Abstract
Past research showed that musical pieces exposed repeatedly are processed more fluently, which increases their liking, at least up to a certain number of repetitions. The present experiment focused on within-stimuli repetitions, that is, repeating musical units within songs, and examined the effects of five types of such repetitions on listeners’ aesthetic reactions. We also investigated processing fluency and perceived stimulus complexity as potentially opposing routes of these influences. Furthermore, we examined the differences between subjective and objective complexity in predicting aesthetic responses, and the relationships between individual musical expertise, preferences, and motivations and participants’ enjoyment of repetition in music. Musical stimuli, varying in repetition and objective complexity, were created using fifty motifs which were standardized and combined into larger stimuli across five repetition types, before being randomly presented in experimental blocks to avoid overexposure to any particular motif. Every stimulus was rated for liking, beauty, and interest, as well as processing fluency and subjective complexity. The main findings indicate significant variations in aesthetic ratings, processing fluency and perceived complexity between repetition types. Three of these repeating patterns increased processing fluency while decreasing perceived complexity, with opposite effects on aesthetic reactions and with perceived complexity having a stronger impact on aesthetic ratings than processing fluency. This suggests that the different types of repetition within songs generate different degrees of confirmation or violations of the expectations that listeners develop, and that variations in subjective complexity, associated with the disconfirmation of these expectations, weigh more than those in fluency for aesthetic appraisals.
Repetition is an inherit part of music, as it can be found at least once in 94% of studied music from across a plethora of cultures (Huron, 2006). Nowadays, music has become simpler and more repetitive than before (Parada-Cabaleiro et al., 2024; Yang, 2020), as a consequence of how music serves commercial purposes and how repetition helps songs place higher on music charts (Yu & Ying, 2015) and sell better (Nunes et al., 2015). The present study aims to investigate the effects of different types of repetitions within musical pieces on listeners’ aesthetic reactions and two potentially concurrent routes of these influences, that is, processing fluency and perceived stimulus complexity, while also considering musical expertise, preferences and motivations as individual characteristics that may be associated with enjoyment of repetition in music.
Types of within-stimuli repetition in music
Most often, research on music appreciation has examined the effects of between-stimuli repetition, by exposing participants multiple times to the same stimuli (Anand & Sternthal, 1991; Mungan et al., 2019; Peretz et al., 1998). The other type of repetition is within-stimuli, such as repeating musical units such as rhythm, melody or harmony at various points within the song or between songs which share stylistic traits and attributes (Titon, 2009).
For repetition within a song in particular, a motif could be repeated indefinitely. Verbatim repetition tends to engage listeners in moving, tapping or singing along with the songs (Margulis, 2014), and there are music genres which are predominantly repetitive, such as loop-based music genres, e.g., disco, minimalism, funk or hip-hop (Hainge, 2002). However, in practice, repetition is rarely presented in an exact form within a song. Most often, motifs or sections of a song are repeated at different points, and they can have large effects on listeners’ aesthetic reactions. For instance, Agres et al. (2017) found that in the case of trance music, participants enjoyed mostly chord progressions that are either very repetitive, such as AABB-AABB or ABCD-ABCD, or fairly complex, such as ABCD-EFAB or ABCD-BCDB; sequences such as AABB-AACC, which presented medium repetition, were not enjoyed as much. Other types of within-stimuli repetition include those used by Margulis (2013), who studied the aesthetic responses to music for either immediate or delayed repetition, and found that people find immediate repetition more enjoyable and artistic, while delayed repetition more interesting. These two types of repetition were also described by Ollen and Huron (2004) and labelled early and late repetition. The rondo repetition type shares the same form as early repetition; however, a new motif is introduced towards the end of the piece to dishabituate the listener (Huron, 2013; Ollen & Huron, 2004).
Facets of aesthetic appreciation of music and processing fluency
The literature describes several facets of the aesthetic appreciation of music (Brattico, 2015; Brielmann & Pelli, 2019). Besides liking, reflecting one’s personal enjoyment, beauty is thought to be the most important component of music’s aesthetic value (Istók et al., 2009). The two facets may, in some instances, be contradictory, as individuals can find artistic stimuli to be pleasurable but not necessarily likable (Muth et al., 2020). Thirdly, the Pleasure-Interest Model of Aesthetic Liking (PIA; Graf & Landwehr, 2015, 2017) highlights interest as the core component of the individual’s positive aesthetic reaction under conditions of focused perception and controlled processing.
