Abstract
In this study we investigated the effect of musical training and emotional content of stimuli on time estimation. Eighty students aged 12 to 15 years were equally divided into two main groups (musicians, non-musicians) based on whether they had formal music education or not. The participants took part in a music listening task and were asked to estimate the duration of a happy or a sad song. The results showed that: musical training affects temporal perception by allowing individuals to make more accurate estimations of time durations; the emotional content of the song affects time estimation, especially in the case of non-musicians; non-musicians overestimate the duration of sad songs while they underestimate the duration in the case of happy songs. These results emphasize the close relationship between time estimation, musical training, and emotion.
Introduction
Subjective time refers to the personal experience of time in contrast to the objective, which can be accurately measured by clocks. The subjective duration of an event can be overestimated or underestimated. Various factors have been shown to affect subjective time: the individual’s normal or pathological state (Wittman, Vollmer, Schweiger, & Hiddermann, 2006), cognitive factors (Ulbrich, Churan, Fink, & Wittmann, 2009), language (Casasanto et al., 2004), age (Gunstad, Cohen, Paul, Luyster, & Gordon, 2006), and the type of stimuli (Klapproth, 2001).
A factor that significantly affects the estimation of a stimulus’s duration is the mood of the individual. Numerous studies have shown that the duration of stimuli or events that evoke positive feelings are underestimated, while those that evoke negative feelings are overestimated (Droit-Volet & Meck, 2007; Flaherty, 2001; Noulhiane, Mella, Samson, & Ragot, 2007; Wittmann & van Wassenhove, 2009). Subjectivization of time (i.e., lengthening or shortening of time duration) is also evident on the events that we experience when we watch a film (De Wied, Tan, & Frijda, 1992; Samartzi, 2003). However, things are less clear when music stimuli are used. Previous studies investigating the role of affect of music in time estimation have come up with unclear and sometimes contradictory results (Angrilli, Cherubini, Pavese, & Manfredini, 2005; Hul, Dube, & Chebat, 1997; Kellaris & Kent, 1992).
Emotion is also related to music and this relationship depends on the structural features of the song (mode, melody, tempo, etc.) and the listener features (musical expertise, personality, etc.) as well (Dalla Bella, Peretz, Rousseau, & Gosselin, 2001; Khalfa, Roy, Rainville, Dalla Bella, & Peretz, 2008; Krumhansl, 1997; Scherer & Zentner, 2001). Many studies have shown (Hunter, Schellenberg, & Schimmack, 2008; Peretz, Gagnon & Bouchard, 1998) that the emotional valence of music depends on the mode (major or minor) and the tempo (slow or fast). Major mode and fast tempo define happy songs, whereas minor mode and slow tempo define sad songs. It’s not yet known whether tempo or mode is more important in perceiving emotion in music and changing the mood of the individual (Husain, Thompson, & Schellenberg, 2002). Even though the relationship between emotion and music is well established, theorists still disagree about its exact nature. Two positions are the most prevalent: emotivism and cognitivism. Emotivists support that music can elicit real emotional (non-aesthetic) responses in listeners (Krumhansl, 1997). On the other hand, cognitivists argue that music simply represents emotions (Scherer & Zentner, 2001). Evidence from electrophysiological and neuroimaging studies seems to offer support for emotivism. In particular, different emotions expressed through music have been associated with particular patterns of increases in skin conductance (Rickard, 2004), and changes in heart rate, temperature, and blood pressure (Krumhansl, 1997). This is the position that we have adopted in this paper.
Musical training may also affect time estimation performance. Studies have shown significant improvements in relative timing (temporal continuity, underlying beat, metrical structure) as musicians practice and become more skilled (Drake & Palmer, 2000). The considerable difference between musicians’ and non-musicians’ time estimation abilities becomes more evident in a number of studies that examine time estimation through time production tasks (Repp, 2005, 2010; Repp & Doggett, 2007; Snyder, Hannon, Large, & Christiansen, 2006) in which musicians outperform non-musicians in synchronizing with a beat and are more sensitive to time deviations. Moreover, superior temporal acuity was shown for musicians compared to non-musicians for auditory fusion, rhythm perception, and for temporal discrimination tasks (Rammsayer & Altenmüller, 2006).
