Complexity in Musical Rhythms: How Do We Like
It?
Iza Korsmit
Date: August 29th, 2017
Student ID: 6048994
Supervisor: Dr. Rebecca Schaefer, Leiden University
Co-Assessor: Dr. Makiko Sadakata, University of Amsterdam
Program: Msc Brain and Cognitive Sciences, Cognitive Science track, University of
Content
Abstract 2
1. Introduction 3
2. Musical Emotion and Pleasure 4
2.1 Emotion Mechanisms 4
2.2 Musical and Everyday Emotion: Different or Same? 5 2.3 Musical Emotion and Aesthetic Pleasure 6 2.4 Musical Pleasure and Reward 7
3. Focus on Single Mechanism: Prediction 8 3.1 The Predictive Brain 8 3.2 Musical Expectation 10 3.3 Error Minimization 11
4. Violation of Expectations: Do we Like it? 11 4.1 Rhythmic Complexity 12 4.2 Affect and Enjoyment in Response to Complexity 14
5. Entrainment: Moving Along with the Rhythm 18 5.1 Rhythm Prediction and Motor Processing 19 5.2 Entrainment Theories & Pleasure 20
6. Integration and Novel Insights 22
6.1 So Far: Predictive Processing in Music 22 6.2 Affective Monitoring Hypothesis 23 6.3 Music as a Supranormal Playground 24
7. Conclusions and Future Research 25
Abstract
This review aims to characterize the relationship between rhythmic complexity and musical enjoyment. Research on affective processing in general and in music is described, including its relation to pleasure and preferences. Expectation is shown to be a consistent mechanism relevant to musical emotion and reward. Expectation in rhythmic complexity has been implemented in various ways, partly due to the ambiguous definition of rhythmic complexity. This has led to some inconsistent findings on the effect of complexity. Expectation, or prediction of what is to come, is also recently considered to be an important factor in general cognitive processing (i.e., the predictive brain). Part of predictive processing theories is the hypothesis that zero error is the end goal of the brain. Some research indicates, however, that a certain optimal amount of prediction error is preferred in music, especially when it comes to rhythmic prediction error (i.e., rhythmic complexity). This is also shown in research on the interaction between music and motor processing; an optimal amount of error, in the form of syncopation, causes a greater urge to move along with the music (i.e., sense of groove). The affective error-monitoring hypothesis and viewing music as a supranormal playground may explain this difference between findings in music cognition and more general cognitive processing. The affective monitoring hypothesis posits that resolved conflict (also: optimal prediction error) elicits positive affect, and unresolved conflict (also: too much error) elicits negative affect. This review hypothesizes that this may be the brain’s way of keeping its models updated and adaptive through play. Further research is needed to test these ideas, while paying attention to clarity in rhythmic complexity measures, affect or preference measurements, and comparisons between music and general cognition.
1. Introduction
Music is a cultural phenomenon that is enjoyed by people around the globe. Musicality, lying at the basis of creating and enjoying music, is a cognitive trait that is possessed by all humans, regardless of whether they are able to sing in tune or not (Honing, 2009). Although the exact function of music, musicality, and its evolutionary origins are subject of ongoing debate, enjoyment and influencing one’s mood appears to be people’s primary motivation to listen to music (Swaminathan & Schellenberg, 2015). Studying the relationship between music and emotion can even be traced all the way back to the Ancient times, where, e.g., Aristotle theorized about the effect of different musical modes on our emotions (Politics, book VIII; Zentner, Grandjean, & Scherer, 2008). Empirical studies have found that music can be used in the treatment of affective disorders (Gold, Voracek, & Wigram, 2004), can be used as a mood manipulator when studying consumer behavior (Bruner, 1990), and that even infants of the age of 4 months have emotional reactions to music (e.g., Zentner & Kagan, 1998). Altogether, this indicates that understanding musical emotion is an important aspect in research on the function and origins of music and musicality.
The aim of this review is to describe the effect of specific musical characteristics on the enjoyment of music. From music creation perspectives, this knowledge can provide guidelines about what music should sound like, if your goal is to be liked by a large audience. From a clinical perspective, this knowledge can help in finding the right music for each patient, to optimize the effects of music therapy. From a scientific perspective, research on musical preferences and enjoyment can contribute to our understanding of music’s importance in daily life and human evolution, and perhaps also teach us more about human cognition in general.
Instead of discussing music as a whole, this review will discuss, more specifically, rhythm, because it is a core feature of musicality and a prominent feature in musical traditions around the world (Savage, Brown, Sakai, & Currie, 2015). Rhythm has the ability to move us, not only emotionally, but also physically. Additionally, with regards to enjoyment, rhythm has been relatively studied less than harmonic aspects of music. Before we continue, it is important to have a clear understanding of the different definitions concerning musical rhythm, because the exact differentiation between rhythm, meter, pulse, and beat is not always clear (Vuust & Witek, 2014). This review considers them according to the following definitions. The rhythm is the pattern of different events and different inter-onset-intervals (IOI). It is the most basic and non-ambiguous concept. Beat is a perceptual concept, also described as the sense of pulse, with stronger and weaker beats at different points in the meter. The meter is abstracted from the rhythm as a hierarchical (mental) representation with multiple levels of isochronous intervals of two and three (in Western music; Lerdahl & Jackendoff, 1983, London, 2012, as cited by Vuust & Witek, 2014).
An important comparison throughout this review is between the mechanisms that are studied in music specifically, and human cognition in general. Both fields of research
have independently studied similar mechanisms (e.g., emotion, predictive, and motor processing), but could also benefit from each other’s knowledge. Firstly, the mechanisms through which emotions are induced and how this relates to pleasure are discussed in the first section. The section will give a broad overview of which factors are important in enjoying music, and whether music is special compared to everyday-life emotion mechanisms. In the second section of the review, the focus is narrowed down to one specific mechanism; prediction. This is relevant in different aspects of human cognition, music cognition, but also especially in rhythm processing, because the meter is thought to provide a predictive framework. In the third section, we will look at the relationship between complexity, or degree of predictability, and affect or preferences. Are they linearly related, or is the relation more complex than that? More related to rhythm specifically, entrainment and embodiment is discussed in the fourth section of the review. Being able to move along with the music, requires some predictability of the music, and hence can also influence the liking for the music. Finally, the discussed literature will be integrated and provided with some alternative perspectives on how to explain the influence of (rhythmic) complexity on the enjoyment of the music, with suggested directions for future research.
2. Musical Emotion and Pleasure
In this section, the mechanisms through which music causes emotions and pleasure will be discussed. Logically, music that leads to positive emotions will be enjoyed, but research shows that the relationship between emotion and pleasure is not as straightforward. Firstly, emotion induction mechanisms and emotion measurements will be discussed in a more general context, not specifically about music. Then the question is asked; is music special in this respect? And how do the emotions that are induced in the listener, relate to which music induced pleasure in its listener?
