Non-musicians recognize unfamiliar contemporary classical music excerpts with increasing repetition

Abstract

This article aims to determine the effects of exposure and title information when applied to music excerpts that are unfamiliar both in terms of the single excerpt, and in the underlying tonal structures used (i.e. contemporary classical music). Twenty-three non-musicians participated in a two-session experiment: 32 musical excerpts were presented on Day 1 immediately followed by presentation of a title varying by type (no title, descriptive, semantic, affective). Music excerpts were varied by number of exposures (1 vs. 3 exposures). Participants were asked whether they remembered the excerpt from earlier in the experiment, and to rate their liking on a 7-point Likert scale. On Day 2, recognition for the 32 “old” (Day 1) excerpts was tested, along with 32 “new” excerpts, which were all presented without titles. Participants were also asked to rate their liking for the excerpts, as per Day 1. Accounting for response bias, recognition was above chance level across all conditions. Results indicated a significant effect of exposure on recognition ( $η_{p}^{2}$ = .79). No other effects were significant. Examining the liking ratings of accurately recognized excerpts, a significant interaction between exposure and title was found ( $η_{p}^{2}$ = .20), where the difference in ratings across exposure levels for affective titles was greater than that of the semantic titles (d = .60). This suggests that the title type, although not influencing memory of the piece, affected enjoyment of music at different exposures.

Keywords

contemporary music exposure liking memory titles

Repetition and liking are related such that listeners like a piece of music they can remember and consequently remember pieces of music that they like (Stalinski & Schellenberg, 2013). While repetition is a strengthening factor in contributing to higher rates of recognition for unfamiliar music (Deffler & Halpern, 2011; Szpunar, Schellenberg, & Planer, 2004), it does not serve as the sole determinant. Listeners’ ability to recognize music is also influenced by the quality of their listening condition (focused vs. incidental) as well as their perceived emotional status of the music (happy vs. sad) (Schellenberg, Peretz, & Vieillard, 2008). Similarly, listeners’ music preferences are influenced by a combination of intrinsic features (e.g., music structure and style) and extrinsic factors (e.g., prior exposure to a piece of music, programme notes) (Juslin, 2003; Kroger & Margulis, 2016). These findings highlight that recognition and liking for music are influenced by a complex interplay between structural qualities inherent within a piece of music, as well as the listener’s context.

Associations between music and text (mainly concerning lyrics) have also been shown to influence recognition, where participants were able to better remember songs in original music and text pairings, than in mismatched pairings (Crowder, Serafine & Repp, 1990; Serafine, Crowder, & Repp, 1984). This finding confirms that two elements presented closely together in time become associated in memory, such that memory for one is aided by the presence of another. In terms of memory for songs specifically, both text and song fragments similarly aid recognition (Peretz, Radeau, & Arguin, 2004). Furthermore, extra-musical information in the form of programme notes (Margulis, Kisida, & Greene, 2015) and subtitles (Silveira & Diaz, 2012) has been shown to influence listeners’ perceptions during music listening.

The effects of repetition and complexity on music recognition

While repetition has been demonstrated as an effective technique for improving memory for unfamiliar music (Deffler & Halpern, 2011), it has been found that stimulus complexity also influences recognition for both auditory and visual stimuli (Bornstein, 1989). Both the mere exposure (Zajonc, 1968) and two-factor models (Berlyne, 1970) posit that complex stimuli require more learning in comparison to stimuli which are simple (Montoya, Horton, Vevea, Citkowicz, & Lauber, 2017). Berlyne’s two-factor model posits that two opposing processes (habituation and tedium developed towards a stimulus) are operating separately with repeated exposure. The more complex a stimulus is, the more exposures it will take for a listener to habituate, and equally the more exposures it will take for a listener to experience tedium. Like the two-factor model, the perceptual fluency model proposes that simple stimuli are learned more rapidly compared to complex stimuli (Montoya et al., 2017). Consequently, smaller numbers of exposures are sufficient for simple stimuli to reach peak fluency, in comparison to stimuli that are more complex (Winkielman, Schwarz, Fazendeiro & Reber, 2003). For musical stimuli, it has been observed that increased exposures led to better memory for orchestral pieces (Szpunar et al., 2004), suggesting that repetition is vital for establishing familiarity with musical pieces which are complex. Therefore, it is expected in the current study that repeated exposures will result in greater familiarity with the excerpts and consequently greater recognition, compared to one exposure.