The findings reported above show that: (1) emotional content of information influences time perception and estimation; (2) musical training influences time perception and estimation; and (3) music influences and changes our emotions. Apart from the knowledge about dual relations between these parameters, an interesting question is to identify the relationships that possibly exist among the three of them (for a literature review see Samartzi & Panagiotidi, 2011). The aim of this study was to investigate the effect of musical training and emotionality on time estimation. In addition to that, we tested this hypothesis on adolescents. The majority of studies on music use professional musicians as participants (Rammsayer & Altenmüller, 2006; Repp, 2010) and compare their performance to that of non-experts. In our study we wanted to examine whether fewer years of musical training could produce significant behavioural effects. Moreover, research so far was more focused on time estimation of shorter stimuli (in the range of milliseconds–seconds) (Grondin, 2008). In this study, we used as stimuli songs, which had longer duration (in the range of minutes).
Specifically, we hypothesized that: (1) the emotional content of the song will influence time estimation; the duration of the “sad” song will be overestimated while the duration of the happy song will be underestimated; (2) the features of the listeners are expected to mediate the emotional content effects supposed by our first hypothesis; musicians will be more accurate in time estimation than non-musicians (Repp, 2005; Repp & Doggett, 2007).
Method
Participants
Eighty public high school students, 12 to 15 years old (average age: 13.46; boys = 37, girls = 43) in a middle-class residential region of Athens, took part in this research. Participants were equally divided into two main groups of 40 subjects each, based on whether they had formal musical training (minimum 2 years) (M = 2.40, SD = 2.559, boys = 19, girls = 21), or not (boys = 18, girls = 22). We chose 2 years as the amount of minimum training based on neuroimaging studies that suggest that observable structural differences are found in the brains of children after only a few years of musical training (Hyde et al., 2009; Schlaug, Norton, Overy, & Winner, 2005). The participants from each main group were randomly allocated to the happy or sad song condition.
Materials
In our study we used Mozart’s Eine kleine nachtmusik (1st movement) as a happy (major mode and fast tempo) song and Albinoni’s Adagio as a sad (minor mode and slow tempo) song. Both songs were instrumental (no vocals) to avoid possible effects of the content of the lyrics. These two pieces are ecologically valid stimuli (Samartzi, Kazi, & Koustoumbardis, in press; Tobin, Bisson, & Grondin, 2010) and were used in previous studies (Thompon, Schellenberg, & Husain, 2001) in their capacity to evoke a sense of happiness or sadness, respectively. Both songs were digitally rerecorded from compact discs onto the hard drive of each computer with a minimum loss of sound quality (mp3 files; 320 kbps, 44 Hz) and were modified in order to have an equal duration of around 6 minutes (~5:58). Furthermore, they were normalized so that their maximum amplitude was identical in an attempt to minimize differences in perceived loudness. No change was made to the original tempo or mode of the songs. Both songs ended with a 3 second fade out and their playback was not cut abruptly. Changing the original duration of each song was also a way to control for possible effects of familiarity with the stimuli.
Procedure
Each of the two groups was divided randomly into two subgroups of 20 participants according to the song they listened to (happy versus sad). The experiment took place in a silent computer room, where every participant had access to a PC and a pair of headphones. The paradigm we used was retrospective and the students did not know in advance that they would perform a time estimation task. They were instructed to listen to the song carefully. The song was played on Microsoft Media Player, which was modified so that no time markers were available to the participants. Furthermore, during the experiment the participants could not see a clock and had no access to cell phones and watches. Immediately after the end of the song, questionnaires were handed out and the participants were asked to fill out personal information (age, sex), and to estimate the duration of the song they listened to. They could choose a number, which expressed time in minutes, from 1 to 13. Moreover, all participants were asked to fill in a post-mood questionnaire.