2.1 Emotion Mechanisms
In emotion science, emotions are studied from a few different perspectives. Research looks at emotions as a biological evolution (affect programs; e.g., Ekman, 1992), as a social construct (Markus & Kitayama, 1991), as the perception of bodily changes (e.g., increase in heart rate is perceived as fear; James, 1884, Lange, 1885, as cited by Fox, 2008), or as a cognitive appraisal (Lyons, 1999, as cited by Fox, 2008). As a result, the field of emotion science has different schematizations of the relevant processes, different terminology (e.g., motion, affect, mood), and different research methods.
An interdisciplinary model that tries to incorporate these different perspectives, is the quartet theory of human emotions (Koelsch et al., 2015). The theory distinguishes four different affect systems, revolved around distinct brain regions; the brainstem, diencephalon, hippocampus, and orbitofrontal cortex. These affect systems interact with each other, and with biological processes, called effector systems. These effector systems contain motor systems that influence action tendencies, motor systems that influence
expressions of emotions, physiological arousal systems, attention systems, and memory systems. As such, it incorporates the many different ways in which emotions are induced, and the many different ways in which they can be expressed. These affect systems, interacting with the effector systems, result in an emotion percept, which is an unverbalized subjective feeling. This emotion percept can then be transformed into a linguistic description, which is a subjective report of the felt emotion.
The different ways of measuring emotions can be divided into four categories; behavioral, physiological, neurological, and introspective. Behavioral measurements measure the motor effector systems that influence action tendencies and facial (or other bodily) expressions. For example, approach and avoidance behaviors are expressions of positive and negative affect (Fox, 2008), and facial expressions can express many different emotions (Ekman, 1992). However, it is possible for a person to suppress these reactions to some extent. Physiological measurements can reliably represent arousal (e.g., heart rate, pupil dilation, skin conductance, and hormones), but not the valence of the emotion (Fox, 2008). Neurological measurements show neural correlates of the four affect systems as discussed by Koelsch et al. (2015). Although activation in different areas provides some information about different types of emotions (e.g., basic or complex emotions), findings remain correlational and we cannot yet infer from neural activation what a person experiences. Finally, introspective measurements can provide information on arousal and valence, and also more specific distinctions (e.g., mixed emotions), but are considered to be less valid because they require a translation from the emotion percept into a linguistic description. Nevertheless, one cannot deny that self-report is an important window into the phenomenological experience of emotions (Feldman Barrett, Mesquita, Ochsner, & Gross, 2007). In addition, some research has shown that subjective reports are actually more reliable than physiological measurements (Dermer & Berscheid, 1972; McNair, Lorr, & Droppleman, 1992; Thayer, 1970; as cited by Husain, Thompson, & Schellenberg, 2002). Considering the pros and cons of these different methods, depending on the question at hand, a combination of methods is often necessary.
2.2 Musical and Everyday Emotion: Different or Same?
Following this more general overview of human emotion mechanisms and measurements, the subsequent question arises whether musical emotions are special in any way. BRECVEMA (Juslin, 2013; Juslin & Västfjäll, 2008) is a model designed to describe the mechanisms that evoke emotions when listening to music, specifically. It entails eight mechanisms; brain stem reflex, rhythmic entrainment, evaluative conditioning, emotional
contagion, visual imagery, episodic memory, musical expectancy, and the later added aesthetic judgment. Although BRECVEMA shows some overlap with the quartet theory of
emotions (e.g., the involvement of the brain stem), and some mechanisms may have a foundation in the quartet theory, BRECVEMA only describes mechanisms that are applicable to music. Juslin (2013) argues that, still, the first seven mechanisms that are described, also lead to emotions in everyday situations unrelated to music. The last mechanism of aesthetic
judgment distinguishes musical emotions from everyday emotions. When listening to music as a piece of art, one can judge the aesthetics of the music based on its novelty, creativity, or style. This has influence on both the emotions that are experienced, and the preference for the music. However, the emotions that are experienced are not different from everyday emotions, according to Juslin (2013). It is merely the causal process leading towards these emotions that is different.
Others, however, claim that the aesthetic nature of music and art in general does cause emotions that are inherently different from everyday emotions. Scherer and Zentner (2008) claim that everyday emotions are high-intensity reactions, which ask for proactive behavior to achieve a specific goal. Aesthetic emotions do not lead to goal-oriented responses or proactive behavior. This is similar to Kant’s description of the aesthetic experience, which is detached from reality and purely concerned with form (Kant, 2001, as cited by Scherer & Zentner, 2008). Flaig and Large (2013) dislike the idea of yet another mechanism to explain musical emotions, but rather argue that music has a special way of engaging with the brain. Although we know that music, and other forms of art, can induce emotions that also occur in daily life (Zentner et al., 2008), research to this day has not been able to show whether there are qualitative differences between musical and everyday emotions. This is partly due to research methods that have the problems we described above; they are either very subjective, transformed due to language, or not very specific. Researchers agree, however, that aesthetic judgment plays a role in emotional responses to music, but its exact effects and importance are yet unknown. Compared to other forms of art, research on music has focused more on emotion whereas research on other forms of art have focused on the aesthetic experience (Tiihonen, Brattico, Maksimainen, Wikgren, & Saarikallio, 2017). An integration of these different disciplines could shed more light on emotional responses in an aesthetic context, and whether music is special compared to other forms of art as well.
2.3 Musical Emotion and Aesthetic Pleasure
Although the mechanisms through which music induces emotions in a listener can say a lot about what people like to listen to, the relationship between emotion and pleasure is not so straightforward. For example, in some cases, negative emotions in music, like sadness, are enjoyed (Garrido & Schubert, 2011, 2013a, 2013b). More specifically, the emotion that music expresses does not necessarily correlate with the emotion that is felt or whether the music is liked (e.g., happy music causes happy emotions and enjoyment; sad music causes sad emotions and dislike).
One explanation for this discrepancy is the aesthetic judgment that occurs when listening to music. The Parallel Processing Hypothesis (Schubert, 2016) argues that in an aesthetic context, certain emotional processes can be uncoupled. For example, motivational action tendency and subjective feeling usually show similar valence. However, the appraisal whether something is a real-life situation or an aesthetic context can change how motivational action tendency and subjective feeling interact, and the latter situation of an
aesthetic context can uncouple the two processes. Although this hypothesis first needs thorough testing, along with other claims about the aesthetic experience (see e.g., Brattico, Brigitte Bogert, & Jacobsen, 2013; Ishizu & Zeki, 2011; Leder, Belke, Oeberst, & Augustin, 2004), it asks for consideration of the role of aesthetics in inducing pleasure when listening to music.