The role of repetition and stimulus complexity in influencing liking

The link between repetition, music complexity and liking has been explored extensively in the literature: findings support that the complexity of the stimulus moderates the extent to which liking increases and decreases as a function of repeated exposures (Montoya et al., 2017). In other words, listeners often prefer music they perceive as not too complex nor overly simple (Heyduk, 1975; North & Hargreaves, 1995). Although the mere exposure theory proposes that liking increases linearly as a function of exposure (Zajonc, 1968), Berlyne’s (1970) two-factor model proposes that liking forms a Wundt inverted-U curve representation which is dependent upon the complexity of the stimulus. When stimuli are simple and more prone to satiation, listeners are quicker to reduce their liking of a musical excerpt after a small number of exposures (Berlyne, 1970; Montoya et al., 2017). Conversely, increases of liking for more complex stimuli require a larger number of exposures (Montoya et al., 2017).

Regarding liking for musical stimuli, the optimal complexity model (Heyduk, 1975) postulates that music preference is inversely related to a combination of the individual’s perceived complexity and their preferred level of complexity of the stimulus. Findings from Hargreaves (1984) reveal that the inverted-U paradigm did not necessarily apply when comparing liking ratings across different musical genres. The study utilized four musical styles that were distinct from one another in complexity: pop and “easy listening” (less complex), and classical and avant-garde jazz (more complex) (Hargreaves, 1984). The authors found that in the case of the avant-garde pieces, with higher perceived complexity, liking ratings remained at low levels and did not increase with repetition. This finding contradicts the inverted-U paradigm which postulates that repetition should decrease subjective complexity over time and therefore increase initial low levels of liking ratings. However, for the other pieces (pop, “easy listening”, and classical), liking ratings followed an inverted-U shaped function.

One recent study challenges the models of optimum complexity and Wundt curve representation of familiarity and liking, where the findings indicated that familiarity is the strongest predictor for music liking, independent of their level of complexity (Madison & Schiölde, 2017). The study utilized music excerpts of varying genres (e.g., jazz and progressive rock) which had been pre-rated for their level of complexity by a group of musical experts. Findings indicated that liking increased monotonically after repeated presentations, across varying levels of complexity. However, the study reported results of only 15 participants, who themselves had a wide range of musical training. This heterogeneity of expertise in the participant sample may have diluted any effects of complexity present.

The set of experiments conducted by Szpunar et al. (2004) demonstrated that the satiation effect is strengthened with more ecologically valid and complex excerpts (i.e., orchestral and chamber music from the Baroque, Classical and Romantic periods) as compared to simple excerpts (short single tone sequences played with a piano tone). Using a 7-point Likert scale to record participants’ liking of each excerpt, findings revealed that repetition of complex excerpts (either 2, 8 or 32 times) resulted in larger initial increases in liking, followed by a reduction in liking, as satiation began to take effect. This finding suggests that repetition allows listeners to engage with musical stimuli of different levels of complexity, providing them with the opportunity to explore new aspects of previously presented music (Margulis, 2014). Where repetition provides the opportunity for familiarization and exploration of a more complex music piece, simpler pieces are more likely to be subject to tedium and boredom, resulting in decreased liking upon less repeated exposures (Stevens & Latimer, 1991). In the current study, 3 exposures of unfamiliar music excerpts (15 seconds in length) are expected to be sufficient to result in increased recognition, as well as increased liking ratings as opposed to only 1 exposure. Taking into account the unfamiliar nature of the excerpts used and listeners’ perceived and preferred levels of complexity (Heyduk, 1975), we do not expect tedium effects at this level of exposure.

The role of extra-musical information on recognition and liking

Music is not meaningless sequences of notes. Listeners often associate pieces of music with a certain scene, occasion or emotion. Is it the case that memory for the associated meaning of a piece of music aids with memory of the actual music? Extrinsic factors, for example, extra-musical information such as titles and programme notes, can influence listeners’ subjective experiences (e.g., enjoyment) of music (Kroger & Margulis, 2017). Not only does extra-musical information have the potential to influence enjoyment, it also assists listeners in overcoming the challenges associated with the unfamiliarity of previously unheard music. Aesthetic responses to artworks suggest that descriptive or elaborative titles help audiences to make sense of what they are seeing, and enable them to enjoy it more (Leder, Carbon, & Ripsas, 2006). Furthermore, textual information can provide a contextual point of reference for the audience, by communicating the expressive intent of the artist by eliciting their intended emotions and meanings (Cupchik, Vartanian, Crawley & Mikulis, 1994).

However, it is unclear whether extra-musical information is sufficient in facilitating later recognition and liking for unfamiliar music. Do they increase audience recognition, and do they increase audience enjoyment? Early studies exploring the associations between music and text (mainly concerning lyrics) have shown that extra-musical associations strengthen recognition, where participants remember songs better in original music and text pairings, than in mismatched trials (Crowder et al., 1990; Serafine et al., 1984). The associative network model of music memory posits that we bind not only the structural properties of a music piece (e.g., rhythm), but its extra-musical associations as well (Stevens & Latimer, 1991). Furthermore, these musical and extra-musical components form an integrated representation in memory, rather than separate components (Crowder et al., 1990). This finding confirms that two elements presented closely together in time become associated in memory, such that memory for one is aided by the presence of another. In terms of memory for songs specifically, both text and song fragments similarly aid recognition (Peretz et al., 2004).