Results
An analysis of covariance (ANCOVA) (between-subjects factor: group (musicians, non-musicians), song (happy, sad); covariate: gender) was conducted and revealed main effects of the song on time perception (see Figure 1 and Table 1), F(1,75) = 23.40, p < .05, η 2 = .235, gender, F(1,75) = 13.45, p < .05, η 2 = .152, and no main effects of group, F(1,75) = 2.00, p = .161, η 2 = .026. Finally, there was an interaction between song and group, F(1,75) = 6.343, p < .05, η 2 = .078.

Error plot of the means of the time estimates in both groups.
Means of time estimates.
We investigated this interaction by simple effects tests which showed that the affect of the song influences time estimation when the participants have no musical training, F(1,72) = 25.80, p < .05. However, in the case of musicians the effect was not significant, F(1,72) = 2.010, p = .152. Simple tests also revealed that musical training led to better time estimations in the case of the happy song, F(1,72) = 19.74, p < .05, but not it in the case of the sad song, F(1,72) = .405, p = 0.527.
Finally, a t-test revealed no significant differences on the post-mood questionnaire ratings between musicians and non-musicians, t(78) = .675, p = .75.
Discussion
Previous research has pointed out the relationships among (1) emotion and time estimation (Chambon, Gil, Niedenthal, & Droit-Volet, 2005); (2) time estimation and musical experience (Repp & Doggett, 2007); and (3) music and emotion (Scherer & Zentner, 2001). In this study we investigated the relationships among music, emotion, and time.
The present study showed that musical training affects significantly the ability to accurately estimate time by reducing the “subjectivity” about time and allowing people with music training to make more accurate estimations of songs’ duration (particularly of happy songs). Overall, we found that emotion has an effect on time perception; the duration of the happy song was underestimated, whereas the sad song’s duration was overestimated. This is consistent with our first hypothesis. More specifically, non-musicians underestimate the duration of the happy song and overestimate the duration of the sad song. Our results are consistent with findings from studies investigating time estimation of real-life experienced events (Chambon et al., 2005) or by watching events occurring in a movie (De Wied et al., 1992; Newtson & Enquist, 1976) but contradicts those observed in the case of duration estimation of events described in texts where subjectivization does not matter (Samartzi et al., in press). These findings indicate that the emotional content of stimuli can influence the time estimation under particular conditions, such as the absence of musical experience.
The results obtained from the post-mood questionnaires confirmed that the stimuli we chose successfully induced happy and sad moods in the participants in both groups. However, musical training seems to reduce or eliminate the effect of emotion. In particular, in our study the affect of the song does not seem to have any effects on the estimations of musically trained individuals. Musical training led to better time estimations in the case of the happy song but not it in the case of the sad song. This finding partly supports our second hypothesis. This is consistent with previous studies (Repp, 2005; Repp & Doggett, 2007) that have shown that musicians are better at making temporal judgments than non-musicians. Interestingly, a minimum of 2 years of musical training seems to affect time estimation abilities. This effect was particularly significant in the case of the happy song. The differences between the happy and the sad song could possibly be attributed to the intensity of the emotional response. Previous studies suggest that slow tempo evokes stronger emotional responses (Oakes, 2003).
We suggest that the difference between the performance of the musicians and the non-musicians observed in this particular time estimation task could reflect differences in the way both groups process musical stimuli. More specifically, studies have showed that individuals with musical training compared to individuals without training activate different brain areas when they are listening to music (Seung, Kyong, Woo, Lee & Lee, 2005). The way that musicians process music stimuli seems to make them less susceptible to the effects of emotionality. This finding could be consistent with recent findings according to which, musicians/artists can modulate specific brain regions involved with emotion and executive function (Heilman, Nadeau & Beversdorf, 2003).
