There is an extensive literature on music and emotions (Juslin and Sloboda, 2001; Swaminathan and Schellenberg, 2015; Juslin, 2019), and contemporary researchers in this subfield rely heavily on mainstream models of affect and emotion. However, there appears to be little reciprocal influence of music research on general theories of emotion, which is perhaps unfortunate. Hevner (1936) reported a circular arrangement of emotion terms to be used in rating musical samples; its vertical axis was marked by merry versus mournful, and its horizontal axis contrasted vigorous with serene. This scheme appears to have been ignored until, half a century later, Russell (1980) offered his circumplex model of valance and arousal.

As Hunter and Schellenberg (2010) pointed out, much research on music and emotions has quantified the latter using Russell’s (1980) two dimensions. Although it is undeniable that valence and arousal are important characteristics of affect, there is a certain irony in the reduction of the powerful, subtle, and ever-changing expression of affect in music to two-dimensional ratings—it is as if one tried to analyze Van Gogh’s Starry Night with a color wheel. Somewhat more elaborate models based on sets of discrete emotions have also been used (Eerola and Vuoskoski, 2011, 2013), but a simple categorical judgment of the emotion portrayed in a passage of music cannot capture the subtleties of the flow of music. Surely a more dynamic theory of emotions is needed, and a theory that can adequately describe the feelings portrayed in music should be of value in understanding all emotional experience. If, as Pratt (1952) argued, “music sounds the way emotions feel” (p. 26), then emotions must feel the way music sounds, and musical analysis must offer insights into affective experience.

Although the psychology of music has had a limited influence on general theories of emotion, music itself has played a major role in research on affective neuroscience. In these studies, neuroimaging and electrophysiological techniques are used to identify areas of the brain that are activated by emotions. Music has frequently been used to induce these emotions (e.g., Blood and Zatorre, 2001; for reviews, see Habibi and Damasio, 2014; Koelsch, 2014), and effects have been documented that depend in part on features of the music (e.g., consonant vs. dissonant, happy vs. sad) and in part on characteristics of the listener (e.g., musical training, music preferences). Koelsch (2013) has pointed to several advantages of using music as a stimulus in this research, including the fact that it can be used to study the time course of affective processes over short (seconds) and longer (minutes) intervals.

There is controversy, however, over whether the feelings induced by music are genuine emotions—and thus whether they are informative about how the brain processes emotion in daily life. Konečni, 2008, (p. 115) argued that, at best, “music may induce low-grade basic emotions through mediators” such as personal associations (e.g., of a song with the loss of a loved one). Koelsch (2014) countered by citing evidence that music affects endocrine responses, facial expression, and action tendencies as well as subjective feeling, and thus “music can evoke real emotions” (p. 178).

However, the affective experience derived from music surely differs somehow from that of everyday life, which is thought to be generated by appraisals of the valence and personal relevance of events (Ellsworth and Scherer, 2003). The distinction between what is felt when listening to music and what is felt in real life is underlined by the curious fact that people often enjoy sad, even tragic music (Sachs et al., 2015); people do not enjoy tragedy when it befalls them. The only class of emotions that music clearly generates are the esthetic emotions such as awe and esthetic chills (McCrae, 2007; Konečni, 2008; Sherer and Zentner, 2008). Here the music is appraised as being beautiful or moving, and the listener shows a distinctive emotional response. Reybrouck and Eerola (2017) pointed out that the esthetic experience has perceptual, cognitive, and affective aspects, and that esthetic emotions are “more contemplative, reflected and nuanced” (p. 9) than core affects and basic emotions. One may be in awe of a Bach fugue, but only if one has acquired knowledge of fugal techniques and become intimately familiar with the music.

Some exceptions should be noted. Watson (2000) has reported an extensive program of research on mood, including daily, weekly, and seasonal variation. In a measure of emotional intelligence, Mayer et al. (2000) included subscales that assessed the ability to predict sequences of emotional events: progressions (as from anger to rage) and transitions (as from fear to tranquility through acceptance). Appraisal theories (e.g., Lazarus, 1966; Ellsworth and Scherer, 2003; Moors et al., 2013) see emotion episodes as unfolding processes, with appraisal of an event and of our ability to deal with it leading to an emotional state that affects attention, memory, and physiological arousal in preparation for a behavioral response. But those theories often imply that the emotion feeling itself is a static state, neither growing nor waning in intensity, nor altering in character. If emotions were truly like this, it is not music but painting or sculpture—or photographs of faces—that would be the language of emotion.

