A functional MRI study of happy and sad affective states induced by classical music

INTRODUCTION

Previous research has shown that music evokes comparable emotional responses across different musical categories and cultures [Peretz and Hebert, 2000; Trehub, 2003]. It has also been argued that affective states induced by music are stronger, longer lasting, and more pervasive than affective states induced through other modalities [Krumhansl, 1997].

However, only a small number of studies have focused on the neural correlates of emotions elicited by music in healthy individuals [Blood and Zatorre, 2001; Blood et al., 1999; Brown et al., 2004; Koelsch et al., 2006; Menon and Levitin, 2005; Peretz et al., 1998]. A prominent finding of these studies has been an increased engagement of the ventral striatum in response to pleasant emotional music [Blood and Zatorre, 2001; Brown et al., 2004; Koelsch et al., 2006; Menon and Levitin, 2005].

The ventral striatum is involved in reward generation, attention, and motivation, whereas the dorsal striatum is thought to be involved in the initiation of movement towards potentially rewarding stimuli in humans and animals [Elliott et al., 2004; Roitman et al., 2005; Schultz, 2000; Tobler et al., 2003]. Dorsal striatal activation has also been found during suppression of motor reactions [Raemaekers et al., 2002]. The mesolimbic dopaminergic system plays an important role in mediating the activation of neurons in this region through nucleus accumbens afferents [Carelli and Wightman, 2004], thus regulating the reinforcing effect of reward stimuli. In primates, neural firing in the ventral and dorsal striatum occurs immediately prior to approach behavior and continues until the reward has been obtained [Berns et al., 2001; Schultz and Romo, 1990]. Activation in these structures has also been observed during mating/copulation in rodents [Pfaus et al., 1995], in humans in association with stimulants such as food [Small et al., 2001], recreational drugs [Breiter et al., 1997] and while viewing pleasant pictures or facial images [Aharon et al., 2001; Lane et al., 1999; Morris et al., 1996, 1998; Phillips et al., 1998].

Another structure of interest in the context of emotion research is the anterior cingulate cortex (ACC) with its afferents from the midline thalamus and brainstem nuclei [Paus, 2001]. Cytoarchitecture and function suggest a division into a dorsal cognitive area and a rostral–ventral affective area [Bush et al., 2000]. Neuroimaging research has reported ACC activation during positive emotions [Morris et al., 1996; Phillips et al., 1998], sad affect [George et al., 1995; Levesque et al., 2003], as well as across both happy and sad affects [Killgore and Yurgelun‐Todd, 2004]. Using musical pieces, Blood and Zatorre [ 2001] found a positive correlation with ACC activation and the intensity of the physiological experience of a highly rewarding sensation with a selection of music. Moreover, Blood et al. [ 1999] found that activity in the ventral portion of the ACC showed an association with decreases in levels of tonal dissonance.

A further region that is frequently reported with positive and negative emotional processing is the amygdala [Phan et al., 2002; Zald, 2003]. Amygdala engagement has been reported in response to fearful [Breiter et al., 1996; Morris et al., 1996; Whalen et al., 1998] and sad [Blair et al., 1999; Goldin et al., 2005; Killgore and Yurgelun‐Todd, 2004; Posse et al., 2003] emotional stimuli. Interestingly, Blood et al. [ 2001] observed decreased cerebral blood flow in the amygdala in relation to increased “chills” induced by highly pleasurable classical music, while Koelsch et al. [ 2006] found greater amygdala activation with unpleasant musical excerpts. Additional structures recruited during processing of emotional music include the hippocampus [Koelsch et al., 2006], parahippocampal gyrus [Blood et al., 1999] and orbitofrontal cortex [Blood and Zatorre, 2001; Blood et al., 1999; Menon and Levitin, 2005].