Previous research exploring the impact of programme notes and titles on music enjoyment has yielded mixed results. Extra-musical information has been shown to enhance emotions induced in music listening (Vuoskoski & Eerola, 2015). However, extra-musical information in the form of subtitles can hinder listeners’ perceptions of expressive moments in a performance (Silveira & Diaz, 2012). Findings from Margulis (2010) indicated that for adult non-musicians, presenting programme notes before a performance was detrimental to their enjoyment. Similarly, presenting programme notes to children (who have less cumulative musical experience than adults) before a performance did not impact on their levels of enjoyment, although they aided in the orientation of their attention and consequently, increased their comprehension of the performance (Margulis et al., 2015).

However, programme notes do have the potential to change how audiences listen to a piece of unfamiliar music (Bennett & Ginsborg, 2018). In this study, participants were provided with song titles before the first listening of the performance, followed by oral delivery of brief programme notes after the first performance. This was to ensure that the audience was able to attend to the performances without being distracted by reading printed programme notes. All participants in the study reported that the orally-delivered programme notes impacted on their second listening: this information influenced them to listen to the second performance differently. Despite this, only 39% of listeners reported that the programme notes had a positive impact on their listening experience, in cases where this information allowed them to confirm their initial interpretations. This finding suggests that perhaps programme notes dictate too strong an interpretation for most listeners: this leaves little opportunity for them to develop their own interpretation. Here, expertise may also play a role. Musically experienced listeners were more likely to reject programme notes in favour of their own interpretations, compared to listeners with less experience. Listeners with less musical experience preferred an ontological interpretation (i.e., relating the music piece to the real-world context) over interpretations that either focused on sounds (syntactical) or the meaning behind the relationships between sounds (semantic). This finding suggests that there may be potential benefits for liking and memory, associated with the type of extra-musical information presented with an unfamiliar piece of music.

Purpose

The current study aimed to determine the effect of exposure and type of titling information on non-musicians’ recognition accuracy and liking for unfamiliar musical excerpts. Using a within-subjects design, we presented participants with 15-second musical excerpts which varied in exposure (one or three exposures), and accompanying title (none, descriptive, semantic, affective). Higher recognition rates of excerpts presented more frequently would confirm the mere exposure hypothesis. As non-musicians have demonstrated a tendency to prefer descriptions of the ontological meaning of unfamiliar music on repeated performances, we hypothesized that the type of information presented would affect participants’ learning of new musical excerpts. We predicted that exposure and titles would interact with greater recognition accuracy and liking in increased exposures with titles that describe the underlying emotional elements (affective). By contrast, with limited exposure, music excerpts accompanied by descriptive or semantic titles, may better aid recognition.

Method

Participants

Participants were 23 musically-untrained undergraduate students at Western Sydney University (17 female, 6 male; age M = 22.3 years, SD = 6.1) who received course credit for participating.¹ Participants were recruited from the Research Participation System (SONA) at Western Sydney University, where they voluntarily signed up for the study. All participants confirmed that they had less than 1 year of musical training on any instrument, and met the standard Australian university requirements for English proficiency.

Materials

The musical materials in the present study comprised 64 excerpts of contemporary classical music (classical music composed post-1945). Each excerpt was approximately 15 seconds in duration, and selected under the following criteria: (i) the excerpt had no clear harmonic key,² (ii) the excerpt was taken from a piece of music whose title or accompanying programme notes suggested a scene or emotion. The excerpts also varied according to four different instrument groups to ensure enough distinction between excerpts (solo piano, violin and piano, flute and piano, wind ensemble). Each exposure × title condition contained one excerpt for every instrument grouping.

The authors constructed titles for each musical excerpt according to three categories, where the content matter focused on a specific aspect of the music excerpt: (i) descriptive – emphasizing musical instrumentation, patterns and technical details in the excerpt. This type of title focused on the surface elements of the excerpt such as ascending/descending passages of notes, pitch-range, speed and density of notes, e.g., “A low rumbling melody followed by a slow succession of quiet, soft notes”; (ii) semantic – emphasizing information illustrating either a concrete scenery of the excerpt e.g., “The moon rises, the birds are woken and begin to call” or abstract subject matter, e.g., “A song about purity and innocence”; and iii) affective – emphasizing the emotional content of the song, which was either positive, e.g., “Joyous warmness and acceptance” or negative, e.g., “A sense of fear and loss of control”. All titles used information taken from CD inlays and programme notes provided by the Australian Music Centre.³ A full list of excerpts and corresponding titles is available (see Table S1 in the Supplemental Material Online section).

Equipment

The experiment was conducted in individual sessions using a laptop with Sennheiser HD 280 Professional headphones. The volume was set to the same level for all participants.