People’s enjoyment of any stimuli partly depends on the fluency or ease they experience during their processing, that is, processing fluency (Graf et al., 2018). Generally, fluently processed stimuli are liked more (Forster et al., 2013; Westerman et al., 2015), and this has been observed for visual art (Reber et al., 2004; Vissers & Wagemans, 2021), as well as musical pieces exposed repeatedly, as shown by a recent review (Popescu & Holman, 2024). Nevertheless, listener’s type of engagement with the stimuli is an important factor for music, as liking either increases directly alongside fluency when people listen to the repeated stimuli inattentively while focusing on something else, or it increases until a point, plateaus and then decreases, resulting in an inverted U-curve, when people attentively listen to the stimuli (Chmiel & Schubert, 2017; Graf & Landwehr, 2015).
These findings are interpreted in the Habituation–Fluency Theory of Repetition (Huron, 2013) as indicating that processing fluency evokes a positive hedonic reaction in music listeners even during conscious perception, but stimuli repetition may also generate habituation, which leads to a reduction in responsiveness and thus to lower enjoyment. The PIA model (Graf & Landwehr, 2015, 2017) takes a different approach to these findings, asserting that aesthetic liking stems from processing fluency only in the unconscious processing mode. Oppositely, during conscious listening, aesthetic liking is rather driven by processing disfluency, which sparks feelings of novelty that, in turn, foster another facet of aesthetic appreciation noted above, namely interest. Furthermore, interest incites the perceiver to process the stimuli in more depth (Alter, 2013; Sung et al., 2022).
Repetition, fluency and complexity
Past research has shown that repeating musical stimuli increases the fluency of their processing (e.g., Mungan et al., 2019). Nevertheless, as highlighted above, this does not guarantee a parallel positive effect on enjoyment during focused listening. Besides examining the relationship between repetition-induced fluency and aesthetic appreciation of musical pieces including such repetitions, our study also aims to investigate the role of stimulus complexity in these dynamics. Complexity has also emerged as positively related to the enjoyment of visual stimuli such as graffiti (Husselman et al., 2024) and abstract art (Ball et al., 2018). This suggests that more complex music, which presents lower degrees of repetitiveness, may also receive higher aesthetic ratings (Marin et al., 2016).
These roles of fluency and complexity in music enjoyment may stem from the expectations listeners develop during the musical piece about its subsequent unfolding (Thorpe et al., 2012; Tillmann et al., 2014), and repetition can represent a base for such expectations. The violation of musical expectations generated by within-stimuli repetitions may decrease the fluency of processing the musical material while simultaneously rendering it subjectively more complex because of the uncertainty generated by its deviations from expectancies. Furthermore, past research found that listeners experience not only surprise but also enjoyment when their expectations are disconfirmed, and uncertainty and perceived unpredictability are associated with more positive aesthetic ratings (Cheung et al., 2023; Clemente et al., 2024; Gold et al., 2019, 2023).
Secondly, subjective or perceived complexity may differ from objective complexity (Heyduk, 1975), as any stimulus with a certain level of complexity might be appreciated as very complex by some individuals but simplistic by others. Research has highlighted several individual characteristics that affect how people interact with complex stimuli, such as expertise (Orr & Ohlsson, 2005), cultural background (Eerola et al., 2006) or contextual information (Carbon, 2023). Furthermore, this subjective appraisal of stimulus complexity may have direct effects on aesthetic appreciation, along a distinct yet concurrent pathway of influence to that of processing fluency. Listeners may appreciate musical pieces with a certain degree of complexity more than very repetitive and thus simplistic ones, specifically due to their perceived complexity and beyond the influence that processing fluency may have on their aesthetic reactions.
Individual characteristics associated with enjoyment of repetition in music
Musical expertise has been found to be a prevalent individual characteristic considered when studying processing fluency-based aesthetic liking (Popescu & Holman, 2024) because people with higher expertise have more experience in relation to complex music, which changes their perception of complexity (Güçlütürk & van Lier, 2019; North & Hargreaves, 1995). However, some individuals without formal musical training may still exhibit cognitive and perceptual abilities similar to those of musicians. This suggests that even in the absence of musical expertise, non-musicians can still display sophisticated musical behaviour, such as active engagement with music, accuracy of musical listening skills or singing abilities (Müllensiefen et al., 2014) or emotional responses to expectancy manipulations in music (Sauvé et al., 2018).
Another important aspect of peoples’ engagement with music on a daily basis is their preferences, which are thought to be stable over time and are often determined by social factors or the individual’s personality (Jacobsen & Beudt, 2017). Certain genres are considered to be more complex, for instance classical, jazz, blues and folk music, while others such as rap/hip-hop, soul/funk and electronica/dance music to be more energetic and rhythmic, but more repetitive (Rentfrow & Gosling, 2003). This suggests that people who enjoy reflective and complex music may appreciate more stimulus complexity as opposed to repetition.