Following a series of pleas for research on temporal sequences in affect (Frijda, 2007; Ebner-Priemer and Trull, 2009; Larsen et al., 2009), researchers have begun to analyze affective patterns over time (Kuppens, 2015). The major tools used in these studies are recollections collected at the end of a day or week, or real-time ambulatory assessments usually involving telephone prompts and responses. Typically, affect dynamic studies assess positive affect (PA) and negative affect (NA) repeatedly over the course of a day or week; data are summarized in terms of theoretically-based statistical indices, such as intrindividual variability; and the indices are used to predict psychological well-being (Houben et al., 2015) or psychiatric diagnoses (Santangelo et al., 2016).

Unfortunately, results to date from these studies have been meager. Dejonckheere et al. (2019) reanalyzed a series of studies and showed that “all affect dynamic measures add little to the prediction of psychological well-being once the explanatory power of mean levels of PA and NA is taken into account” (p. 483). They concluded that “conventional emotion research is currently unable to demonstrate independent relations between affect dynamics and psychological well-being” (p 478). It is understandable that contemporary emotion researchers would turn to statistical indices to quantify emotion dynamics, but at this stage of research it may be more useful to describe emotion patterns qualitatively, and for this music may offer some hints to a way forward. Some steps have been taken in studies of prosody, in Clynes’s (1977) notion of sentics, and in Frijda’s (2007) discussion of emotion episodes.

One area of emotion research that has explicitly drawn upon music is the study of emotional prosody (Quinto et al., 2013; Weniger et al., 2013). Prosody refers to the acoustic features of speech—pitch, loudness, rate, and so on—that are determined in part by semantics (as in the rising pitch at the end of a question) and in part by the emotion the utterance expresses (as in the harsh timbre and clipped words of a verbal threat). Prosody has been widely studied, having applications in fields ranging from speech pathology to human/robot interactions. Students of prosody have often used research methods that parallel those used in studies of music (e.g., Coutinho and Dibben, 2013). Results show that the acoustic cues that express a particular emotion in music generally express the same emotion in speech (Weniger et al., 2013). This phenomenon has long been recognized, and theorists from Spencer (1857) to Juslin (2013) have argued that musical expression mimics emotional prosody, which in turn reflects the underlying physiology of emotion: Slow music sounds sad because it resembles the slow speech caused by psychomotor retardation of depressed people.

Manfred Clynes was a concert pianist whose colleagues included Yehudi Menuhin and Pablo Casals; an engineer, inventor, and neologist who coined the term “cyborg;” and a neuropsychologist who co-edited a volume on emotions with Jaak Panksepp (Clynes and Panksepp, 1988). But his theory of sentics—his term for emotion—and the associated therapy, sentic cycles (Clynes, 1977), are far from mainstream. Only a handful of independent replications of his work have been reported, with mixed results (Nettelbeck et al., 1989; Hama and Tsuda, 1990). Some of his ideas seem relevant only to a rarified group of musicians or artists, and some of his claims strain credulity—practicing sentic cycles is said to cure anxiety, prevent suicide, and increase energy and creativity.

Nevertheless, his ideas merit consideration. Calling on great musicians as expert witnesses in the communication of emotion seems like a useful strategy, at least for the generation of hypotheses. Clynes’s (1992) focus on the temporal shape of emotions is surely appropriate. And his basic findings have a certain plausibility.

Sentics, from the Latin sentire, to feel, is Clynes’s term for emotion. He argued that the experience of emotion is inevitably tied to its expression or perception, so the natural unit of emotion has the duration of a single expression—a smile, a shout, a sigh. Longer emotion episodes are compounded of repeated instances. Clynes focused on basic emotions, and argued that each has a distinctive temporal form, regardless of the modality by which it is expressed (tone of voice, facial expression, musical phrase).