Procedure

The experiment comprised two sessions, spaced one day apart. Prior to the experiment, participants were asked to complete a short paper questionnaire providing details of their age, gender, and information regarding previous music experience. This questionnaire allowed for confirmation that all participants received no formal musical training and were eligible to participate in the study.

For each trial, participants heard a 15-second music excerpt followed by the text of one of the titling conditions appearing onscreen for a duration of 5 seconds. The rationale for presenting the text titles after the music excerpts was that participants would be blind to the titling condition whilst listening to the music. Results from a short pilot test⁴ confirmed that presenting titles after the music excerpt did not affect recognition accuracy as compared to presenting titles before the excerpt.

Participants indicated whether they had heard the excerpt previously in the experiment (“Yes/No”). Similar to liking ratings recorded by Szpunar et al. (2004) and Schellenberg et al. (2008), participants then rated how much they liked the excerpt using a 7-point Likert scale (1 = “Strongly Dislike”, 2 = “Dislike”, 3 = “Moderately Dislike”, 4 = “Neutral”, 5 = “Moderately Like”, 6 = “Like” and 7 = “Strongly Like”).

On Day 1, participants heard 32 musical excerpts in a random order. On Day 2, participants heard the 32 “old” and 32 “new” excerpts in a random order. The participants’ task for Days 1 and 2 was identical with the exception that on Day 2, no titles followed after the music excerpts. Presenting the music excerpts without titles on Day 2 ensured that participants’ recognition rates reflected their memory for the music and not for the text titles.

Music excerpts were counterbalanced in terms of appearance in the exposure × title conditions on Day 1, and excerpts appearing as “old” or “new” across Days 1 and 2, across all participants. Each participant rated four excerpts for every exposure × title condition, one for each instrument timbre. There were equal distributions of concrete/abstract descriptions in the semantic title condition, as well as positive and negative emotion descriptions in the affective title condition.

The experimenter did not provide participants with details regarding the purpose or design (i.e., recognition and liking re-tested on Day 2, or that excerpts from Day 1 would be tested on Day 2) of the study to reduce possible demand characteristics.

Data analysis

The corrected recognition score accounting for response bias, or discrimination measure (P_r), was calculated from the Yes/No recognition responses. The discrimination measure is equal to the conditional probability of responding Yes to an “old” item (a “hit”, or correctly recognized excerpt), minus the conditional probability of responding Yes to a “new” item (a “false alarm”, or a falsely recognized excerpt) (P_r = P_H - P_FA). False alarm rates were also calculated across responses for Days 1 and 2, where these values denote the proportion of incorrect “Yes” responses on Day 2 for music excerpts that were not presented on Day 1, i.e., the participants had falsely recognized the excerpts as having been presented in the previous session.

Parametric assumptions checking and repeated measures analysis of variance (ANOVA) were performed on recognition (calculated discrimination measure, P_r) and liking scores using the statistics software package IBM SPSS Statistics 23. Analysis was conducted on recognition and liking scores, collapsed according to the variables of interest: exposure (1 exposure vs. 3 exposures) and title (none, descriptive, semantic, affective).

Normality for recognition and liking data was assessed by calculating the z-scores of skewness and kurtosis (dividing their statistic value with their standard error) and comparing them to the critical value, ±3.29 (Field, 2013). The Shapiro-Wilk test was used to assess normality with the significance value p < .05, due to its conservativeness and sensitivity in detecting significant deviations from normality (Field, 2013). Mauchly’s test was used to assess the assumption of sphericity among recognition and liking scores, for variance of differences between titling conditions by comparing them against the critical p-value (p < .05) (Field, 2013). Normality and sphericity assumptions were met for both discrimination measure scores and liking ratings.

Due to non-violation of normality distributions, parametric t-tests and repeated measures were appropriate for this analysis. Paired-sample t-tests were performed on the false alarm rates for Day 1 and Day 2.

To test the main hypotheses of the study, two separate 2 × 4 repeated measures ANOVAs were conducted on the discrimination measure (P_r) and liking, respectively, to determine the differences and interaction between exposure (1 vs. 3) and title (none, descriptive, semantic and affective). Post-hoc Bonferroni comparisons were used to further analyse interactions and main effects. For the ANOVA conducted on the liking ratings, only ratings for correctly identified excerpts on Day 2 were included.