Finally, our study also considers music use motivations as a factor of individual’s aesthetic reactions to music varying in repetitiveness. Schäfer and colleagues (2013) identified three general categories of such motivations: self-awareness, expressing a very private relationship with music listening, social relatedness, consisting of social bonding and affiliation, and arousal and mood regulation, consisting of using music as background entertainment or as a means to regulate emotions. This latter motivation was highlighted by several studies (Cook et al., 2019; Thoma et al., 2012), and it was suggested that people often use low complexity music for emotional purposes and more complex music for cognitive purposes (Sallavanti et al., 2016). Therefore, the prevalence of a certain type of music motivation may also relate to the individual’s habitual exposure to musical stimuli with a certain degree of repetitiveness, which may further influence one’s perception and aesthetic appraisals of musical pieces that vary on this characteristic.
The current study
This experimental study first aims to examine the effects of within-stimuli repetition on aesthetic appraisals of music in a sample of non-experts, that is, individuals with no formal musical training. To this aim, we compare five types of repetitions, that is, verbatim, alternate, early, late, and rondo on three facets of the aesthetic appreciation of music, that is, liking, beauty and interest. Building on previous results of comparisons between different types of repetition, we expect aesthetic appraisals to be the lowest for verbatim repetition and the highest for no repetition. In addition, previous research suggests that aesthetic appreciation is influenced by both processing fluency (the ease with which a stimulus is processed) and subjective complexity (the perceived unpredictability or intricacy of a stimulus) (Nadal et al., 2010; Reber et al., 2004). These variables function in parallel, with processing fluency typically enhancing aesthetic evaluations (Reber et al., 2004), and subjective complexity contributing to a more nuanced response, depending on context and individual differences (Leder et al., 2004). As such, our research further investigates the influence of the specific types of repeating pattern on processing fluency and perceived stimulus complexity as potential mediators of their aesthetic effects. In this respect, we expect the different types of repetition to vary in the processing fluency and the perceived complexity of the stimuli, with the no repetition stimuli being reported as the most complex and disfluent and the verbatim type stimuli as the least complex and most fluent. Furthermore, we hypothesize that liking, beauty and interest scores will be significantly predicted by processing fluency and subjective complexity, and that the relationship between repetition type and aesthetic ratings is mediated by processing fluency and subjective complexity.
We are also interested in studying the weight of both objective and subjective complexity in aesthetically judging musical stimuli. While objective complexity refers to measurable properties of the stimuli, such as pitch variance or rhythmic density (Clemente et al., 2020), subjective complexity reflects the listener’s perceptual and cognitive evaluation of the music. Since aesthetic judgements are shaped not only on structural features but also on individual interpretation and experience, we expect subjective complexity to account for more variance in aesthetic ratings than objective complexity (Güçlütürk & van Lier, 2019; Marin & Leder, 2013). As such, we hypothesize that regardless of repetition type, both objective and subjective complexity will significantly predict liking, beauty and interest ratings. However, we expect subjective complexity to predict a greater variance of aesthetic responses than objective complexity.
Method
Stimuli creation
As, to the best of our knowledge, there is no comprehensive database of musical stimuli that vary on how repetitive they are, we created our own stimuli. As a starting point, we used the MUST set (Clemente et al., 2020), which consists of 200 musical motifs varying in balance, contour, symmetry, and complexity. This database provides stimuli in the same key (C major) in MIDI format, which allows for easier manipulation and rendering in order to create coherent stimuli. We used only the 50 motifs that vary in complexity and while creating the stimuli, we used adjacent stimuli as to not create large musical contrasts between motifs, increasing the complexity. This also helps us in analysing how does subjective complexity differ from objective complexity (Marin & Leder, 2013) in evaluating the stimuli. As all the 50 motifs were also created to be fully symmetrical (meaning that the second half of the motif is exactly as the first half, but reversed), we decided to halve them in order to avoid overexposure and fatigue due to experiment length. We also standardized their length to one bar long at 120 beats per minute and we modified them to all start on the C4 note. Each motif was then analysed and adjusted so it would maintain its musical and rhythmic structure.
Following this initial cleaning step, we were left with 50 motifs that can be connected in order to create larger pieces that vary in repetition. Each motif retained its original name, with K1 being the least complex and K50 the most complex. To render the MIDI files into audio files, we used the Reaper digital audio workstation. The files were rendered into stereo .wav files, at a sample rate of 44,100 Hz and a bit depth of 24-bit PCM. The audio library used was the Autograph Grand sound library that is freely offered by Spitfire Audio.