His procedure for detecting that form employs a sentograph, a pressure gauge on which participants rest their middle finger; the finger remains on the finger rest throughout the procedure. The pressure gauge is wired to a computer that records data and produces graphs of pressure-by-time curves. Participants are instructed to express a given emotion on a signal by pressing their finger once. (Recall that Clynes was a pianist, for whom the pressure of fingers is virtually the only way to express musical emotion). Signals are given by the experimenter at random intervals of a few seconds, and pressure is recorded both vertically (up vs. down) and horizontally (i.e., inward vs. outward). Each response is called an E-acton. When averaged over 50 trials, characteristic vertical and horizontal curves are obtained, distinct for each emotion, but generalizable across persons and cultures (at least in Clynes’s own research).

Consider three examples. Over a period of 2 s, an individual expressing anger gives a quick, sharp press down and outward that is quickly released. It is hardly a stretch to call this a jab. In contrast, an individual expressing love gives a slow, gentle touch that fades away; there is a slight motion inward. Joy is seen in a quick downward press that rebounds upward—a jump for joy. These are intuitively reasonable results, comparable to the findings that anger is expressed musically by loud and staccato sounds, and sadness by low and slow melodies (Gabrielsson and Lindström, 2001). What Clynes added was a specific temporal sequence and duration.

Many students of prosody have datasets in which the acoustic properties of utterances are recorded over time, and the duration of a typical sentence or pseudo-utterance is comparable to the duration of an E-acton. Clynes’ curves thus provide specific hypotheses that might be tested with available data. Do the pitch, or loudness, or rate contours of emotional sentences match his predictions?

In a sentic cycle, an individual seated at a sentograph is prompted to express (with finger pressure) a series of emotions in a fixed order: anger, hate, grief, love, sexual desire, joy, and reverence (Clynes, 1977). Multiple trials are given for each emotion before moving to the next. About 4 min is allotted to each emotion, so a single session requires about one-half hour. After 4 min, the felt emotion begins to fade—a phenomenon Clynes (1988) attributed to saturation of its neurohormonal substrate. Participants find it easy to switch to a new emotion, and any emotion can succeed any other, but with some carry-over effects: “Each state appears to cast its shadow on the following ones” (Clynes, 1977, p. 147).

In his chapter on “Time,” Frijda (2007) made a great deal of a dissertation by Sonnemans (1991), because it was a rare and early source of information on the duration of emotions. Sonnemans asked informants to recall an emotion they had had the previous week and to describe what caused it, how they felt, and how the experience unfolded over time; informants created a graph of feeling intensity over time. Many of these accounts covered hours or days; many included more than one feeling intensity peak; and many mentioned a number of different emotions, simultaneously or successively, as part of the experience. For example, a woman reported on her response to a demeaning catcall: Over the course of 11 h she felt disgusted, angry, and humiliated and had disturbing dreams; the next morning, however, she thought the harasser had shown himself ridiculous and laughed the incident off.

Frijda (2007) called such experiences emotion episodes, which is perhaps an unfortunate word choice. Other researchers (e.g., Konečni, 2008) use the term “episode” quite differently, to refer to a single occurrence of a basic emotion, many of which might be included in each of the experiences Sonnemans (1991) analyzed. In literature, an “episode” is typically one of a series of distinct but coordinate units, such as chapters—but Frijda’s emotion episodes presumably are often isolated incidents, not necessarily continuous in time.

For musicians, the lexical ambiguity is even more awkward: A musical episode in a fugue is an incidental diversion from the ongoing development of the theme, whereas Frijda’s concern is precisely with the developing emotional experience. From a musical perspective, what Frijda described is in fact the analog of a theme, a recognizable tune that appears and disappears over the course of development, that (particularly in Romantic music) is emotionally expressive, and that often evolves, changing from soft to loud or from major to minor in successive statements. Frijda’s phenomenon is a series of emotions bound together by a common theme (e.g., reaction to a catcall).

However, the word theme is at least as ambiguous as episode, so I suggest this phenomenon be called an emotion thread. A plot thread is the narrative parallel to a musical theme: a distinct subplot that can appear, disappear, and reappear; evolve in tone or significance; and interact with other plot threads. When people are asked to recall emotional experiences, Sonnemans’s (1991) findings suggest that they organize them by emotion threads. If so, this poses yet another challenge to ambulatory assessment: Perhaps researchers need to sample feelings when, and only when, a particular thread is activated—which may be difficult to predict.