To further analyse the relationship between liking and recognition, as well as accounting for variance due to individual differences, a generalized linear mixed effects model was used to predict the binary outcome of participants’ recognition (Yes/No) for items that had been presented on Day 1. The model controlled for a random effect of Participant ID, and fixed effects of instrument, first liking, final liking, title and exposure. The variation across all participants was accounted for in the model by treating Participant ID as a random effect. Here, we were not interested in the differences between specific participants, but instead aimed to account for the variation across participants as a whole. Title and exposure were treated as fixed effects as these were the two main factors manipulated in the experiment, and we wanted to quantify the differences in recognition across the different levels of each factor. Instrument was included as a fixed effect as the type of instrumentation in the excerpt may have been a confounding factor for recognition. By treating this as a fixed effect, we looked to investigate the difference in recognition across different categories of instrument. Similarly, the liking rating given by participants on both the first and final presentations of the excerpt may have been confounding factors and so were also treated as fixed effects. Interpretation of these results was then based on the influence of all fixed effects, controlling for random variation across participants (Participant ID). All models were fitted with the lme4 package in R. The final model was compared to a base model containing only the intercept using the ‘anova’ function. Reported estimates for the final model show the influence of a one-unit increase in each fixed effect on the outcome, whilst holding the other effects constant. Each predictor’s p-value was estimated through theoretical tests $(χ^{2} (1))$ of the logarithm of the likelihood ratio between the model with that predictor and the model without that predictor using the ‘drop1’ function. Confidence intervals were calculated with the ‘confint’ function.

Results

Recognition

A paired sample t-test performed on the discrimination measure (P_r) from Day 2 confirmed that participants correctly recognized “old” excerpts significantly better than chance, t(22) = 16.95, p < .001, d = 3.53. A paired sample t-test was also conducted between the false alarm rates for Day 1 and Day 2 to determine whether there were significant differences between incorrectly identifying “old” excerpts for Day 1 and Day 2. False alarm rates were significantly different to each other between sessions, t(22) = -2.59, p = .017, d = -.54, as the rate increased from Day 1 (M = 0.11, SD = 0.14) to Day 2 (M = 0.24, SD = 0.19).

Figure 1 shows the data for mean discrimination measure scores on Day 2 across each exposure and title condition. A 2 × 4 repeated measures ANOVA was conducted to compare the main effects of exposure (1 exposure and 3 exposures) and title (no title, descriptive, semantic and affective) on the discrimination measure scores (P_r) for Day 2. There was a significant main effect of exposure, F(1, 22) = 82.72, p < .01, $η_{p}^{2}$ = .79, on the corrected recognition scores for Day 2. No significant main effect for title was found, and there was no significant interaction between exposure and title.

Liking

To confirm that participants used all response categories in the liking scale, Table 1 summarizes the frequency of ratings for each liking category (out of the total responses across all items and participants), as well as the number of participants who used each category across Day 1 and Day 2.

Table 1.

Frequency of responses and number of participant responses for each liking category collapsed across responses over the 2 sessions.

Liking category	Frequency of responses (out of 2944 total responses)	Number of participants
1 – “Strongly dislike”	161	11 (47.8%)
2 – “Dislike”	223	18 (78.2%)
3 – “Moderately dislike”	486	21 (91.3%)
4 – “Neutral”	757	23 (100%)
5 – “Moderately like”	721	23 (100%)
6 – “Like”	420	23 (100%)
7 – Strongly like”	176	18 (78.2%)

Analyses were conducted on liking ratings for correctly identified “old” excerpts on Day 2, to determine whether exposure and title conditions affected liking on the final exposure. Seven participants were removed from analysis, as they had missing values where they had not correctly identified an “old” excerpt in one or more exposure × title conditions. Figure 2 shows the data for liking ratings on Day 2 of correctly identified “old” items across each exposure and title condition for the remaining participants.

A 2 × 4 repeated measures ANOVA was conducted to compare the main effects of exposure (1 exposure and 3 exposures) and title (no title, descriptive, semantic and affective) on the liking ratings of the remaining 16 participants. A significant interaction effect was found between exposure and title, F(3, 45) = 3.78, p = .17, $η_{p}^{2}$ = .20. This finding indicates that the type of title had different effects on liking ratings depending on the level of exposure.

To test our hypothesis regarding the affective titles, Bonferroni-corrected post-hoc comparisons were conducted on the differences between liking ratings (between exposure levels), between the affective title and other title conditions (descriptive, semantic and no title) It was found that the difference between liking ratings at the two exposure levels for affective titles (M = 0.97, SD = 1.58) was significantly greater, t(22) = 2.86, p = .009, d = .60, than the difference between liking ratings at the two exposure levels for semantic titles (M = 0.08, SD = 0.15). No significant differences were found between liking ratings for no label, affective and descriptive titles.

The relationship between recognition and liking

A generalized linear mixed-effects model was used to assess how well recognition of music excerpts on Day 2 related to participants’ liking either on first (Day 1) or last (Day 2) exposure. Only items that had been presented on Day 1 were included in the model. The mixed effects model allowed assessment of all fixed effects controlling for variation across participants (Participant ID). Table 2 shows the fully specified model, which is significant $(χ^{2} (9) = 100.3, p < . 001)$ in comparison to a base model containing only the intercept. The coefficients for exposure and final liking were both significant (p < .001). Essentially, for every one unit increase in exposure (i.e., the difference between the two levels of exposure in the study), the log odds of accurate recognition increased by 1.64. This means that accurate recognition was 5.16 times as likely when participants were exposed to the excerpts at the higher level of exposure as compared to the lower level of exposure. For every one unit increase in the final liking rating, the log odds of accurate recognition increased by 0.28. This means for every step increase in the liking rating on the final exposure, accurate recognition was 1.32 times as likely.