Just as in the mechanisms through which music induces emotion, the mechanisms through which music induces aesthetic pleasure are dependent on the content of the music, the characteristics of the listener, and the context in which the listening takes place. Brattico et al. (2013) propose that an aesthetic experience requires the appraisal that the listening takes place in an aesthetic (non-everyday-life) context, which is also influenced by the attitude, intentionality, and attention of the listener. When this combination of factors is present, it is thought that the aesthetic experience can induce emotions in the listener, and an individual judgment of beauty, liking, and preference will be made. Although this explanation seems sufficient in explaining both similarities and discrepancies between every day and musical emotions, it does require more direct testing with an explicit context manipulation (daily life vs. aesthetic context). For example, it is not clear whether the causality described here leaves room for emotions in response to music that seem to be very direct and involuntary, arising perhaps before an aesthetic appraisal can be made. In addition, the context of listening can be ambiguous, when for example a person is distracted by another task that does not entail an aesthetic appraisal, (such as a conversation) and notices music that he or she likes in the background
2.4 Musical Pleasure and Reward
A final mechanism that is related to musical preferences and pleasure, is reward. A significant amount of research has shown that there is a core neural network related to reward mechanisms. This core network consists of dopaminergic brainstem nuclei, i.e., the ventral tegmental area, ventromedial and orbitofrontal cortices, amygdala, insula, and the striatum (Zatorre, 2015). Activity in these areas is related to primary reinforcements, which are rewarding because they are essential for our survival (i.e., food, sex, social contact), and secondary reinforcements, which are essential because they give access to primary rewards (i.e., money). Interestingly, feelings of pleasure or enjoyment in response to music are correlated with activity in this same reward network (Koelsch, 2014). More specifically, musical enjoyment is related to dopaminergic activity in the striatum (Salimpoor, Benovoy, Larcher, Dagher, & Zatorre, 2011). However, music is not clearly considered as a primary or secondary reinforcement, but rather, the music is thought of as a reward in itself.
The core network of reward is also able to dissociate between the anticipation or prediction of an event (dorsal striatum), and the perceived pleasure (reward) for fulfilling that prediction (ventral striatum; Salimpoor et al., 2011). Research on temporal prediction and pleasure shows that prediction plays an important role in the experience of pleasure in response to music (Trainor & Zatorre, 2009, as cited by Zatorre, 2015). The role of prediction and expectation had already emerged in the BRECVEMA model on musical
emotion. Thus, it appears on a theoretical and neurological level, that prediction is an important factor in inducing musical pleasure.
Interestingly, although musical reward and general award seem to involve the same core mechanisms, there are people who show anhedonia in response to music specifically, independent of general anhedonia, abnormal reward sensitivity, or amusia (Mas-Herrero, Zatorre, Rodriguez-Fornells, & Marco-Pallarés, 2014). Musical anhedonia is also present in some subjects with lesions in temporal, frontal, or parietal regions instead of the reward related areas (Agustus et al., 2015; Satoh, Nakase, Nagata, & Tomimoto, 2011). One possible explanation for this discrepancy could be that musical reward requires more frontal cortical mechanisms, because it is more complex and abstract than the described primary and secondary reinforcements. This might also be related to the role of aesthetic appraisal, arguably a higher-level mechanism that is involved in inducing emotions and pleasure in response to music.
To summarize, when it comes to emotions, many different factors are at play, and many different theoretical and methodological approaches exist, making it difficult to study emotion as a whole. This section shows that music cognition and cognition in general show similarities when it comes to emotional processing and reward mechanisms. What seems to dissociate musical emotion from emotion in general is the aesthetic situation of the music. Prediction and reward mechanisms are important factors in emotion and pleasure induction. Again, reward mechanisms related to music and general cognitive processing, show similarities (e.g., in associated brain areas), but also differences (e.g., based on lesions). The role of prediction will be the subject of the next section.
3. Focus on Single Mechanism: Prediction
As becomes clear from the literature described above, music induces emotions and pleasure depending on many different mechanisms, based on factors that are present in both the music and the listener. As a result, the individual preference that people show for certain pieces of music is a complicated subject. Therefore, in this review we will not approach the subject as a whole, but rather focus on one of its mechanisms; expectation. Expectation is one of the first described mechanisms in music cognition (Meyer, 1957), subsequently featured in BRECVEMA (Juslin, 2013; Juslin & Västfjäll, 2008), and explored more thoroughly in the ITPRA theory of musical expectation (Huron, 2006), which is discussed in more detail in this section. First, however, this section will discuss predictive processing as a more general mechanism. Afterwards, to keep in line with the comparisons made in section two, predictive processing in music will be discussed and compared.
3.1 The Predictive Brain
The idea of the brain as a prediction machine has a long history in cognitive research, as has been thoroughly reviewed by Clark (2013). The idea is that higher-level mechanisms create predictions of what is going to be observed. In a hierarchical, cascading fashion,
predictions are made for each level below it. When information from these lower level processes does not coincide with the higher-level predictions, an error is detected and the higher-level prediction models need to be adjusted. In this sense, the predictive brain works in a bidirectional manner; top-down to infer predictions, and bottom up to translate error into new prediction models. This error is also known as surprisal, which is a subconscious and basic form of the conscious experience of surprise. The end-goal of the predictive brain is formulated as having zero prediction error, or no surprisal.
An example of predictive processing lies in the retinal ganglial cells of the human eye, as demonstrated by Hosoya, Baccus, and Meister (2005). Based on incoming light in one location of the retina, the brain will predict which light will fall onto the surroundings of that location, both in space and time. Which light actually falls on those surroundings, will be subtracted from the predicted light, and be processed. Thus, the brain makes top-down predictions about its surroundings, and processes the error bottom-up to form new predictions for the future. This is a more efficient way of processing, but also enables the brain to detect the unexpected (surprisal) events that are perceived.
Clark (2013) also describes ‘action-oriented predictive processing’, where action is incorporated in the predictive model. Motor actions show the same characteristics as just described. To perform a goal oriented action, the brain makes predictions about what actions need to be performed to reach the goal, and adjusts these predictions based on feedback on the way there. Furthermore, goal-directed actions influence perception, and hence the two cannot be regarded as separate entities. Perception and action work together, as Clark hypothesizes, to make optimal predictions with as minimal error as possible.
As a theory, the predictive brain combines perception, action, and attention into one unified model. It is able to explain a wide range of phenomena and is parsimonious in its formulated rules. However, as Clark (2013) describes, neural evidence for the predictive brain is still in its infancy. One source of evidence for the predictive brain is the fact that some forms of perceptual processing occur according to Bayesian statistics. For example, color constancy and some color illusions are the effect of Bayesian statistics in that there is a prior expectancy interacting with perception (Brainard, 2009). This behavioral data is indirect evidence, as is another area of research, namely computational models that simulate predictive processing and demonstrate the same outcome as human behavior. Examples of this are in the non-classical receptive field effects (Rao & Sejnowski, 2002). One of these effects, for example, concerns the orientation of objects in the receptive field of a cell. If one object of a certain orientation is surrounded by objects of a different orientation, this causes a stronger effect in the cell than when all objects are oriented similarly. The response of such cells is best simulated by a prediction model that concerns error, instead of the fixed content of what is perceived. Thus, although there is some indirect evidence for the theory of the predictive brain, more research is necessary.
Interestingly, and as has already been mentioned in previous sections, prediction and expectation have a rather longer history of research in music cognition. Its findings may
prove to be useful for the theory of the predictive brain. These music cognition studies, too, largely consist of indirect evidence of behavioral and computational studies, although there is also neurological evidence that shows brain areas related to prediction.
