “melody,” “chords,” and “chord progression” to make sure those without musical training understood what would be asked of them. Participants then took a trial question. In this trial, participants were told, “You will hear a chord progression and a melody together. The next question will ask you to select the example with the matching chord progression.” After listening to the progression, participants would click next and see the text “Looking below, which example has the matching chord progression?” This trial was the model used for Experiment 2. The experiment then began, asking participants nine questions. Participants were presented with a chord progression with a melody and asked to match it with one of four chord progressions without a melody, the inverse of Experiment 1. The questions and answers were randomized. The key difference from the pilot studies is that participants could not replay the initial progression once they moved on from it. After the nine questions, participants were asked the same demographic information in Experiment 1.
We found that the results for Experiment 2 were consistent with our first hypotheses. Using a Welch Two Sample t-test, we compared musicians (N = 47, M = 6.06, SD = 2.59) and nonmusicians (N = 43, M = 4.74, SD = 2.36) and found significant results, t(88) = 2.52, p = .013, d = .53. This can be seen in Figure 4. To assess differences based on their self-identity of musical ability, we grouped participants self-identity into three groups, pro musician (professional musician and semiprofessional musician), amateur (serious amateur musician and amateur musician), and nonmusician (music loving nonmusician and nonmusician). Unlike Experiment 1, participants' self-identity as musicians did not produce any significant results, F(2, 87) = 2.66, p = .07, Cohen’s f = .19. The results can be seen in Figure 5.
Fig. 4. Participants scores for Experiment 2 shown in box and whisker plots
Fig. 5. Participants scores for Experiment 2 shown in box and whisker plots, grouped into those who identify as any form of professional, amateur, and nonmusician
In these experiments, we found musical training to have a small to medium effect on the perception of popular-music harmonic loop-based progressions. However, in terms of performance, musicians average only about one additional correct trial in each experiment, suggesting that although musical training has an effect, its benefits are relatively modest. The descriptive scores of how participants self-identify continue to show this slight increase of only about one correct question more from nonmusicians to amateurs, and amateurs to professionals. It is also worth noting that those without any musical training score significantly better than chance during all experiments. A potential area for further inquiry would be to run this experiment with a larger number of questions, to get a greater understanding of the effect of musical training on the population. Additionally, replicating this study with a more thorough investigation into musical sophistication could also prove useful. It is currently unclear as to why the results of self-identity in Experiments 1 and 2 are different, but it could be due to the small sample size. It could also suggest that the style of Experiment 2 is harder than that of Experiment 1. Given that participants started by hearing more musical information in Experiment 2 (i.e., a melody and harmonic progression together), their focuses were split across two different musical items. Even still, the scores are quite similar across the two experiments.
These experiments did see a statistically significant difference in how musicians and nonmusicians scored. However, for these experiments, self-identity was not always found to be a key method for predicting participants scores. Taken together, our results suggest that some type of musical training plays a role in people’s ability to match loop-based harmonic progressions together. Two important points can be drawn from these findings. The first is that while the results from many of our hypotheses are significant, across both experiments, musicians and nonmusicians’ average scores were around the same. This is to say, participants are generally good at matching looped-based harmonic progressions together, regardless of musical training. The second is that both forms of our experiment (i.e., starting with just the harmonies, or starting with the harmonies with a melody) score about the same. This means that both of our experiments do address the questions we sought to ask.
The closeness of the results across both experiments matters because it suggests that musical training might not be the ideal way to judge and assess people’s ability to engage with music. While in both experiments we found significant results, both group’s average scores were still close, and they did not always hold up when considering self-identity. There were also nonmusicians who scored perfectly and musicians who got everything incorrect (in Experiment 1, only a single participant got everything incorrect, and they were a musician). This means there might be other determining factors, such as how much people listen to music, or the genres of music to which they listen. Phrased another way, formal musical training might not be a prerequisite for musical understanding. While using a musical sophistication test could provide further insight, it would still be necessary to disentangle which aspects of musical sophistication grant participants these abilities. However, given research on single item measurements for musical ability (Zhang & Shubert, 2019), it would still seem likely that participants would score the same regardless of how their musical ability is measured. Our future work in this area will include more questions about people’s musical habits to see if there are other factors that might indicate some level of their musical abilities beyond formal training.