This body of evidence thus suggests an important role of musical stimuli in studying human emotion processing [Blood and Zatorre, 2001; Blood et al., 1999; Brown et al., 2004; Koelsch et al., 2006; Menon and Levitin, 2005; Peretz et al., 1998]. The present study sought to investigate the neural correlates of happy and sad affective states induced by music using fMRI. Baseline neural activity was measured during presentation of musical pieces of emotionally unevocative, neutral valence. An important methodological consideration of the present study concerns the choice of musical stimuli. As humans actively choose to listen to certain types of music in different settings, we believed that utilizing existing musical pieces may be advantageous in provoking emotional reactions similar to those encountered in natural environments. The present research aimed to first establish a set of musical pieces that can reliably induce affective states in healthy individuals. In a pilot behavioral validation study, musical pieces reliably inducing happy, sad, and neutral affective states were determined. These musical stimuli of happy, sad, and neutral valences were then presented to a separate group of healthy volunteers to investigate the neural correlates of music of an emotive nature, using fMRI.

We hypothesized that music from these emotional categories would evoke neural activation in cortical and subcortical regions associated with processing of relevant emotional stimuli. We expected that the happy musical pieces would be associated with greater activation in the ventral striatum [Blood and Zatorre, 2001; Elliott et al., 2004; Koelsch et al., 2006; Menon and Levitin, 2005; Roitman et al., 2005; Schultz, 2000; Tobler et al., 2003] and anterior cingulate [Blood et al., 1999; Morris et al., 1996; Phillips et al., 1998] and the sad stimuli would elicit greater activity in the amygdala [Blair et al., 1999; Killgore and Yurgelun‐Todd, 2004; Koelsch et al., 2006; Posse et al., 2003], compared to emotionally neutral musical excerpts.

METHODS

RESULTS

DISCUSSION

The present study was designed to investigate the neural basis of modulation of affective states through music by utilizing a set of validated happy, sad, and neutral musical pieces. Behavioral valence ratings provided during the scanning procedure point to a successful induction of happy and sad emotional states. The ratings, however, revealed that order of presentation (happy pieces first, sad pieces first) influenced the emotional experience of subsequent musical pieces. Participants who started with happy musical pieces provided more pronounced subjective ratings of happy and sad music. On the other hand, when sad musical pieces were presented first diminished affect ratings for both happy and sad musical pieces were observed. Both groups were matched for trait measures, such as extraversion, neuroticism and hedonia, factors which may affect subjective perception of emotional state. These findings are thus comparable with previous observations of permeating and lingering effects of exposure to sad stimuli on subsequent processing of happy stimuli, which has been observed in the visual modality [Ekman et al., 1980].

Given the findings in the behavioral data, it was expected that this would be reflected in the fMRI data. However, an analysis controlling for order effects found consistent activation patterns between the two groups in all contrasts of interest. The lack of differences represents a replication of the activation patterns in two independent samples; this finding thus allowed the data to be combined in a conjunction analysis.

As expected, happy emotional music showed greater activation in the ventral striatum relative to neutral music. Recruitment of the ventral striatum has been observed in many forms of positive emotional states [Aharon et al., 2001; Blood and Zatorre, 2001; Brown et al., 2004; Elliott et al., 2004; Koelsch et al., 2006; Menon and Levitin, 2005; Schultz, 2000]. Moreover, animal research [Schultz, 2000] has shown that neurons in the ventral striatum display a higher percentage of reward related responses, compared to those in the nucleus caudate or putamen. The activation found in the ventral striatum could therefore reflect the rewarding effect of music.

Although one could argue that the musical pieces used in this study were very well known and could, therefore, have been conditioned with rewarding experiences, a similar finding of ventral striatum activation was found in a recent PET study with popular musical pieces [Brown et al., 2004] in response to unfamiliar music. This finding would add support to the notion that non‐natural stimulants activate our reward system. This view receives further support by a connectivity study by Menon and Levitin [ 2005] who found a strong correlation between activation in nucleus accumbens and the ventral tegmental area, which is the site of mesolimbic dopamine neuron cell bodies that receives input from brainstem, medial and occipital prefrontal cortices, amygdala and from striatal areas [Carelli and Wightman, 2004; Wise, 2002].