3.2 Musical Expectation
The first to propose a role of expectation in music perception was Leonard B. Meyer (1957). Huron (2006), in exploration of this proposed relationship, described four different types of expectations that are at play when listening to music; veridical expectations, schematic expectations, dynamic expectations, and conscious expectations. Veridical expectations are derived from episodic memory on how a (familiar) piece of music should continue. Schematic expectations are formed based on generalizations from previous listening experiences. Dynamic expectations are formed based on more short-term memories of the music that is being played at that time, and how it should logically progress. This type of expectation is updated as the music continues. Finally, conscious expectations are a more explicit type of expectations about how a piece of music will sound.
In further exploration, Huron formulated the ITPRA theory (2006). ITPRA describes five types of responses that are related to musical expectation: imagination, tension,
prediction, reaction, and appraisal (Huron, 2006). As Huron describes, each of these
responses were developed to be able to respond to our (auditory) surroundings, and composers use these responses to catch the attention and evoke emotions in their listeners. The responses of imagination and tension occur before the actual expected event. Prediction is a transient state of reward (positive) or penalty (negative) in response to whether the event was correctly predicted or not. Reaction and appraisal are responses that evaluate the predicted response, and might be more so related to pleasure than emotion. The appraisal, for example, can label the event as aesthetic, a feature that also is present in BRECVEMA.
Whether people actually have expectations when listening to music and the specific characteristics of these expectations, has been studied for musical harmony most extensively. Event-related brain potentials (ERPs) show differences in components in response to variability in harmonic expectations; namely the late positive component (Besson & Faïta, 1995), P3 and early negativity (Steinbeis, Koelsch, & Sloboda, 2006), and P300 (Carrión & Bly, 2008; Janata, 1995). Behavioral studies found that violation of harmonic expectations had an influence on reaction times and preferences (Bharucha & Stoeckig, 1986; Loui & Wessel, 2007). A more recent physiological, ERP, and behavioral study found differences in tension, overall subjective emotion, electrodermal activity (EDA), and early negativity in the ERP when harmonic unexpectedness was manipulated (Steinbeis et al., 2006). Although there is still ongoing debate about, e.g., whether musicians and non-musicians form the same expectations, these studies all agree on the fact that people have harmonic and melodic expectations when listening to music.
Expectations on musical rhythm have been studied only more recently, but show conclusions similar to the studies above. Previous research shows that people form
hierarchical and preattentive expectations of the beat and meter (Ladinig, Honing, Háden, & Winkler, 2009). These expectations also become apparent in high-frequency gamma band activity in the auditory cortex (Zanto, Snyder, & Large, 2009). A functional magnetic resonance imaging (fMRI) study also showed that parts of the basal ganglia are associated with beat prediction (Grahn & Rowe, 2013), a brain area that is also part of the core emotion network discussed in the previous section. Another ERP study showed that the same responses are found (although not identical in latency and magnitude) with an actual and an imagined beat pattern (Schaefer, Vlek, & Desain, 2011), providing further evidence that a top-down mechanism from the listener imposes a metrical structure on the external input. A study that measured surprise through pupillary response, found that participants, while distracted by another task, showed different surprise responses to omissions of beat at different points of metrical salience, with no differences between musicians and non-musicians (Damsma & van Rijn, 2017). Finally, a computational study showed that a probabilistic model that infers meter on the rhythm, is better at predicting beat onsets than a probabilistic model without an imposed meter (van der Weij, Pearce, & Honing, 2017). Altogether this indicates that people form expectations based on harmony and rhythm, which may lie at the core of the aesthetic response to musical stimuli. There are some indications that this happens without necessarily directing attention to the music. Although differences between musicians and non-musicians are found, the expectations appear to be present in all people to some extent.
3.3 Error Minimization
One critique, or question mark, regarding the theory of the predictive brain, is formulated by Clark (2013) as follows:
How can a neural imperative to minimize prediction error by enslaving perception, action, and attention accommodate the obvious fact that animals don’t simply seek a nice dark room and stay in it? Surely staying still inside a darkened room would afford easy and nigh-perfect prediction of our own unfolding neural states? Doesn’t the story thus leave out much that really matters for adaptive success: things like boredom, curiosity, play, exploration, foraging, and the thrill of the hunt? (Clark,
2013, p. 13).
In other words, if the ultimate goal of the brain is to minimize its prediction error, why is it adaptive to explore or play, a situation in which error is inevitable? Translating this to music; if we are constantly trying to predict how the music will continue, why don’t we just listen to the same song over and over again, to make it easy? Or, why would we listen to music at all?
One reply to this critique, as formulated by Friston (2010; Friston, Daunizeau, & Kiebel, 2009) is that exploration and search are inevitable, because our surroundings are constantly changing, and thus, in minimizing error, organisms cannot stay still in the same place doing the same thing. It is simply not possible to survive in this darkened room. However, this does not explain the voluntary exposure of organisms to situations where
prediction error will inevitably occur. In music, violation of expectation is bound to happen. Anecdotally, this is also preferred to some extent; nobody wants to listen to perfectly predictable music. Again, the aesthetic context of music listening may explain this distinction. One could see music as a harmless form of play, where the brain can learn to form new prediction models. However, the type and size of the prediction error are likely crucial in understanding the resulting response. The error, i.e., violation of expectation, in music and its effect on liking of the music will be the topic of the next section.
4. Violation of Expectations: Do we Like it? 4.1 Rhythmic Complexity
In the introduction of this review, clarifications were provided about the exact definitions of rhythm, meter, and beat. When talking about rhythmic or musical complexity, further clarifications are also necessary, as previous studies have shown variable uses of the term. In this review, a rhythm is considered to be more complex, when it becomes harder to
predict. There may be no sense of beat or pulse because the metric structure is unpredictable, or error predictions may occur because the rhythm deviates from what the beat and meter imply.
Some studies regard rhythmic complexity as the amount of syncopation (Chapin et al., 2010; Witek, Kringelbach, & Vuust, 2015). Witek describes this interpretation as follows:
Syncopation provides an example of the structural complexity of groove that opens up empty spaces in the rhythmic surface which invite the body to fill in through entrainment. By filling in temporal gaps, bodies extend into the musical structure, and the desire to fill the gaps and complete the groove affords a participatory pleasure. (Witek, 2016, p. 1)
Syncopation is also described as “rhythmic events which violate metric expectations” (Vuust & Witek, 2014, p. 6). Thus, the complexity of a syncopated rhythm is determined by the deviation from the expected metric structure. A subtler form of syncopation is performer timing, a deviation from the rhythmic structure instead of the metric structure. One could also say this is a less or unconscious form of timing deviation, possibly more related to surprisal, as compared to syncopation. For example, micro-timing is generally incorporated in drum computers’ sequences, to humanize the sound (Frühauf, Kopiez, & Platz, 2013).