The fact that the experiments swapped the type of presentation and test stimuli while maintaining similar scores suggests that people’s ability to match loop-based harmonic progressions is a flexible rather than task-specific. It also tells us that participants are not guessing on the answers and that they are aware of the musical features they are hearing, regardless of which experiment they took.
As stated in our introduction, there are limitations to our experiments, namely the simplicity. We used blocked chords with simple melodies and a slow harmonic rhythm. This was done on purpose so that we could build an initial understanding of the role of musical training in the domain of schema theory, but we acknowledge the pitfalls of this methodology. In future work, we plan to build on this by exploring musical stimuli that make use of actual music and not simple piano timbres. Another possible limitation is the number of questions asked. Perhaps if we used more stimuli, there might be a greater difference between musicians and nonmusicians. Future work will address these limitations.
Much of popular music is done in loops. The recognition of these loops, or schemata, is consistent throughout our experiments. This is probably due to the loop-based nature of popular music, which is a defining feature and not a flaw. Music genres often use the same couple of chord progressions through their genres. It stands to reason that these schemata bind the styles of these genres together. This means that it would behoove listeners, even those without musical training, to hear musical schemata as genre signifiers. As statistical learning would suggest, everyone takes in the musical information they hear over time. This is all to say, if the schemata, or loops, get lost in interpretation, the music might not get understood at all.
We would like to thank Oberlin College Conservatory for funding and support for this project. We would also like to thank Daniel Shanahan for early comments on this project, and Daniel Müllensiefen for his comments and encouragement on this paper.
[1] Correspondence can be addressed to: Dr. Samuel Gardner, Oberlin College Conservatory, 135 W. Lorain St. Oberlin, Ohio 44074, sgardner@oberlin.edu
[2] This can be seen in online publications such as Classical FM, and their 2019 article “These four chords are at the heart of every pop song.
[3] https://www.youtube.com/watch?v=OMshvUReunc
[4] https://www.youtube.com/watch?v=JdxkVQy7QLM&t=144s
Biamonte, N. (2010). Triadic modal and pentatonic patterns in rock music. Music Theory Spectrum, 32(2), 95–110. https://doi.org/10.1525/mts.2010.32.2.95
Byros, V. (2012). Meyer’s anvil: Revisiting the schema concept. Music Analysis, 31(3), 273–346. https://doi.org/10.1111/j.1468-2249.2012.00344.x
Castellano, M. A., Bharucha, J. J., & Krumhansl, C. L. (1984). Tonal hierarchies in the music of North India. Journal of Experimental Psychology: General, 113(3), 394–412. https://doi.org/10.1037/0096-3445.113.3.394
Creel, S. C., Newport, E. L., & Aslin, R. N. (2004). Distant melodies: Statistical learning of nonadjacent dependencies in tone sequences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30(5), 1119–1130. https://doi.org/10.1037/0278-7393.30.5.1119
de Clercq, T., Temperely, D. (2011). A corpus analysis of rock harmony. Popular Music, 30(1), 47–70. https://doi.org/10.1017/S026114301000067X
de Clercq, T. (2017). Interactions between harmony and form in a corpus of rock music. Journal of Music Theory, 61(2), 143–170. https://doi.org/10.1215/00222909-4149525
Deutsch, D. (1980). The processing of structured and unstructured tonal sequences. Perception & Psychophysics, 28(5), 381–389. https://doi.org/10.3758/BF03204881
Eitel, M., Ruth, N., Harrison, P., Frieler, K., & Müllensiefen, D. (2024). Perception of chord sequences modeled with prediction by partial matching, voice-leading distance, and spectral pitch-class similarity: A new approach for testing individual differences in harmony perception. Music & Science, 7. https://doi.org/10.1177/20592043241257654