Another possible explanation for our findings could reflect differences in the listening modes of musicians and non-musicians. Evidence from studies on healthy volunteers and patients suggests the existence of more than one listening modes in music. More specifically, there seems to be dissociation between processing the structural features of music and its affect. One particular case study highlights this. I. R., a patient who after sustaining brain damage was left severely impaired in processing music (Peretz et al., 1998), was still able to respond emotionally to music. Peretz and her colleagues investigated the origin of this neuropsychological dissociation and concluded that I. R. could perform close to normal in tasks that the change affected the mode and the tempo of the excerpt. Brattico, Jacobsen, De Baene, Glerean, & Tervaniemi (2010) using EEG found that different listening modes (cognitive versus affective) elicited different brain responses. The reduced effects of the emotion on the time estimation judgments of musicians could possibly suggest that they spontaneously engage in a different listening mode. This could be the result of their training and their familiarity with the structural features of music. It could also possibly explain their less accurate performance in the sad song condition. In particular, previous studies suggest that slow tempo leads to a greater degree of inaccuracy in time estimation in no specific direction overestimation/underestimation) compared to faster tempo (North, Hargreaves, & Heath, 1998). This is similar to the time estimates provided by the musicians in our study.
The gender of the participants was another factor that influenced their temporal judgements in both groups. More specifically, women in both groups tended to overestimate the duration of both songs. This is a common finding in previous studies that have reported significant sex differences (Espinosa-Fernández, Miró, Cano, & Buela-Casal, 2003; Hancock & Rausch, 2010; but also see Gilliland & Humphreys, 1943). The majority of those investigations reported that women overestimate time intervals relative to males. Future studies could attempt to examine whether extensive musical training can mediate this effect.
Retrospective timing paradigms are usually neglected by time perception psychologists (Grondin, 2008). This is one of the reasons that led us to use a retrospective paradigm. Such a paradigm, though, has its limitations, as it does not allow the participants to make numerous time estimates. The choice of two songs as stimuli increased the ecological validity of our study (Tobin et al., 2010). However, this approach did not allow us to control for other musical variables that were different between the two selected pieces. For example, previous research has found that apart from differences in mode and tempo, sad songs are often more complex than songs with happy emotional valence (Balkwill, Thompson, & Matsunaga, 2004). Complexity in music has been found to affect time estimation. In particular, North et al. (1998) suggested that the more complex the music was the more inaccurate were the time estimates. Moreover, the paradigm we used was retrospective. Recent studies suggest that retrospective and prospective timing depend on difference brain and cognitive mechanisms (Zakay & Block, 2004, but also see Brown & Stubbs, 1992). These findings might have implications for the generalizability of the results of the present study.
A limitation of our study was the use of a 1–13 scale. It has been shown that participants tend to choose values near the mean of a range of numbers (Duffy, Huttenlocher, Hedges, & Crawford, 2010). Even though, this cannot fully explain the findings of our study, future studies could repeat the experiment using an open question instead of a scale. In addition to that, follow-up studies could control for familiarity with the stimuli. Even though previous studies have shown no significant effects of familiarity on time estimation (Schiffman & Bobko, 1977), we cannot exclude the possibility that musicians’ performance was better in our task because of their familiarity with the stimuli that were used.
The above findings could be additionally explored in future research by using a bigger number of musical pieces in order to specify the role of tempo and mode in perceiving emotion in music. Moreover, future studies could possibly take into account the effects of familiarity with the stimuli and arousal that have also been shown to influence duration judgements but were not controlled in detail in our research. This could be achieved by using newly composed songs and manipulating tempo and mode to create the desired emotional valence. Arousal could be investigated using electrophysiological recordings to determine whether musicians and non-musicians react differently to musical stimuli. Also, it would be interesting to investigate whether different levels of musical training lead to differences in time estimation. A future study could see whether overexposure to music and not musical training alone could possibly affect time estimation. The results from our study shed light on the relationship between musical training and time estimation and can be applied in education and other areas of everyday life. For instance, this relation may support possible tools in treatments aiming at aiding time perception, which seems to be compromised in many developmental and neurological disorders (Himpel et al., 2009; Lee et al., 2009; Nicolson, Fawcett & Dean, 1995; Overy, 2000).
Footnotes
Acknowledgements
We thank Argiro Vatakis for her suggestions and comments on an early draft.