We created the stimuli for the following repetition types, as derived from the literature on musical repetition (Huron, 2013; Margulis, 2013, 2014; Margulis & Simchy-Gross, 2016; Neuhaus, 2013; Neuhaus et al., 2009; Ollen & Huron, 2004:): verbatim repetition, alternate repetition (which is more of an extension made to verbatim repetition, made out of our curiosity), early repetition, late repetition and rondo repetition. For a more detailed account of how the stimuli for each repetition type were created, refer to Figure 1. Following the combination of motifs, we ended up with 245 musical fragments, each being 8 bars long with a duration of 18 seconds. The C4 note was placed at the end of each fragment to musically close it and to avoid increases in tension or expectancy from the participants, which were shown to affect how they aesthetically judge stimuli (Cheung et al., 2023; Wiersema et al., 2012). Control stimuli were also created, in which no motif was repeated. Given that the motifs used for the no repetition stimuli came from the same pool of motifs as the other repetition types, the motifs were reversed to avoid overexposure, as similarity and recognition judgements are highly dependent on the musical context in which they are made, meaning that there is a slim chance of them being perceived as similar between stimuli and within stimuli (Bartlett & Dowling, 1988; McAdams & Matzkin, 2001; Taher et al., 2018). To further avoid overexposure, we decided to create experimental blocks so that participants would be exposed to a motif only once. To create the blocks, the stimuli were grouped in such a way that each motif was present only once between all repetition types. These combinations were created using a matrix, starting at verbatim_K1 and rondo K1_K2_K3, respectively, in parallel (the selection matrix is attached as an auxiliary file to this article). This was done to ensure a somewhat even distribution of stimuli based on objective complexity. Selection was done diagonally until 5 stimuli were chosen (e.g., the first combination consists of verbatim K1, alternate K2_K3, early K4_K5, late K6_K7, rondo K8_K9_K10, while the sixth one consists of rondo K1_K2_K3, late K4_K5, early K6_K7, alternate K8_K9 and verbatim K10). If no complete combination of 5 stimuli could be made (meaning that one type of repetition was already used in another combination), we would skip until we found the next combination that allows for all five types of repetition to be present. At the end, all unallocated stimuli were discarded. We ended up with 29 combinations consisting of stimuli varying in repetition and objective complexity. Objective complexity was calculated by averaging the complexity rating of each motif used to create each stimulus (e.g., verbatim K1 has an objective complexity of 1, alternate K2_K3 has an objective complexity of 2.5, early K4_K5 has an objective complexity of 4.5, and so on). For each combination, a no-repetition stimulus was also inserted, resulting in a total of 168 stimuli (28 of each repetition type and 28 with no repetition).
Stimuli creation diagram.
To expose the participants to multiple stimuli of the same repetition type but different motifs and complexities, we created experimental blocks. This procedure resulted in 9 experimental blocks, varying in both repetition type and constituent motifs. Once again, unallocated stimuli were discarded. All blocks had 18 stimuli in total, with 3 different stimuli for each repetition type (except for the first block, which had 24 stimuli and 4 stimuli for each repetition type, as a result of the selection procedure).
Participants
A total of 298 students (M = 20.6, SD = 3.1; 87.9% women, 11.8% men and 0.3% non-binary; 65.26% urban) volunteered to take part in the experiment, in return for credit course. In the final data analysis, only participants with zero formal musical experience were included, and data from 81 participants who reported having at least a year of formal music training or a year of practical music training was discarded. Despite research stating that it takes about 6 years of music training to become a musician (Zhang et al., 2020), we adopted a more conservative approach in selecting our participants, similar to Correia et al. (2023). The final sample size consists of 216 participants with no formal musical experience (M = 20.7, SD = 3.5; 90.7% women; 62.42% urban). Demographic data can be consulted in Table 1. This study was conducted in accordance with the Declaration of Helsinki, and it was approved by the Ethics Committee of the Faculty (No. 650/20.05.2024), where the authors are affiliated.
Demographic characteristics of the participants (N = 216).
Data analysis
Analyses were conducted using the R Statistical language (version 4.4.1; R Core Team, 2024) on Windows 10 x64 (build 19045). A full list of the packages we used can be found in the R project file. For the mediation analysis, we used the PROCESS macro package for R (Hayes, 2022).
Procedure
Before starting the experiment, each participant was seated at a computer and handed a pair of headphones (Beyerdynamic DT 770 Pro, 80 Ohms). After being generally informed about how the experiment would run, the experiment was initiated on every computer. The experiment was created using the PsychoPy programme (Peirce et al., 2019).