Table 2.

Generalized linear mixed-effects model predicting Recognition (Yes/No) on Day 2 from fixed effects of title, exposure, instrument, liking on first exposure and liking on final exposure. The model includes a random effect to control for individual variation across participants.

Fixed effect	Estimate	SE	p	Confidence Intervals
				[2.5%]	[97.5%]
Intercept	−2.38	0.53	4.22e–3**	−3.46	−1.35
Title: Descriptive	−0.02	0.26	0.93	−0.53	0.48
Title: Semantic	0.11	0.26	0.66	−0.39	0.62
Title: Affective	−0.01	0.25	0.69	−0.60	0.40
Exposure	1.64	0.20	< 2.2e–16***	1.27	2.03
Instrument: Flute	0.07	0.26	0.80	−0.45	0.58
Instrument: Piano	−0.11	0.25	0.65	−0.61	0.38
Instrument: Violin	−0.24	0.25	0.35	−0.74	0.26
First Liking	−0.02	0.08	0.80	−0.17	0.13
Final Liking	0.28	0.09	8.61e–4***	0.11	0.45
Log Likelihood	−387.4
Num. Observations	736
Num. group: ID	23
Var: ID (Intercept)	0.71

p < .01. ***p < .001.

Discussion

The present study aimed to determine the contribution of exposure and titling information to non-musicians’ recognition and liking of unfamiliar music excerpts. The findings indicate first that non-musicians can remember complex excerpts of classical contemporary music significantly better than chance, despite being completely unfamiliar with both the stimulus items, and their underlying harmonic structures. Our results confirm the hypothesis that repeated exposures will result in increased recognition. This finding indicates that three repetitions were sufficient for overcoming the initial challenges for non-musicians hearing an unfamiliar, complex music excerpt. This is in line with findings from Szpunar et al. (2004) which suggests that repetition plays a vital role in establishing familiarity for complex musical pieces.

We did not find support for the hypothesis that titles with descriptive, semantic and affective content presented after the musical excerpt affected non-musicians’ recognition accuracy. There was neither any benefit or impairment of memory from any of the title types in comparison to no titles being present. Although the lack of differences found between title categories could be a result of some overlapping elements in the titles (see Limitations), this does support Deffler and Halpern’s (2011) findings that associative facts provided no additional benefit on young adults’ memory for unfamiliar tunes.

The findings did not support the hypothesis that exposure and titles will interact with greater accuracy in recognition. However, an interaction was found for liking ratings where exposure appears to have a stronger effect on liking ratings for music excerpts presented with affective titles describing the emotional aspects of the excerpts, rather than semantic titles which focused on the underlying scene or context. Essentially, there was a larger increase in liking for music excerpts with affective titles over increased exposures, compared with semantic titles. Although the mean liking ratings were close to the “Neutral” rating (rated as 4) across most exposure × title conditions, the mean liking for the affective titles for one exposure was below the neutral rating (i.e., participants mainly disliked the excerpt), whereas for three exposures, the mean liking for the same title condition was above the neutral rating (i.e., participants mainly liked the excerpt). A possible explanation for this difference between semantic and affective titles with increasing exposure is that the semantic titles conveying a particular scene or context may have too strongly instilled one interpretation of the piece. Over repeated listening, these potentially hindered listeners’ abilities to develop their own interpretation. Affective titles on the other hand may have allowed audiences to formulate their own understanding. This is in line with previous research which has shown that non-musicians have a tendency to focus on broader meanings of music pieces, and align their initial interpretations with information provided by programme notes (Bennett & Ginsborg, 2018).

Finally, accounting for individual differences, participants’ liking of the music excerpt at the final exposure appears to be related to their recognition accuracy of that excerpt. This is in line with findings from Szpunar et al. (2004) in that it suggests that participants like pieces they remember. However, the lack of predictive power for the initial liking rating on the final recognition accuracy does not suggest that participants necessarily remember pieces of music they like (at least on first presentation). An explanation for this could be the majority of relatively neutral responses given to these musical excerpts, i.e., on first exposure, there was not a large amount of difference in liking ratings across musical excerpts. We cannot discount demand characteristics as a potential contributor to the predictive power of the final liking rating, as participants may have been aware from the procedure of the study (particularly by Day 2 testing) that we were interested in this relationship. Further investigation into how liking for unfamiliar music changes over time (and higher numbers of exposures) would better determine this relationship and elucidate at what timepoint liking begins to predict recognition.