As long ago as 1929, Köhler proposed that psychologists adopt musical terminology: “the inner processes, whether emotional or intellectual, show types of development which may be given names such as crescendo and diminuendo, accelerando and ritardando” (Köhler, 1929, pp. 248–249, quoted in Langer, 1957). Such developments correspond to what Mayer et al. (2000) called progressions in emotion and to Kuppens and Verduyn (2017) intensity profile shape. Köhler’s notion has been extended by Stern (1985) under the name “vitality affects” or “vitality contours” (Stern, 1999).

Stern pointed out that emotions have a temporal shape that can be described by such terms as surging, fading away, fleeting, and explosive. He considered vitality affects to be entirely distinct from basic emotions, capturing the manner in which experience was felt rather than the content: Joy, anger, fear, or excitement may all fade away.

As Køppe et al. (2008) complained, in the course of his writings Stern used a variety of terms for vitality affects and extended their scope beyond affect to include temporal patterns of behavior, cognition, and motivation. In his last work on the topic, Stern (2010) called them “dynamic forms of vitality,” and defined them as a Gestalt of “movement, time, force, space, and intention/directionality” (p. 4) that gives the feeling of being alive. For the purposes of this article, a more limited definition is useful. I will use affect contours to refer to “the continual shifts in arousal, activation, and hedonics occurring split-second-by-split-second” (Stern, 1999, p. 70). Note that these are actually not new affects, but descriptions of the temporal course of the familiar dimensions of affect.

Stern, a developmental psychoanalyst who was chiefly concerned with the details of interactions between infant and parent, believed that these experiences have a very brief duration, “rarely over 5 s” (Stern, 1999, p. 68). If so, an extended expression of emotion, such as a full movement from a symphony might show, would need to be construed as a chain of affect contours.

Consider a musical example. The emotional tone of the Andante “teneramente, molto cantabile, con espressione” in Tchaikovsky’s Pathétique Symphony (Figure 1) is shaped by many musical features: the pure diatonic melody over a chromatic harmony, the major mode, the timbres of horns and muted strings. But the passage is brought to life—given vitality—by musical versions of affect contours. Beginning piano, the volume swells and sinks gently in the second measure and again in the fourth, creating a wistful, sighing effect. The fifth measure (incalzando, i.e., urgent, pressing) shows a rapid increase in volume and tempo, a surge of passion and longing that dissipates, at least in tempo, in the seventh measure. The seventh and eighth measures repeat the melody of the third and fourth, but now, played forte, they have a noble strength and dignity.

Although both Clynes and Stern were concerned with temporal forms over a brief interval, they differ fundamentally on the fixity of form. For Clynes, anger is always expressed by the equivalent of a jab, although the intensity of the jab may vary; for Stern, anger might be smoldering or explosive or undulating. Affect contours do not define basic affects, they merely shape them.

The flow of music is structured by expectations and outcomes (Meyer, 1956, 2001; Narmour, 1991; Huron, 2008), which Sloboda and Juslin (2001) call intrinsic sources of emotion. Perhaps the most basic of these (at least to Western ears) is the chord progression from consonance to dissonance and back to consonance (e.g., I—V⁷—I), but any musical component (rhythm, melodic interval, harmonic density) that generates an expectation of some resolution creates (and can express) tension. The time scale may be quite extensive: The entire development section of a classical symphony creates a tension that is resolved by the return of the original theme and key in the recapitulation at the end of the movement.

A number of musical devices can be used to generate suspense. Figure 2 shows a passage from the Allegro non-troppo of the Pathétique’s first movement. Here a small motif is repeated in successively higher pitches; key modulations introduce uncertainty; phrasing is progressively compressed; and the volume increases dramatically in the last two measures before attaining the climatic F-sharp and a subsequent reduction in tension. The passage might be described as showing an increase in arousal—which it does—but there is also a sense of urgent seeking that is more than mere arousal.

The psychological basis of Meyer’s (1956) and Narmour’s (2000) theories of musical expectation is essentially perceptual-cognitive; a computer program could specify the musical expectations at any given moment in a musical passage. But from an affective perspective, these expectations can be understood as volitions: Listeners not only expect a dominant seventh chord to be resolved into the tonic; they want it to be resolved. It is this tension—the volitional pushes and pulls built into a composition—that holds the listener’s attention. The affective equivalents of musical expectation are expressed in such words as longing, anticipation, suspense, and dread, which might be termed volitional affects. Cowen et al. (2019, Figure 2) showed a cluster of emotions including yearning, eagerness, and interest that can be interpreted as positive volitional affects, and another group characterized by detachment, boredom, and fatigue that might be seen as negative volitional affects. (It is likely that negative volitional affects play a much larger role in human life than in music, because they would appear to have limited esthetic appeal).