Within the experiment, each participant was instructed to adjust the volume according to their preferences. They were then presented with a few slides intending to familiarize them with the listening and rating procedures. During the audition, each participant had to look at a fixation cross situated in the middle of the screen. Stimuli were randomly presented. Participants were also instructed to press the space bar on their keyboard the instant they felt that they like the stimulus they are currently listening to and, after pressing, to continue listening to it until it ended. This key press gave us a spontaneous felt liking rating, which could later be used to map the moment in which conscious liking ratings were made based on the type of repetition.
For the rating procedure, each participant had to rate each stimulus based on how much they liked it, as well as how beautiful, interesting, complex and fluent (defined as easy to listen) they found it. Ratings for each question were made on a 10-point Likert-type scale, with 1 being the lowest possible score and 10 the highest. After answering all questions, the next stimulus would automatically play (after a blank period of 500 milliseconds). These listening and rating procedures would loop until all stimuli from the experimental block were exhausted, and then this first, experimental, part of the study would end and participants were redirected to its second part of the experiment. In total, the experiment would last around 20 minutes.
Measures
In the second part of the study, participants were asked to respond to questions regarding demographic data (age, gender, residence, marital status, studies, personal income and religion), and then to fill in the Music Use and Background Questionnaire (MUSEBAQ) developed by Chin et al. (2018). This scale measures music engagement on 4 different modular dimensions: musicianship, musical capacity, music preferences and music use motivations. For the third module (music preferences), we opted to present the participants only with the broad categories (e.g., ‘How often do you listen to rock and metal music?’), without going into detail about specific genres (for the broad categories, if the participant reported that they listen to rock or metal music ‘Often’ or ‘Always’, they would have been prompted to report how much they listened to specific genres such as ‘Rock and roll’, ‘Soft Rock’ or ‘Classic Rock’, to name a few). This was done in order to avoid confusion regarding specific subgenres that are almost non-existent in the Romanian culture (e.g., ‘Early music’, ‘Britpop’, ‘Easy listening’, ‘Motown’, etc.). After completing the questionnaire, each participant was debriefed and allowed to leave the lab. This part of the study lasted around 5 minutes.
Results
Preliminary data analysis
Power analysis
Power analysis was done using the pwr R package. For the group comparison analyses, in order to observe a medium effect size at a power level of .80 and a significance level of .05, the pwr.anova.test function suggests that a minimum of 36 participants in each group is required (or a total of 216 participants). For the correlation analyses required for the mediation analysis, in order to observe a medium effect size at the same power level of .80 and significance level of .05, the pwr.r.test function suggests a total of 85 participants are necessary. Thus, our total sample size of 216 participants is able to capture medium effect sizes, as well as large and small effect sizes.
Normality of data
Normality of data was determined by using the dlookr R package. We both visually and statistically inspected the responses to our variables in order to test for normality. The function normality and the Q-Q plots created by the plot_normality function indicated that our measured variables are not normally distributed.
Correlations between measures
We found significant positive and strong correlations between liking, beauty and interest ratings for all repetition types. All these three aesthetic measures were also positively associated with subjective complexity and processing fluency, respectively, except the nonsignificant correlation between fluency and interest in the case of the alternate repetition stimuli. No significant correlation was found between processing fluency and subjective complexity. regardless of repetition type. A more detailed view over the correlations between the variables can be seen in the correlation matrix in Figure 2.
Correlation matrix between processing fluency, subjective complexity, liking, beauty and interest, separated by each experimental condition.
Within-stimuli repetition and processing fluency
We tested the effect of within-stimuli repetition on processing fluency using the nonparametric Kruskal–Wallis test (see Figure 3). The test revealed significant differences between repetition types when it comes to processing fluency, χ2(5) = 16.78, p = .005. A Wilcoxon Signed Rank test (Table 2) revealed significant differences specifically between verbatim repetition type and the no repetition stimuli, as well as verbatim repetition type and late repetition type, with higher processing fluency in the verbatim repetition than in the other conditions. The differences between alternate repetition type and the no repetition stimuli as well as the difference between alternate repetition type and late repetition type were also significant.
Visual summary of the mean differences between each experimental condition, separated by processing fluency, subjective complexity, liking, beauty and interest.
Mean and standard deviation of every dependent measure for each repetition type, as well as mean pairwise differences between conditions.
p < .05, ** p < .01, *** p < .001.