Limitations

A limitation in the current study was the small sample size. A priori power analysis indicated that 19 participants were sufficient for a large size effect. This minimum number of participants was met for the strong effect elicited by the exposure rate. However, the analysis for liking ratings discarded participants who failed to accurately recognize excerpts in one or more exposure × title conditions. With 16 remaining participants for analysis where liking effects tend to be subtler, this was potentially underpowered. Despite this, a significant interaction of exposure × title was found in the liking ratings for these 16 participants, indicating that the type of title information may influence liking at different exposure levels.

The inclusion of the “Neutral” response in the Likert scale as a measure for liking ratings could also have contributed to the small differences seen between title and exposure categories. Despite the intention of the study providing participants with the option to indicate that they felt neutral or indifferent towards an excerpt, it may be possible that participants selected the “Neutral” option as a result of not knowing whether they enjoyed the excerpt. The current study lacked a measure to capture “I don’t know” responses, therefore limiting appropriate interpretation of “Neutral” ratings where it is not possible to discern from the findings whether participants felt truly indifferent about the excerpts or whether they were unsure.

The lack of effects found for title information concerning recognition accuracy could be a result of the small overlaps in title information, e.g., descriptive titles containing some semantic elements while describing the musical features of the excerpt. Although this may explain the lack of difference between certain title categories in terms of recognition, the significant interaction between exposure and title for the liking ratings indicates that the semantic and affective categories elicited different responses at different exposure levels. Further studies using heavily reduced title information to ensure distinctness of separate categories would assist in confirming whether this is the case. Another explanation for the lack of effect of title on recognition accuracy could be concerning the timing of presentation of the titles (i.e., after the music excerpt). Although the rationale for presenting titles after the music excerpts was to encourage participants to focus on the music excerpts and not the text, and maintain blinding of the participants to the condition until after they heard the music, this in tandem with low sample size for the liking ratings analysis in particular may have tempered potential effects. This also serves as a limitation for generalizability to real-world performances, where audiences typically read titles and programme notes before a performance, rather than after.

Although this study has demonstrated effects on recognition accuracy and liking by varying the exposure and type of titling information, these results pertain to 15-second excerpts of music limited to one specific genre (contemporary classical music). While this allows us to comment on effects for non-musicians listening to completely unfamiliar music, it is unknown how this would generalize to longer musical excerpts, and variation in musical genre.

Future directions

To further determine the effect of varying exposure and title information on non-musicians’ ability to recognize unfamiliar music (and how their liking of these pieces develop), future research could utilize a wider variety of musical excerpts and length of these excerpts. In determining the extent of familiarity with the underlying structures of a piece, i.e., classifying the strength of particular harmonic keys and their relationship to those found in Western popular music, studies could explore the contribution of participants’ familiarity with the underlying elements of the music rather than simply controlling familiarity for the item by manipulating the number of exposures.

As mentioned in the limitations, further exploration of the effect of titling information is required, particularly where categories can be made more distinct from one another. Our results suggest that affective titles with emotional content may allow more potential for increases in liking over several exposures than titles focusing on underlying scenes or context. The mean liking ratings across all conditions were far from satiation: higher numbers of exposures for this type of complex, unfamiliar music may further clarify the extent of this effect.

Conclusions

Non-musicians can better recognize 15-second excerpts of contemporary classical music with repeated exposure. Additional textual information presented after listening to the excerpt does not appear to aid this recognition accuracy. Participants’ liking for a musical excerpt, however, varies depending on both the type of textual information provided, and the number of exposures: titles with emotional content resulted in significantly larger differences in liking across exposure levels than titles containing information about underlying scenes or context. Further investigation may illuminate whether titles with affective content can further increase enjoyment of initially unfamiliar excerpts with increased exposure.

Supplemental Material

Supplementary_Material – Supplemental material for Non-musicians recognize unfamiliar contemporary classical music excerpts with increasing repetition

Supplemental material, Supplementary_Material for Non-musicians recognize unfamiliar contemporary classical music excerpts with increasing repetition by Verena S. Wu, Jennifer MacRitchie and Catherine J. Stevens in Musicae Scientiae

Footnotes

Acknowledgements

The authors would like to thank the Australian Music Centre for allowing access to their library of music recordings and Dr. Steffen Herff for assistance with analyses.

Declaration of Conflicting Interests

There are no conflicting interests for any of the authors.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Supplemental Material

Tables and figures/audio files with the index “S” are available as Supplemental Online Material, which can be found attached to the online version of this article at . Click on the hyperlink “Supplemental material“ to view the additional files.

Notes

ORCID iDs

Verena S. Wu

Jennifer MacRitchie

References

Bennett

Ginsborg

(2018). Audience reactions to the program notes of unfamiliar music. Psychology of Music, 46(4), 588–605.