It is rather puzzling that suspense has not been recognized as a basic emotion, like fear and anger; surely it is a common and powerful feeling. Under the guise of tension, it has occasionally appeared in dimensional models. Wundt (1902) included tension/relaxation (Spannung/Lösung) along with dimensions resembling valence and arousal, but most subsequent models—notably Russell’s (1980) circumplex—have omitted it. Because much research on music and emotion has been informed by Russell’s model, a crucial element of musical expression seems to have been missed. Eerola and Vuoskoski (2011) tried to correct this oversight by proposing that three dimensions (valence, energy arousal, and tension arousal) are needed to describe the emotion expressed in music, and one might suppose that tension arousal is related to the degree of expectation generated by the music. There is, however, a possible confusion here, because tension can mean either suspense (which can be resolved) or harsh unpleasantness (which can only be endured or ended). Eerola and Vuoskoski asked participants to rate excerpts from film scores and found that ratings of tension were strongly negatively correlated with valence (r = –0.83), suggesting that tension was interpreted by the raters to mean music that was ugly and jarring, suitable perhaps for horror films—indeed, a selection from the score of Hellraiser was one of those chosen to represent high tension. Compositions certainly do vary in the amount of musical suspense they generate, but this is presumably unrelated to valence; it is therefore essential to distinguish tension-as-distress from tension-as-suspense.

Suspense has been treated by appraisal theorists as a combination of hope and fear aroused by cognitive uncertainty (Ortony et al., 1988), and Madrigal and Bee (2005) provided evidence that both hope and fear occur and fluctuate when viewing a television advertisement deemed to be suspenseful. This would seem to suggest that suspense can be decomposed into the feeling dimensions of valence (hope vs. fear) and arousal (strong vs. weak) and the cognitive appraisal of outcome uncertainty. But in music, and presumably in life, there is also an affective component of suspense—something that feels like holding one’s breath—and it can be felt without either hope or fear. We sometimes want to know how something will turn out even when it is of no consequence to us.

Goal-directed striving is central to Carver and Scheier (1998) and Carver’s (2015) account of emotion, but the role of volition in that model is quite different from what is suggested here. Carver proposed that people monitor their progress toward goals (or away from adversity), and that the rate of progress is signaled by emotion: “Feelings with a positive valence mean that you are doing better at something than you need to, and feelings with a negative valence mean you are doing worse” (Carver, 2015, p. 302). Suspense is completely absent from this model, because it is precisely the feeling that accompanies uncertainty about whether (or when, or in what degree) you are doing better or worse.

The premise of this article—that music can offer a guide to understanding affect—is perfectly illustrated by the work of Huron (2008), who stated that “music provided me with a serendipitous starting place for theorizing more generally about the psychology of expectation” (p. viii). Huron was concerned with cognitive and physiological concomitants of expectation, but also centrally with emotion. He proposed a model in which five responses—imagination, tension, prediction, reaction, and appraisal (ITPRA)—account for expectation effects. Imagination consists of envisioning the consequences of a behavior and foretasting the pain or pleasure it would likely lead to; tension is an anticipatory arousal that is felt as suspense. After the event occurs, people feel a degree of satisfaction if their prediction was correct; this corresponds to the esthetic satisfaction that Meyer (1956) described when listeners correctly anticipate a musical event. The immediate, reflexive evaluation of an outcome Huron called the reaction response; its counterpart is the pleasant surprise that sabotaged musical expectations provide. The final affective state results from a conscious, reflective assessment of the event; this is the appraisal response, corresponding perhaps to the perceived overall affective tone of a piece of music. Huron’s (2008) book provides a detailed exposition of his model, with extensive musical examples.