Within-stimuli repetition and subjective complexity
We ran the same Kruskal–Wallis test to examine the effects of within-stimuli repetition on subjective complexity. The test revealed significant differences in subjective complexity between repetition types, χ2 (5) = 288.84, p < .001. Compared to the no-repetition stimuli, a Wilcoxon Signed Rank test showed significant differences for every other repetition type, with the highest subjective complexity scores being observed for the no-repetition stimuli (Table 2). In addition, the verbatim repetition type significantly differs from alternate, early, late and rondo repetition types. In these instances, verbatim repetition had the lowest subjective complexity. Rondo repetition stimuli were also rated as being significantly more complex than those containing alternate, early and late repetitions.
To summarize the first two sets of results, we found that repetition positively influences processing fluency, while negatively influencing subjective complexity. Specifically, the no-repetition stimuli were the most complex and the least fluent, while verbatim-repetition stimuli were the least complex but among the most fluent.
Within-stimuli repetition and aesthetic responses
Differences between each repetition type in liking, beauty and interest ratings were tested with Kruskal-Wallis tests. The test revealed significant differences between repetition types for liking, χ2(5) = 87.63, p < .001, beauty, χ2(5) = 92.59, p < .001 and interest, χ2(5) = 124.71, p < .001. Looking at liking in particular, a Wilcoxon Signed Rank test (Table 2) revealed that the no repetition stimuli and the rondo repetition are liked more than all the other repetition types. At the opposite end, alternate, early, late and rondo repetition types significantly differ from the verbatim repetition, with the latter being the least liked. Next, examining beauty ratings, a Wilcoxon Signed Rank test mirrored the results obtained for liking, the no repetition stimuli and the rondo repetition were appraised as more beautiful than the other repetition types, while verbatim repetition was considered the least beautiful. Similarly, the results of the Wilcoxon Signed Rank test on interest ratings highlighted the same pattern of differences, with the most interesting stimuli being the no repetition and the rondo repetition stimuli, and the least interesting stimuli being the verbatim repetition stimuli.
Mediation analyses
We further examined whether the relationships between various repetition types (verbatim, alternate, early, late, rondo) and liking, beauty and interest ratings are parallelly mediated by processing fluency and subjective complexity. As repetition type is categorical, it was dummy coded, and the no repetition stimuli were set as the comparison group. We also included each MUSEBAQ dimension as covariates to control for any effect these dimensions might have on the relationships we examine. To calculate the indirect effects, each analysis was bootstrapped 10,000 times.
The results of the three mediation analyses were similar. For the indirect paths of repetition type on liking, beauty and interest through processing fluency, the mediation models revealed significant relationships for verbatim, alternate, and rondo repetitions. The coefficients of the individual paths (see Tables 3, 4) showed that these types of repetition generate higher processing fluency in comparison to the no repetition stimuli, and that the three aesthetic ratings were positively predicted by processing fluency. Examining the effects of repetition type on liking, beauty and interest through subjective complexity, we found significant indirect effects for all five repetition types, which were evaluated as lower in subjective complexity than the no repetition stimuli, while subjective complexity positively predicted all three aesthetic ratings. It is also worth highlighting that for all three aesthetic appraisals, the differences between the regression coefficients indicate that subjective complexity had stronger effects on these ratings than processing fluency Finally, the rondo repetition also had significant positive direct effects on liking and interest, but not on beauty ratings.
Results of the mediation analysis for repetition type (IV) and liking, beauty and interest ratings (DVs) through processing fluency (M1) and subjective complexity (M2).
Note. CI: confidence interval.
p < .05, ** p < .01, *** p < .001.
Coefficients for the direct effects (a and b), for repetition type (IV) and liking, beauty, and interest ratings (DVs).
Note. CI: confidence interval.
p < .05, ** p < .01, *** p < .001.
Objective versus subjective complexity
We found that objective and subjective complexity are positively correlated, r = .29, p < .001, suggesting that there is a slight overlap between the two. To test which kind of complexity is more important in the aesthetic appraisal of musical repetition, we included them as simultaneous predictors in three separate linear regression analyses, one for each aesthetic response (liking, beauty and interest). The results revealed that for all three aesthetic responses, subjective complexity predicts more variance compared to objective complexity. Specifically, subjective complexity significantly predicts 39% of the liking variance, 41% of the beauty variance and 51% of the interest variance, while objective complexity significantly predicts 2% of the liking variance, 2% of the beauty variance and 3% of the interest variance. In the case of processing fluency, both subjective and objective complexity significantly predict 1% of the total variance of processing fluency. All regression coefficients are reported in Table 5.
Regression coefficients for subjective complexity and objective complexity for each dependent measure.
Note. SD: standard deviation; CI: confidence interval; SE: standard error.
p < .05, ***p < .001.