An fMRI study on performing motor sequences quantified rhythmic complexity by increasing the number of IOIs that are present in the rhythm to be performed; ranging from only one IOI (isochronous rhythm) up to 6 different IOIs present in the rhythm (Lewis, Wing, Pope, Praamstra, & Miall, 2004). This form of complexity can be seen as rhythmic complexity, but also more specifically as sequence complexity, which is not necessarily related to the meter in the rhythm or rhythmic expectations. Rather, it would test the ability to memorize or keep a certain amount of information in working memory. As such, rhythmic complexity can also be described as the amount of information that is embedded in a rhythmic sequence (Shmulevich, Yli-Harja, Coyle, Povel, & Lemström, 2001). Predictability is
decreased in that case, due to the sheer amount of information. In the case of syncopation, expectations are violated. In this case, however, expectations are difficult to form. This measure of complexity, as entropy, is a distinct measure of complexity from measures of expectation violations. Another fMRI study did a similar manipulation with different IOIs, but also kept in mind the metric complexity, by providing three conditions of simple metric, complex metric, and non-metric sequences (Chen, Penhune, & Zatorre, 2008b). This provides complexity quantifications on the levels of the rhythmic sequence, but also on the metric level. Nevertheless, the manipulation of number of IOIs makes use of two capacities – working memory and prediction - whereas this review aims to focus on prediction only.
Another study tested participants’ ability to synchronize to simple and complex rhythms by changing the complexity of the rhythmic structure (Large, Fink, & Kelso, 2002). Simple rhythms, in this case, were isochronous rhythms. The complex rhythms were
random, in the sense that the participant could not learn the sequence. However, the
rhythms were still metrically sound (to Western music standards), and thus possible to synchronize with in terms of the beat. In this case, the complexity was quantified as predictability of the rhythm, but not predictability of the meter. By showing that people are able to tap along to rhythmically predictable and unpredictable sequences, which remain metrically predictable, this study implies that complexity is more influential at the level of meter, than at the level of the rhythm sequence.
When looking at the complexity of the metric structure itself, one could argue that some cultures have more complex structures than others. This is most often described with regards to the hierarchical structure of the meter (Lerdahl & Jackendoff, 1983, London, 2012, as cited by Vuust & Witek, 2014), a concept that was shortly introduced in the introduction. This hierarchy is built up out of different levels of isochronous rhythms, where each level multiplies or subdivides the other levels. Multiplication or subdivision ratios of 1:1 or 1:2 (Western music) are considered to be simple, and ratios of 3:2 complex (music from South Asia, Africa, Middle East, or Eastern Europe; Clayton, 2001; London, 1995; Pressing, 1983; as cited by Hannon, Soley, & Ullal, 2012). In this respect, complexity is a term described from a Western perspective. However, even more complex ratios of, e.g., a 7:4 ratio, are found to be difficult to process by infants, which indicates that it is not merely a learning bias that influences perceived ratio complexity (Hannon, Soley, & Levine, 2011).
Table 1 shows an overview of the studies that used complexity as an operationalization. As is also described by Sadakata, Desain, and Honing (2006), some quantifications and definitions of complexity concern the structure of the rhythmic pattern, or the ratios of the metrical structure, or sometimes both. To summarize, what is important in light of this review is to regard complexity as a measure of the predictability of a rhythm. This predictability can be influenced by (a) syncopation/micro-timing (deviation from the rhythm or meter), (b) the amount of information present in the rhythm, (c) the randomness of the rhythmic structure, while metric structure remains stable, or (d) the ratio of the metric structure. Especially the latter is influenced by culture, but not entirely so, as was shown by Hannon et al. (2011). The amount of information present in the rhythm, appears
to be a less ideal quantification, because it is confounded by working memory capacity. The complexity of the rhythmic structure, did not appear to influence predictability, as long as the metric structure remained intact (Large et al., 2002). Thus, metric structure appears to be the most valuable component of rhythmic complexity in terms of predictability, in that either there is a violation of expectation (syncopation), or an inability to form expectations (metric structure ratios). The standard representation of rhythm, with more and less salient beats, thus provides the grid for generating predictions.
Table 1
Summary of Different Types of Complexity Manipulations and Their Relationship to Predictability
Complexity Manipulation
Predictability References
Syncopation Metric Expectation Violation Chapin et al. (2010), Witek,
Kringelback, & Vuust (2015), Vuust & Witek (2014)
Micro-timing Rhythmic Expectation Violation
Frühauf et al. (2013)
Amount of different IOIs
Hard-to-Predict Rhythm & Entropy
Lewis et al. (2004)
Amount of different IOIs
Hard-to-Predict Rhythm & Hard-to-Predict Meter
Chen et al. (2008b)
Isochronous vs. Random rhythm
Hard-to-Predict Rhythm Large et al. (2002)
Metric Hierarchy Ratios
Hard-to-Predict Meter Hannon et al. (2011)
4.2 Affect and Enjoyment in Response to Complexity
The main question of this review is whether complexity is enjoyed. One study on complexity and hedonic value theory (Berlyne, 1970), led to an influential concept in music cognition stating that the relationship between complexity and enjoyment is described by an inverted U-curve, with complexity on the x-axis and enjoyment on the y-axis (see e.g., North & Hargreaves, 1995; Schaefer, Overy, & Nelson, 2013; Vuust & Witek, 2014; Witek, Clarke, Wallentin, Kringelbach, & Vuust, 2014). In other words, the complete absence of
complexity (or prediction error) does not lead to arousal, positive affect, or enjoyment, nor does an abundance of complexity. Instead, there is an optimal level of complexity that is enjoyed and preferred. Note that not the absolute absence of error, in this case, seems to be the goal, but rather a certain amount of optimal error that can be processed. Research on the relationship between rhythmic complexity and enjoyment is rather scarce. Therefore, studies on general musical complexity or harmonic complexity are discussed as well.
One of the earliest studies on the effect of complexity on preference did not separate melody, harmony, and rhythm from each other, but rather asked their participants to rate the subjective musical complexity of different pieces of music (North & Hargreaves, 1995). They indeed found an inverted U-relationship between subjective complexity and self-reported liking. The most recent study on the effect of complexity on preference also did not separate different types of complexity from each other (Madison & Schiölde, 2017). However, they found that it was familiarity that influenced self-reported liking, but not complexity, when presenting participants with different levels of subjective complexity in existing (but unfamiliar) popular music. Some reasons for these conflicting results could be that complexity was measured based on self-report and the term musical complexity was too broad to be reliable.
Research that did look at more specific complexity of harmony also shows some conflicting results. For example, Koelsch, Fritz, and Schlaug (2008) found that irregular chord progressions (i.e., unpredicted chords) were perceived as less pleasant than regular chord progressions. This effect was correlated with amygdala activity, signifying that emotional processing was involved. Two studies (Koelsch, Kilches, Steinbeis, & Schelinski, 2008; Steinbeis et al., 2006) found physiological arousal responses for unexpected chords, but did not measure differences in valence. A more recent study with live music, found that unexpected moments induced self-reported arousal and negative affect, as well as differences in physiological arousal when compared to expected moments (Egermann, Pearce, Wiggins, & McAdams, 2013).
Interestingly, however, studies on musical chills, a desired and positive emotional reaction to music, did find a positive affective reaction to violation of harmonic expectations. For example, Sloboda (1991) found that music sections where people reported to experience chills were characterized by unexpected harmonies. Similarly, physiological studies on the chill response found that these responses occurred at sections in the music that were harmonically ambiguous and unexpected (Grewe, Kopiez, & Altenmüller, 2009; Grewe, Nagel, Kopiez, & Altenmüller, 2007; Guhn, Hamm, & Zentner, 2007).