Gjerdingen, R. (2007). Music in the galant style (1st edition). Oxford University Press. https://doi.org/10.1093/oso/9780195313710.001.0001
Gjerdingen, R., & Bourne, J. (2015). Schema theory as a construction grammar. Music Theory Online, 21(2). https://doi.org/10.30535/mto.21.2.3
Hughes, B., & Lavengood, M. (2021). Introduction to harmonic schemas in pop Music. https://viva.pressbooks.pub/openmusictheory/chapter/intro-to-pop-schemas/
Jimenez, I., Kuusi, T., & Ojala, J. (2022). Veridical and schematic memory for harmony in melody-and-accompaniment textures. Music Perception, 40(2), 89–111. https://doi.org/10.1525/mp.2022.40.2.89
Jordan Raskopoulos (Director). (2009, November 29). The axis of awesome 4 chords Melbourne comedy festival [Video recording]. https://www.youtube.com/watch?v=OMshvUReunc
Krumhansl, C. L. (1983). Perceptual structures for tonal music. Music Perception, 1(1), 28–62. https://doi.org/10.2307/40285249
Krumhansl, C. L. (1990). Tonal hierarchies and rare intervals in music cognition. Music Perception, 7(3), 309–324. https://doi.org/10.2307/40285467
Krumhansl, C. L., & Kessler, E. J. (1982). Tracing the dynamic changes in perceived tonal organization in a spatial representation of musical keys. Psychological Review, 89(4), 334–368. https://doi.org/10.1037/0033-295X.89.4.334
Krumhansl, C. L., & Shepard, R. N. (1979). Quantification of the hierarchy of tonal functions within a diatonic context. Journal of Experimental Psychology: Human Perception and Performance, 5(4), 579–594. https://doi.org/10.1037/0096-1523.5.4.579
Lörch, L. (2022). Chunking in tonal contexts: Information compression during serial recall of visually presented musical notes. Psychology of Music, 50(3), 691–708. https://doi.org/10.1177/03057356211013396
Loui, P., Wessel, D. L., & Kam, C. L. H. (2010). Humans rapidly learn grammatical structure in a new musical scale. Music Perception, 27(5), 377–388. https://doi.org/10.1525/mp.2010.27.5.377
Müllensiefen D., Gingras B., Musil J., Stewart L. (2014) The musicality of non-musicians: An index for assessing musical sophistication in the general population. PLOS ONE 9(2), e89642. https://doi.org/10.1371/journal.pone.0089642
Ollen, J. E. (2006). A criterion-related validity test of selected indicators of musical sophistication using expert ratings [The Ohio State University]. https://etd.ohiolink.edu/acprod/odb_etd/etd/r/1501/10?clear=10&p10_accession_num=osu1161705351
Richards, M. (2017). Tonal ambiguity in popular music’s axis progressions. Music Theory Online, 23(3). https://doi.org/10.30535/mto.23.3.6
RobPRocks (Director). (2006, November 21). Pachelbel Rant [Video recording]. https://www.youtube.com/watch?v=JdxkVQy7QLM
Shea, N., White, C., Hughes, B., & Vuvan, D. T. (2023). Metric accent affects perception of key center in pop-music chord loops. OSF. https://doi.org/10.31234/osf.io/spj6h
Temperley, D. (2018). The musical language of rock. Oxford University Press. https://doi.org/10.1093/oso/9780190653774.001.0001
Longdon, V. (2019) These four chords are at the heart of every pop song. Classic FM. Retrieved September 29, 2024, from https://www.classicfm.com/discover-music/music-theory/four-chords-every-pop-song/
Vuvan, D. T., & Hughes, B. (2019). Musical style affects the strength of harmonic expectancy. Music & Science, 2, 1–9. https://doi.org/10.1177/2059204318816066
Vuvan, D. T., & Hughes, B. (2021). Probe tone paradigm reveals less differentiated tonal hierarchy in rock music. Music Perception, 38(5), 425–434. https://doi.org/10.1525/mp.2021.38.5.425
Zhang, J. D., & Schubert, E. (2019). A single item measure for identifying musician and nonmusician categories based on measures of musical sophistication. Music Perception, 36(5), 457–467. https://doi.org/10.1525/mp.2019.36.5.457
♭VII–IV–I, Ascending
♭VII–IV–I, Descending
i–♭VII–♭VI–♭VII, Ascending
i–♭VII–♭VI–♭VII, Descending
i–♭VII–♭VI–V, Ascending
i–♭VII–♭VI–V, Descending
I–iii–IV, Ascending
I–iii–IV, Descending
i–iv–VII–III, Ascending
i–iv–VII–III, Descending
I–vi–IV–V, Ascending
I–vi–IV–V, Descending
II♯–IV–I, Ascending
II♯–IV–I, Descending
IV–V–vi–I, Ascending
IV–V–vi–I, Descending
vi–IV–I–V, Ascending
vi–IV–I–V, Descending