Discussion
The main objective of this study was to investigate how different kinds of within-stimuli repetition elicit liking, beauty and interest when people listen attentively to music. Past research on fluency-based liking of between-stimuli repetition showed that liking increases up to a certain number of such repetitions. Our study examined the aesthetic effects of within-stimuli repetitions operationalized through five specific types of repeating patterns and further investigated the perceived complexity of the stimuli alongside processing fluency as potential mediators of these effects. Overall, our findings indicated consistent variations in aesthetic appraisals according to the type of within-stimuli repetition used, with the stimuli including no repetition and the rondo repetition ones being enjoyed more than all the other repetition types, while the verbatim repetition stimuli received the lowest aesthetic ratings. Furthermore, processing fluency and subjective complexity were found to exert significant and opposing indirect effects in these relationships between repetition and aesthetic appraisals.
Firstly, stimuli in the verbatim repetition and alternate repetition forms were considered as significantly more fluent than those in the no-repetition form. This suggests that repeating the same motif (verbatim) or alternating between two motifs (alternate) facilitates ease of processing compared to presenting entirely new material throughout. The alternate repetition stimuli, in which two motifs are repeated together, creating a larger sequence, are also experienced more fluently than others, particularly in comparison to the late repetition type, suggesting that presenting a repeating pattern later in the sequence hinders ease of processing compared to simply alternating between two motifs. Although the rondo repetition stimuli had the highest mean fluency rating, this difference was not statistically significant when compared to all other repetition forms and no repetition. Overall, these differences indicate that within-stimuli repetition can influence processing fluency, paralleling the effect of between-stimuli repetition (Popescu & Holman, 2024). Secondly, regarding subjective complexity, we observed that the no-repetition stimuli were thought to be the most complex, while the verbatim stimuli were the least complex. The differences between the types of repetition we examined further indicate that interrupting the exact repetition of the motifs at different points leads to changes in perceived complexity, but results in changes in processing fluency only for certain repetition types; for example, rondo repetition increased perceived complexity without reducing processing fluency, while verbatim and alternate repetition both decreased perceived complexity and increased processing fluency.
Furthermore, the results of the mediation analyses on liking, beauty and interest ratings revealed that for the verbatim, alternate and rondo repetition types, both processing fluency and subjective complexity mediated the effects of repetition on how much participants liked, as well as how beautiful and interesting they found the musical stimuli. Specifically, these repeating patterns increased processing fluency, thus making the stimuli more pleasurable. At the same time, they also showed indirect negative effects via subjective complexity, meaning that higher perceived complexity reduced liking, even when overall complexity was not always lower. For example, the verbatim repetition stimuli, while being among the most fluently processed and the least complex, were rated lowest in liking, beauty, and interest. In contrast, rondo stimuli, which were rated as more complex, still benefitted from increased fluency and showed positive overall effects on aesthetic appraisal. These findings complement our results that indicated subjective complexity is a stronger predictor of aesthetic ratings than processing fluency. Together with the low aesthetic status of verbatim repetition, they suggest that the variations in subjective complexity weigh more than those in fluency for aesthetic appraisals.
It is worth noting that all participants in this study were non-musicians. Despite this, they demonstrated sensitivity to differences in repetition structure, complexity, and fluency, and were able to make differentiated aesthetic judgements. This supports previous work suggesting that musical understanding is not limited to formally trained individuals. Non-musicians can still display varying degrees of musical sophistication, such as perceptual accuracy, which may shape how they experience and aesthetically evaluate musical stimuli (Müllensiefen et al., 2014; Sauvé et al., 2018). The present findings indicate that formal expertise is not a necessary condition for responding to structural features of music in meaningful ways.
Fluency in processing musical material and its subjective complexity are related to the confirmation or violation of the expectations that the individual develops during listening. Past research showed that while listening to unfolding music, people develop expectations about what and when will happen musically (Thorpe et al., 2012; Tillmann et al., 2014) and violations of this expectancy result in greater pleasure (Cheung et al., 2023; Gold et al., 2019). Uncertainty and surprise engage the reward neural circuitry and enhance liking ratings (Cheung et al., 2019; Gebauer et al., 2012; Gold et al., 2023), and perceived unpredictability was found to mediate the relationship between stimulus properties and liking (Clemente et al., 2024). Furthermore, they also develop expectations about how easy to listen would the song be, that is, fluency expectations, and their disconfirmations also contribute to aesthetic reactions by increasing interest (Jiang & Hong, 2014). In the same realm, past studies suggested that disfluency can also be valuable in processing a song (Yoo et al., 2023). Our results suggest that repetition is a base for the generation of expectations and support these positive aesthetic effects of expectancy violations, while also indicating processing fluency and the stimuli subjective complexity as experiential underpinnings of these effects.