Thus, evidence on the preference for or liking of harmonic expectancy violations is conflicting as well. One explanation for this is that musical chills are from a completely different category of emotions than the more ‘regular’ emotions in response to music that also range from negative to positive. However, although chills are perhaps different when it comes to emotion, chills are still related to the enjoyment of music. Another explanation is
that it may be very difficult to manipulate expectedness of harmonic progressions, as was done in the studies that did not find positive affective responses. The studies on chills used existing music and hence may have been influenced by confounds that were not specific to violation of expectations. Finally, the type of expectancy violations is variable as well, comparing, e.g., the unexpected harmonies from Sloboda (1991) with the unexpected moments from Egermann et al. (2013). Overall, this research shows that preference, enjoyment, arousal and complexity all are connected, but the exact relationship between these factors is something that needs to be studied more thoroughly.
As was noted at the beginning of this section, there have been fewer studies that examine the influence of rhythmic complexity on affect and enjoyment. However, some evidence does point to a preference for ‘slight surprisal’. Firstly, anecdotal evidence, as was mentioned in section 4.1, is that drum machines ‘humanize’ beat patterns by applying slight micro-timing deviations, because these were preferred over perfectly-timed patterns (Frühauf et al., 2013). A web-based study that manipulated different degrees of syncopation in funk drum-breaks found an inverted-U relationship between the amount of syncopation and self-reported pleasure (Witek et al., 2014). A similar study found that syncopation was enjoyed more than no syncopation, also based on self-report (Keller & Schubert, 2011).
To the best of my knowledge, no studies have been published about enjoyment of rhythmic complexity concerning complex metrical structures, i.e., whether there are enjoyment differences between simple (2:1), complex (3:2), and highly complex (7:4) metric structure ratios (note that the example ratios differ in complexity based on Western standards). Thus, in regard to rhythmic complexity, the violation of metric expectation (i.e., the note appeared slightly before or after a metrically salient point) appears to be preferred to some extent, but such claims cannot be made yet about the difficult to form metric expectations (i.e., the listener has difficulty predicting what are the metrically salient time points).
To summarize, in the case of music complexity, findings on preference are inconsistent, as also becomes apparent in the overview of studies in Table 2. Looking at complexity as a musical whole is rather unspecific in its manipulation, and would be improved by looking at specific musical parameters. The importance of this distinction is also reflected in Huron’s (2006) different types of expectations (see section 3.2 for an explanation). That is, asking to rate the complexity of a piece of music requires the participants to report on their conscious expectations. However, manipulating harmonic or rhythmic complexity does not need to rely on conscious expectations, but can address veridical, schematic, and dynamic expectations more directly. Furthermore, it seems plausible that harmonic complexity relies more on veridical and schematic expectations, whereas rhythmic complexity might concern dynamic expectations more strongly, especially with regards to syncopation or micro-timing. This distinction could also indicate a difference in preferred amount of prediction error when comparing, e.g., rhythmic with harmonic complexity (see also Schaefer et al., 2013).
To further summarize, when looking at harmonic expectations more specifically, it is easier to operationalize and manipulate them. However, in this case too, findings were inconsistent. Studies on measuring the affective responses to unexpected chord progressions generally found negative affect in response to unexpected chords. On the contrary, studies on musical chills found that this positively valenced emotional experience is actually predicted by a violation of harmonic expectations. This discrepancy may either be caused by the fact that musical chills are an emotional category in itself, or by the fact that the studies on musical chills used existing music without explicitly manipulating different levels of complexity, whereas the other studies created their own chord progressions with more subtle complexity manipulations. This latter difference implies that different types of expectation were tested.
Finally, the scarce literature on rhythmic complexity did find a preference for slight violation of expectation and an inverted U-curve for the effect of complexity on enjoyment. Future research in this area is necessary to refine our understanding of the effect of rhythmic complexity. Furthermore, the theory of the predictive brain aims to function as a unifying, all-encompassing mechanism, and would thus predict that what is true for rhythmic meter, is also true for harmony. On the other hand, different types of expectations (conscious, veridical, schematic, dynamic) may have different interactions with preference. Future studies should use manipulations that show subtle differences in complexity (instead of coarse differences), use reliable affective measurements, and use clear concepts of different types of expectations. One possible source that may also shed more light on the effect of rhythmic complexity is the relationship between rhythm and movement. This is an area of research that has been studied more extensively, and will be discussed in the next section.
Table 2
Overview of Studies on the Interaction between Complexity and Preference, Enjoyment, or Emotion
Reference Type of Complexity Type of Measurement
Outcome North & Hargreaves
(1995)
Overall:
Self-reported complexity
Self-reported complexity & liking
Inverted U-curve
Madison & Schiölde (2017)
Overall:
Self-reported complexity
Self-reported complexity & liking
No effect
Koelsch, Fritz, & Schlaug (2008) Harmonic: Unexpected chord progressions Self-reported valence, fMRI Negative affect, amygdala activity
Koelsch, Kilches, et al. (2008); Steinbeis et al. (2006) Harmonic: Unexpected chord progressions
Physiological arousal Heightened arousal
Egermann et al. (2013) Melodic: Computational expectedness of melodic progression Self-reported
valence and arousal, facial EMG,
physiological arousal
Self-reported arousal and negative effect, physiological arousal
Sloboda (1991) Harmonic: Unexpected harmonic progressions
Self-reported chills More chills
Grewe, Kopiez, & Altenmüller (2009); Grewe et al. (2007); Guhn et al. (2007) Harmonic: Unexpected or ambiguous harmonic progressions Physiological chill response Physiological chill peaks
Witek et al. (2014) Rhythmic:
Syncopation (scale)
Self-reported pleasure
Inverted U-curve
Keller & Schubert (2011) Rhythmic: Syncopation (with or without) Self-reported enjoyment Syncopation more enjoyed than no syncopation
5. Entrainment: Moving Along with the Rhythm
In Clark’s (2013) review of the predictive brain, he emphasized the importance of action in the theory of predictive processing. That is, not only do we have predictive models for what we will perceive, we make similar Bayesian inferences when performing actions. Even more so, action and perception work interactively with each other in making predictions. Clark hypothesizes that they work together to minimize error. As became clear from the previous section, minimal error may not be the end goal when it comes to music. In music, action also has a role in processing. For example, Leman (2016) describes how sensorimotor synchronization boosts enjoyment of music in his book The Expressive
Moment. In addition, in the quartet theory of emotion (Koelsch et al., 2015) the different
affect systems interact with (amongst others) the motor effector systems. Thus, predictive action processing is related to enjoyment and emotion. Witek et al. (2014), who were already cited in the section on preferences for syncopation, also related prediction, action, and enjoyment by describing the desire to move along to the music (i.e., groove). The
current section will focus on perceptual, motor, and neural entrainment; the synchronization of two independent oscillating processes.