Berlyne

D. E.

(1970). Novelty, complexity, and hedonic value. Perception & Psychophysics, 8(5), 279–286.

Bornstein

R. F.

(1989). Exposure and affect: Overview and meta-analysis of research, 1968–1987. Psychological Bulletin, 106(2), 265–289. doi:10.1037/0033-2909.106.2.265

Cupchik

G. C.

Vartanian

Crawley

Mikulis

D. J.

(2009). Viewing artworks: Contributions of cognitive control and perceptual facilitation to aesthetic experience. Brain and Cognition, 70(1), 84–91.

Crowder

R. G.

Serafine

M. L.

Repp

(1990). Physical interaction and association by contiguity in memory for the words and melodies of songs. Memory & Cognition, 18(5), 469–476.

Deffler

S. A

Halpern

A. R.

(2011). Contextual information and memory for unfamiliar tunes in older and younger adults. Psychology and Aging, 26(4), 900–904.

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175–191.

Field

(2013). Discovering statistics using IBM SPSS Statistics: London, UK: Sage.

Hargreaves

D. J.

(1984). The effects of repetition on liking for music. Journal of Research in Music Education, 32(1), 35–47.

10.

Heyduk

R. G.

(1975). Rated preference for musical compositions as it relates tocomplexity and exposure frequency. Perception & Psychophysics, 17(1), 84–90.

11.

Juslin

P. N.

(2003). Five facets of musical expression: A psychologist’s perspective on music performance. Psychology of Music, 31(3), 273–302.

12.

Kroger

Margulis

E. H.

(2016). “But they told me it was professional”: Extrinsic factors in the evaluation of musical performance. Psychology of Music, 45(1), 49–64.

13.

Leder

Carbon

C. C.

Ripsas

A. L.

(2006). Entitling art: Influence of title information on understanding and appreciation of paintings. Acta Psychologica, 121(2), 176–198.

14.

Madison

Schiölde

(2017). Repeated listening increases the liking for music regardless of its complexity: Implications for the appreciation and aesthetics of music. Frontiers in Neuroscience, 11(147). doi:10.3389/fnins.2017.00147

15.

Margulis

E. H.

(2010). When program notes don’t help: Music descriptions and enjoyment. Psychology of Music, 38(3), 285–302.

16.

Margulis

E. H.

(2014). On repeat: How music plays the mind. Oxford, UK: Oxford University Press.

17.

Margulis

E. H.

Kisida

Greene

J. P.

(2015). A knowing ear: The effect of explicit information on children’s experience of a musical performance. Psychology of Music, 43(4), 596–605.

18.

Montoya

R. M.

Horton

R. S.

Vevea

J. L.

Citkowicz

Lauber

E. A.

(2017). A re-examination of the mere exposure effect: The influence of repeated exposure on recognition, familiarity, and liking. Psychology Bulletin, 143(5), 459–498.

19.

North

A. C.

Hargreaves

D. J.

(1995). Subjective complexity, familiarity, and liking for popular music. Psychomusicology, 14, 77–93.

20.

Peretz

Radeau

Arguin

(2004). Two-way interactions between music and language: Evidence from priming recognition of tune and lyrics in familiar songs. Memory & Cognition, 32(1), 142–152.

21.

Schellenberg

E. G.

Peretz

Vieillard

(2008). Liking for happy-and sad-sounding music: Effects of exposure. Cognition & Emotion, 22(2), 218–237.

22.

Serafine

M. L.

Crowder

R. G.

Repp

B. H.

(1984). Integration of melody and text in memory for songs. Cognition, 16(3), 285–303.

23.

Silveira

J. M.

Diaz

F. M.

(2012). The effect of subtitles on listeners’ perceptions of expressivity. Psychology of Music, 42(2), 233–250.

24.

Stalinski

Schellenberg

(2013). Listeners remember music they like. Journal of Experimental Psychology: Human Learning and Memory, 39(3), 700–716.

25.

Stevens

Latimer

(1991). Judgments of complexity and pleasingness in music: The effect of structure, repetition, and training. Australian Journal of Psychology, 43(1), 17–22.

26.

Szpunar

K. K.

Schellenberg

E. G.

Pliner

(2004). Liking and memory for musical stimuli as a function of exposure. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30(2), 370–381.

27.

Vuoskoski

J. K.

Eerola

(2015). Extramusical information contributes to emotions induced by music. Psychology of Music, 43(2), 262–274.

28.

Winkielman

Schwarz

Fazendeiro

Reber

(2003). The hedonic marking of processing fluency: Implications for evaluative judgment. In Musch

Klauer

K. C.

(Eds.), The psychology of evaluation: Affective processes in cognition and emotion ( pp. 189–217). Mahwah, NJ: Lawrence Erlbaum.

29.

Zajonc

R. B.

(1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology, 9(2p2), 1–27.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.13 MB