The different types of repetition we investigated vary in the in the manners in which they create and violate expectancies. Verbatim repetition entails the simple reoccurrence of the preceding motif (similar to between-stimuli repetition); thus, the musical material includes no disconfirmation of the early generated expectancies. As listeners anticipate what musical passage is going to be played next, they get bored due to the lack of perceived complexity and disfluency (Graf & Landwehr, 2015), at this boredom reflects in their aesthetic judgement. In this respect, our findings on the detrimental effect of verbatim repetition on liking are in line with past research on aesthetic ratings in relation to between-stimuli repetition and processing fluency (Huron, 2013; Nunes et al., 2015; Reber et al., 2004). In the case of alternate repetition, however, its effects can be explained through the call and answer technique, often used in music composition, where a motif serves as a ‘call’ that creates tension, only for a second motif, serving as ‘answer’ to release the tension. The interlapping motifs create expectation and a context in which they are evaluated and, while at first unpredictable, they become predictable after a short time. For the rondo repetition, the unpredictability is kept high by introducing a novel motif towards the end of the stimulus as to dishabituate the listeners. As such, their expectations are violated, which further heightens their positive aesthetic reactions.
Broadly, our results suggest that disfluency, rather than fluency, is at play in aesthetically judging repeating music motifs. When listening attentively, the patterns of repetition that violate expectations and increase unpredictability and thus subjective complexity result in a more positive aesthetic appraisal. This suggests that for music to be appreciated the best, it needs to be presented in such a way to avoid exact repetition within a song, consistent with past suggestions that positive aesthetic responses originate from the tension between the need for repetition and the need for variety in music (Percino et al., 2014). Alternatively, for exact repetition to be pleasurable, it needs to be experienced outside the listener’s focus of attention, which is more suited for inducing relaxed and unfocused or trance states (Huron, 2013).
Practical applications
Our results provide musicians in particular information on how breaking exact repetition with new musical passages could, in fact, enhance liking for a song by fostering interest in the listener, despite making the musical piece more complex. There is a long-standing idea that making music as repetitive as possible would lead to more engagement, which is backed up by repetitive songs being placed higher on charts (Yu & Ying, 2015), which creates a precedent of composing formulaic music for the sake of maximum engagement. Our research provides evidence against this idea and shows that exact repetition is not as pleasurable for listeners as other forms of repetition, such as early, late or rondo, particularly for those who listen to music for social reasons, to regulate emotions, and to achieve a transcendental state.
Limits and future directions
Given there is no comprehensive database of stimuli that objectively vary on how repetitive they are, we resorted to creating our own to study different types of repetition. However, we believe that a general ‘repetitiveness’ parameter may have been better suited to classify stimuli from the least to the most repetitive, similarly to how stimuli could be arranged from the least to the most complex. In addition, as our stimuli were all in the key of C major, there is another facet of musical emotional valence that our study didn’t explore: that of minor keys. In general, major keys are attributed to happy music, while minor keys, to sad music and research shows that, in general, people tend to like happy music over sad music (Bella et al., 2001). However, at times, the opposite is true, where people enjoy and appreciate sad music more (Kawakami et al., 2013; Sachs et al., 2015; Vuoskoski et al., 2012). While our results offer new insights into how internal repetition of major musical stimuli affects their aesthetic ratings, it would be interesting to study how repetition of minor musical stimuli influences enjoyment.
Conclusion
Our results indicate that within-stimuli repetition, similar to between-stimuli repetition, can foster processing fluency but at the same time decrease the perceived complexity of musical stimuli, with significant effects on listeners’ aesthetic reactions. Processing fluency and subjective complexity were found to mediate the aesthetic effects of very repetitive music, in the case of verbatim and alternate repetition, as well as music that implies repetition while incorporating novel motifs that promote dishabituation, such as in rondo repetition. Our findings also confirm the positive influence of processing fluency induced by within-stimuli repetition on music enjoyment, but they also highlight perceived complexity of the musical material as a stronger factor of aesthetic appraisals in the case of such repetitions. Exact repetitions render the musical piece as too simplistic, that is, low in subjective complexity, which further drastically affects aesthetic reactions despite the parallel gains in processing fluency.
Footnotes
Ethical Considerations
This study was conducted in accordance with the Declaration of Helsinki, and it was approved by The Ethics Committee of the Faculty (No. 650/20.05.2024) where the authors are affiliated.
Consent to Participate
All participants provided written informed consent prior to enrolment in the study.
Consent for Publication
All participants provided written informed consent prior to enrolment in the study.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.