5.1 Rhythm Perception and Motor Processing
Music and movement are inevitable intertwined. One the one hand movement is necessary to create music, and on the other hand music can make us move. Research investigating the neural underpinnings of rhythmic processing has shown a link with motor processing, based on overlap in brain regions (Chen et al., 2008b; Dhamala et al., 2003; Karabanov, Blom, Forsman, & Ullén, 2009; Thaut et al., 2009). Even without any overt movement, motor areas are active when listening to a rhythm; namely the premotor cortex (Bengtsson et al., 2009; Chen, Penhune, & Zatorre, 2008a; Grahn, 2009; Grahn & Brett, 2007; Grahn & Rowe, 2009; Schubotz, Friederici, & von Cramon, 2000), the cerebellum (Bengtsson et al., 2009; Chen et al., 2008a; Grahn & Brett, 2007; Schubotz et al., 2000), the pre-supplementary motor area (Bengtsson et al., 2009; Grahn & Brett, 2007; Schubotz et al., 2000), supplementary motor area (Bengtsson et al., 2009; Chen et al., 2008a; Grahn, 2009; Grahn & Brett, 2007; Grahn & Rowe, 2009; Schubotz et al., 2000), and the basal ganglia (Grahn, 2009; Grahn & Brett, 2007; Grahn & Rowe, 2009; Schubotz et al., 2000). More specifically, the basal ganglia and supplementary motor area show more activity when a pulse is perceived, compared to rhythms where no pulse is perceived (Grahn & Brett, 2007; Grahn & McAuley, 2009; Grahn & Rowe, 2009). In addition, when people listen to rhythms with a beat (compared to rhythms with no beat), functional connectivity between motor areas (premotor cortex, supplementary motor area), basal ganglia, and auditory cortex increases (Grahn, 2009; Grahn & Rowe, 2009). Altogether, these findings show that based on brain activity and connectivity, rhythm and movement seem to have a clear connection.
A clinical demonstration of the link between rhythm and motor processing is the use of rhythmic auditory stimulation (RAS). RAS is a form of gait rehabilitation, mostly applied in Parkinson’s or stroke patients. In movement therapies with RAS, patients are supposed to synchronize their movements to a metronome or musical beat pattern (Benoit et al., 2014; de Bruin et al., 2010; Nieuwboer et al., 2007; Thaut et al., 1996). The intervention shows some mixed results, which may be due to individual differences or differences in musical stimuli characteristics between studies (Lim et al., 2005; Nombela, Hughes, Owen, & Grahn, 2013; Wittwer, Webster, & Hill, 2013). Another factor that may influence the efficacy of RAS or other musical interventions is the influence music has on patients’ motivation (Schaefer, 2014). This influence of motivation indicates that it is important to understand which factors influence musical preferences, in order to optimize the effect of musical interventions.
One feature of the music stimuli that could be related to the efficacy of RAS is the groove (Janata, Tomic, & Haberman, 2012; Madison, 2006; Madison, Gouyon, Ullén, Hörnström, & Umeå, 2011). Groove is described as “a musical quality that makes us want to move with the rhythm or beat” (Stupacher, Hove, Novembre, Schütz-Bosbach, & Keller, 2013, p. 127). In this definition, there is no description of what exactly this musical quality entails. Iyer describes groove as “an isochronous pulse that is established collectively by an
interlocking composite of rhythmic entities” and an “attentiveness to an additional unifying rhythmic level below the level of the tactus” (Iyer, 1998, Chap. 2, p. 7, as cited by Madison, 2006). As such, groove is the beat that urges its listeners to move along with it. Music that has groove is generally also associated with swing music (Madison, 2006), which is a syncopated form of music. Research shows that groove is induced by several music factors, one of which is syncopation (Madison & Sioros, 2014), but surprisingly not by the more subtle micro-timing deviations (Davies, Madison, Silvo, & Gouyon, 2013; Frühauf et al., 2013). This discrepancy may be due to differences between musical genres, which is also suggested by Frühauf et al. (2013). The relation between groove and syncopation can be described by the same inverted U-curve as was introduced to explain the relation between complexity and enjoyment (Sioros, Miron, Davies, Gouyon, & Madison, 2014; Witek et al., 2014).
Thus, rhythm, movement and enjoyment are related, and also seem to show a relation to prediction, in that syncopation is related to both groove and liking. The exact connection between these different factors needs further testing. This interaction does translate nicely to Clark’s (2013) action-oriented predictive processing interacting with perception. However, in the case of music, there is an optimal level of prediction error (i.e., syncopation) that will lead to a peak level of groove and pleasure. Such a relation was not found for micro-timing, and has not been studied for other types of rhythmic prediction error (e.g., complex metric ratios). The next section will describe the theories describing the relationship between rhythm, moving, and enjoyment, and relate this back again to action-oriented predictive processing.
5.2 Entrainment Theories & Pleasure
The Dynamic Attending Theory (DAT), as formulated by Jones (Jones & Boltz, 1989; Large & Jones, 1999), argues that beat perception is a result of neural oscillations resonating with and entraining to an external rhythm. The idea of resonating oscillations can also be explained from a Darwinian perspective, because it can be seen as a vital mechanism of the brain to attend to and respond to the external environment. For example, some studies corroborate this theory, because they found that high-frequency neural oscillations in some brain areas were temporally correlated with beat perception (Fujioka, Trainor, Large, & Ross, 2009; Iversen, Repp, & Patel, 2009; Snyder & Large, 2005). Other research also showed that attending to a stimulus resulted in perceptual facilitation, because this stimulus was expected (e.g., Barnes & Jones, 2000; Jones & McAuley, 2005; Jones, Moynihan, Mackenzie, & Puente, 2002). The entrainment of the neural oscillations with an external rhythm may be the brain’s way to provide a prediction model for the events to come. Such a way of modelling the future may be specific to auditory rhythm perception, but perhaps generalizes to tactile and visual rhythm perception as well, although these other types of rhythmic entrainment appear to be much more difficult to achieve (e.g., Grahn, 2012).
The Action Simulation for Auditory Prediction (ASAP) hypothesis was formulated in response to DAT (Patel & Iversen, 2014). It argues that the mechanism of simply entraining
neural oscillations to an external rhythmic stimulus is not sufficient in explaining beat perception. The ASAP hypothesis aims to incorporate action into DAT, because of the role of motor processing in rhythm perception, as was described in the previous section. Patel and Iversen claim that this auditory-motor interaction lies at the basis of beat perception, in that beat prediction involves a simulation of action in neural motor areas which bidirectionally interact with auditory brain areas. As such, the ASAP hypothesis on beat perception is a demonstration of Clark’s action-oriented processing, where action and perception interact to form predictions about upcoming events. However, the ASAP hypothesis gives no explanation on the enjoyment of slight prediction error. Our previous sections showed that syncopation plays an important role in wanting to move to music, as well as a role in the enjoyment of music. In order to understand humans’ ability, motivation, and pleasure to entrain and move along with a beat, one needs to incorporate affective mechanisms into the existing theories.