Music of a certain degree of complexity does not resolve all expectations in the obvious way. Instead, composers maintain the listener’s interest by changing the musical direction (Narmour, 1991) through key modulations, thematic variations, change in mode, and so on. Musical transitions, whether rapid or gradual, must be meaningful. For example, modulation—moving from one key (or tonality) to another—usually requires an intervening series of chords, each of which is compatible with the preceding and following chord, thus forming a chain that links the two keys. Clearly, such musical evolutions are akin to, and might express, what Mayer et al. (2000) called emotional transitions. If I have correctly interpreted the passage in Figure 1, it shows a transition from wistfulness through self-assertion to dignity.

The tempestuous Allegro non-troppo of the Pathétique’s first movement segues into the Andante shown in Figure 1 through sixteen bars of progressive slowing, quieting, and thinning of the orchestral texture (Musopen Orchestra, Supplementary Audio 3)—a transition based on affect contours. After the second statement of the Andante, the theme dies away to a passage marked pppppp (impossibly soft) for a solo bassoon, followed immediately by a fortissimo chord in the full orchestra (leading back to the “feroce” Allegro non-troppo). This too is a legitimate transition, in part because the soft and loud chords (D-major and C-minor sixth, respectively) form a meaningful, if unusual, harmonic progression, but it is an abrupt change that presents a dramatic surprise to the listener.

Surprise was considered by Ekman and Friesen (1971) to be a basic emotion. But Watson (2000) noted that surprise differs from other basic emotions in that it does not appear to correspond to a trait affect. People can be anger-prone or laughter-prone, but it appears that no one is surprise-prone. If asked, people can rate how often they experience surprise, but these ratings show little temporal stability, cross-observer agreement, or correlation with basic personality factors (Watson, 2000). Surprise appears to be not a basic emotion rooted in characteristics of the individual, but an affect transition inherent in the emotional experience itself (cf., Ortony, 2021).

Transitional affects are closely tied to emotion threads. The surprise fortissimo pivots from the Andante theme to the Allegro non-troppo theme—a transition across emotion threads. Figure 1 shows affect transitions within a theme: It is the continuity of the thread that allows us to perceive a set of different feelings (e.g., wistfulness, self-assertion, dignity) as different stages of the same emotional experience.

Appraisal theorists are likely to attribute affect transitions to reappraisals. In Frijda (2007) example of a woman whose feelings in response to a catcall were transformed from anger to amusement, the affective change was paralleled by a cognitive reappraisal: The harasser, who had first been seen as malicious, was reappraised as merely ridiculous. But the causal direction is not obvious here. It is entirely possible that the feelings of anger and disgust exhausted themselves in the course of a few hours through processes of affective adaptation, and the new attribution was merely a rationalization for the changed feelings. Indeed, the unmotivated affect expressed in music is problematic for appraisal theorists: Ellsworth and Scherer (2003) admitted that “neither appraisal theory nor any other current emotion theory can easily accommodate emotional responses to music” (p. 588).

Perhaps the most basic question is the duration of affects. Clynes (1977) and Stern (1999) argued that the natural unit of emotions is a few seconds, and Pöppel (1997) has argued that experience is integrated by the central nervous system into a subjective present of about 3 s. Emotion research on this micro time scale seems to be indicated. But emotion episodes are surely longer, and probably vary by type. Scherer and Wallbott (1994) found that fear, disgust, and shame tended to be short-lived, whereas joy and sadness could persist for hours or days. Watson (2000) studies of mood across hours and days suggests that, for most people, positive affect is a chronic state normally distributed around a personal mean, whereas negative affects are sporadic and thus show a skewed distribution.

For most affects, duration can be defined and assessed as the period from inception until “the intensity of the emotional response returns to zero or to a baseline level” (Verduyn et al., 2015, p. 331). But anticipation, dread, and suspense—volitional affects—potentially last from the recognition that an uncertain outcome is looming until it occurs, if ever. The outcome—a medical test result, a job offer—may be days or weeks away. Is the volitional affect continuous over that interval, perhaps backgrounded in consciousness, or does it recur when prompted by relevant cues, or surface spontaneously like intrusive post-traumatic thoughts?

What is the typical course of an emotional experience? Emotions may grow in intensity, as when one works oneself up from annoyance to rage. Some affects—or their foregrounding in consciousness—may wax and wane periodically. Others may have an acute phase—say, the terror that a near-miss accident generates—and recur later as echoed apprehension. These patterns might be called the vicissitudes of affect.