Trost, Labbé, and Grandjean (2017) recently reviewed the relationship between affect and entrainment. Firstly, they distinguish between different types of entrainment, as was formulated by Trost and Vuilleumier (2013). There it was proposed that entrainment can occur on four levels. On the perceptual level, a listener is able to perceptually entrain to (and thus predict) a rhythmic pattern. On an autonomic physiological level, some bodily functions (e.g., heart rate, breathing), entrain to the beat of the external rhythm. On the motoric level, entrainment occurs by synchronizing bodily movements to the beat. Finally, on a social level, people entrain not only with the external music, but also with other people.
The current review is mostly concerned with perceptual and motoric entrainment, and what lies at the basis of this; neuronal entrainment. Regarding perceptual entrainment, Trost et al. (2017) define pulse clarity, which they define as “the ease with which listeners can pick up an underlying pulse”, or in other words; beat perception. This pulse clarity has been shown to influence different affective measures, such as arousal (Luck et al., 2007), decreased subjective anger (Eerola, Lartillot, & Toiviainen, 2009), and pleasure (Trost, Frühholz, Cochrane, Cojan, & Vuilleumier, 2015). Note however, that this clarity is an indication of ease, and thus suggests a linear relation between the ease to entrain and emotions, instead of an inverted U-curve.
Motoric entrainment has also been shown to induce pleasure (Janata et al., 2012). Such an effect has even been shown with infants, who displayed more signals of positive affect (i.e., smiling), when they were coordinated to move to a rhythmic stimulus (Zentner & Eerola, 2010). Another example of pleasure in musical entrainment is a somewhat more uncommon state, namely trance. Trance is generally described as a pleasurable experience and is typically induced by dancing to repetitive rhythmical sounds (Szabo, 2006, as cited by Trost et al., 2017). The relation between entrainment and pleasure is supported by neurological findings. As was already described in the previous section, rhythm and motor processing are related to activation in the basal ganglia (Grahn, 2009; Grahn & Brett, 2007; Grahn & Rowe, 2009; Schubotz et al., 2000), which are in close connection to the limbic
system important for affective processing, albeit in separate circuits. The degree of interconnectedness of these circuits has been a point of contention, which is not entirely resolved (Draganski et al., 2008; Joel & Weiner, 1994).
In summary, in this section of the review the relationship between rhythm and motor processing was described. Studying this relationship can provide insights into how and why people entrain to the beat. DAT and the ASAP hypothesis aim to explain beat processing with entrainment as a prediction model, on a neural, perceptual, and motoric level. Especially the ASAP hypothesis shows similarities with Clark’s (2013) action-oriented predictive processing. Note that the studies reviewed by Trost et al. (2017) did not show any evidence of an inverted U-curve relationship between entrainment and enjoyment. Entrainment, however, may not be a continuous variable, but rather a state that one is in or not. In the next section, we will integrate the findings described in all previous section to describe the interaction between rhythmic complexity and enjoyment, including some additional insights which may further shed light on the interaction.
6. Integration and Novel Insights 6.1 So Far: Predictive Processing in Music
The mechanisms through which music elicits emotions are multi-faceted, relying on musical factors, listener factors, and environmental factors. Expectation, however, is consistently mentioned as a mechanism that is related to eliciting emotions (BRECVEMA; Juslin, 2013; Juslin & Västfjäll, 2008; ITPRA; Huron, 2006), and inducing pleasure or reward. Expectation, or the ability to form predictions about what is to come, is also studied in cognitive sciences as a whole (Clark, 2013). The theory of the predictive brain states that in a hierarchical and cascading fashion, the brain continuously makes predictions and relays back the prediction error to make better predictions in the future. As a theory, it unifies research on perception, action, and attention into one model. Evidence for this type of predictive processing is found indirectly in behavioral studies that show the Bayesian rules of prediction, and computational models that use Bayesian statistics to simulate human behavior.
Additional evidence for predictive processing can be found in music. People form expectations of harmonic progressions, as shown with ERP studies. In musical rhythm, people form expectations about the meter, i.e., a hierarchical layer of isochronous beats superimposed over the perceived rhythm. This type of prediction appears to happen pre-attentively and is found in all people, regardless of their musical expertise. When it comes to prediction error, or the violation of expectations, there is a difference between what is found in music research and what is claimed by the theory of the predictive brain. That is, according to Clark (2013), the end goal of the predictive brain is zero error. However, in music some evidence shows that people prefer ‘slight error’ compared to zero or large amounts of error. This is depicted in Meyer’s inverted U-curve describing the relationship between complexity and pleasure. Complexity can be defined on many different levels.
Research findings on harmonic complexity and pleasure or emotion are mixed, showing either a dislike of expectancy violations, or, contrastingly, positively valenced musical chills in response to violations. When it comes to rhythmic complexity, complexity either arises because expectations are violated (e.g., syncopation), or expectations cannot be formed (e.g., complex meter ratios). Several studies based on syncopation indicate that a certain optimal amount is preferred over no syncopation, thus indicating an inverted U-curve relationship.
This inverted U-curve relationship also appears when incorporating bodily movements into the equation. Especially with regards to rhythm, movement is an obvious partner to music. A large body of research has shown the overlap between motor and rhythm processing (see section 5.1). The amount of syncopation showed an inverted U-curve relationship with groove, i.e., the extent to which the music makes its listener want to move. DAT and the ASAP hypothesis (Jones & Boltz, 1989; Large & Jones, 1999; Patel & Iversen, 2014) show that in music too, perception, action, and attention can be unified into one model that describes how the brain makes prediction models through entrainment (neural, perceptual, and motoric), which are tested and updated in a bidirectional manner. Entrainment is something that brings pleasure to the listener, but how and why violation of expectations might induce pleasure is not yet explained.
Some theories on musical emotions and aesthetic pleasure (Brattico et al., 2013; Juslin, 2013; Juslin & Västfjäll, 2008) claim that it is the aesthetic situation of music that makes the feelings it induces of another nature than feelings in non-aesthetic daily life situations. This may explain, e.g., the enjoyment of sad music or the preference for slight violation of expectations, but perhaps musical emotion and pleasure can also be explained without this extra mechanism. This was also suggested by Flaig and Large (2013) who proposed to not introduce an extra mechanism (i.e., aesthetic appraisal) to explain musical emotion, but rather that music has a special way of engaging with the brain. The rest of this section will introduce new insights that may shed further light on the interaction between complexity and enjoyment.
6.2 Affective Monitoring Hypothesis
The affective monitoring hypothesis (Phaf & Rotteveel, 2012) has its foundation in information processing and how people react to conflict in their environment. From an evolutionary perspective, it is beneficial to react appropriately to, or constantly monitor, the environment. The hypothesis describes three different situations: (1) no conflict, which would not attract attention and thus not lead to an affect response; (2) resolved conflict, which attracts attention at first, and then elicits positive affect because the conflict has been resolved; (3) unresolved conflict, which also attracts attention, but leads to negative affect because the conflict remains. Unresolved conflict, as the authors explain, leads to negative emotions because it acts as a challenge and possible threat to survival. An extreme example of positive affect in response to resolved conflict is laughing at a humorous joke. Jokes usually grab the attention by posing a conflict, but then resolve it. A joke with no