But affects may also change qualitatively, morphing into new feelings. Reisch et al. (2008) used quarter-hourly assessment of borderline personality disorder patients to examine the sequence of specific emotions; they noted that compared to controls, the patients more frequently “switched from anxiety to sadness, from anxiety to anger, and from sadness to anxiety” (p. 42). Volitional affects are presumably resolved by the dreaded or desired outcome and succeeded by the appropriate basic emotion—despair or joy. Do positive and negative anticipations (hopes and fears) alternate, or does one predominate (see Madrigal and Bee, 2005)? Mayer et al. (2000) considered natural affective sequences as emotional transitions, and identified enough examples to construct a scale—though more extensive and systematic research on the nature and frequency of such spontaneous transitions is needed.

Emotional transitions in real life are complicated and ambivalent, because experienced affect is not monophonic, a simple sequence of successive feelings like notes in a melody. It is instead polyphonic, or homophonic, or bitonal, or cacophonic—that is, it is a continuously varying mixture of different feelings that may reinforce, contextualize, or clash with each other (cf., Larsen and Stastny, 2011). The trajectory of any single mood in this ongoing mix is surely affected by the other felt moods, which themselves may vary in their usual intrinsic duration. If Clynes (1977) was right, and “each state appears to cast its shadow on the following ones,” then the intricacies of affective patterns will be baffling.

Hollenstein (2015), however, has proposed that human affect is a complex system characterized by attractors. These are particular configurations of affects that tend to endure or recur regardless of the initial starting point—a phenomenon known as equifinality. The identification of such attractors—ideally in an affective space far more differentiated than the usual two-dimensional model—ought to be an aspiration of students of emotion.

One set of causes for affective attractors has already been identified: personality traits. Personality traits have long been linked to the frequency and intensity of affects (e.g., Costa and McCrae, 1980; Watson and Clark, 1992). There is some evidence that the durations of positive and negative affects (like their frequency and intensity) are linked to Extraversion and Neuroticism, respectively (Verduyn et al., 2015). Penner et al. (1994) found that there were consistent individual differences in the degree of variability of 11 affects, and these variances were themselves intercorrelated. Some people thus show greater emotional lability then others, and in extreme cases this may contribute to Borderline Personality Disorder (Ebner-Priemer et al., 2015).

The fact that music seems designed for the expression of emotion does not imply that every musical utterance or style will be a perfect embodiment of emotion feelings. Humans have painted since they decorated Lascaux, but they did not master linear perspective—an accurate representation of the visual field—until the Renaissance. In the same way, there has been progress in the accurate musical expression of emotion, from simple keening to the intricate, extended, and nuanced portrayals of human passion in works like Wagner’s Tristan and Isolde.

Crocker (1966) traced the development of Western music from monophonic chant to the polyphonic motets and masses of the late Middle Ages. A particularly notable innovation of the 15th century was the shift from the open and neutral harmonies of fourths, fifths, and octaves to the emphasis on third and sixths, in both major (happy) and minor (unhappy) forms. Here at last was a musical structure that reflected the basic dimension of valence. Indeed, as Eerola et al. (2013) have shown, major or minor mode is the most important cue in the identification of the emotion expressed by a musical passage. If and when we have a full understanding of human affect, we will better be able to grasp the historical development of musical styles.

The evolution of music, like that of language, is a relatively recent phenomenon. The evolution of emotion is surely far more ancient, and one might surmise that the lessons offered here are of little relevance to an understanding of how affective systems arose. But the central argument advanced is that emotion is a temporal phenomenon, and its dynamics can and should be studied through cross-species comparisons.

Emotional reactions have durations in all species. The piloerection of a startled cat fades away, but how rapidly, with what contour, and with what adaptive significance? A dog awaiting the return of its owner surely feels the volitional effect of anticipation. Elated by the owner’s appearance, it eventually calms down; is this, as Clynes (1988) supposed, because of neurohormonal saturation? Certainly there are affect transitions in non-human animals: quarrels break out in social groups and are somehow resolved; infant distress is soothed by a parent; apathy turns to agitation when a threat appears. Are there commonalities in these processes across species, and if so, how far back in the evolutionary chain are they shared? A consideration of temporal sequences adds a new set of variables to the comparative study of emotion.