If rewards are essentially “environmental incentives we tend to approach” [Scherer, 2000], music has to be conditioned as a rewarding stimulus in order to recognize its lure. The beginnings of this reward experience could lie in early mother–child communication of prelinguistic children. Developmental research has provided evidence that mothers of different cultural backgrounds, when singing to their babies, use a repertoire of simple pitch contours [Trehub, 2003] and repetitions and that these particular melodies are recognizable as lullabies even when they are from an unknown cultural background [Trehub, 2003]. It could be argued that these early positive experiences with music within a social context have conditioned us to display an approach reaction to happy musical tunes as they signal rewarding social interactions. Furthermore, in evolutionary research the importance of music has been interpreted as a group‐building agent, helpful in mating rituals and social group‐building functions [Darwin, 1871; Miller, 2000]. Thus, ventral striatum activation during happy music may be connected to the anticipation of social reward in a wide variety of settings.

A main effect of happy music was also observed in the dorsal striatum, namely in the caudate. The basal ganglia system, in particular striatum and substantia nigra, are known to be involved in motor function. As the mesolimbic system is closely connected to the dopamine system in these regions, it has been argued that this structure mediates motivational and motor mechanisms [Carelli and Wightman, 2004] of behavior approach and withdrawal. Activations in the caudate and ventral striatum were only observed with happy music. This may reflect the rewarding effect of happy musical pieces and perhaps their higher incentive to move to or move towards them.

However, ventral striatum activation could also be an indication of physiological arousal in response to happy musical pieces, due to higher tempi compared to neutral or sad musical pieces. To address this possible concern, SCR data were collected from a subsample (N = 11). The findings suggest that mean SCR were higher during the emotional valence that was presented first (happy first, sad first). Hence, volunteers' SCR levels rose in response to the novelty of the experimental set‐up rather than in response to the emotional valence of the musical pieces.

Significant activation during happy music was also observed in the rostral pregenual and dorsal ACC. The ACC is implicated in controlling of autonomic arousal, error monitoring, conflict processing, nociception, and appetitive behavior (for reviews please refer to: [Ongur and Price, 2000; Paus, 2001; Vogt, 2005; Vogt et al., 2005]). The ACC holds anatomical connections with the ventral and dorsal prefrontal cortices, parietal cortex, amygdala, striatum, and thalamus [Cavada et al., 2000; Tekin and Cummings, 2002; Vogt and Pandya, 1987; Vogt et al., 1987].

ACC activation in response to enjoyable music has previously been reported [Blood and Zatorre, 2001; Menon and Levitin, 2005]. Moreover, Blood and Zatorre [ 2001] found a positive correlation between ACC activation (BA24&32) and the intensity of “chills” as well as an association between physiological arousal and regional cerebral blood flow in the ACC. Furthermore, increased ACC and ventral striatum activation have been observed during drug administration in cocaine dependent individuals [Breiter et al., 1997] suggesting involvement in reward generation.

A recent study investigating reward in response to high risk behavior suggested that the anatomical links between striatum and ACC are also reflected in the functional connectivity of those areas [Cohen et al., 2005]. Bush et al. [ 2002] suggested an involvement of the ACC in reward based decision processes. Furthermore, a study investigating single‐unit firing in ACC neurons [Shidara and Richmond, 2002] has suggested that ACC activity relates directly to reward expectation, whereas Rogers and colleagues [Rogers et al., 2004] observed pregenual ACC activation during reward‐related decision processes.

In the context of the present findings, it might be that the pregenual ACC activity during listening to pleasant music is a reflection of a reward‐related decision process preceding the actual rating process in the subsequent phase. An alternative explanation could be its involvement in vocalization. The ACC is involved in speech‐related tasks [Paus et al., 1993; Petersen et al., 1988]. Together with the increased activation found in the supplementary motor area the findings may suggest that an analysis and reproduction of the music's melody was initiated. As inspection of Table I shows, happy musical pieces may have been better known, relative to neutral pieces. Their melodies may therefore have been easier to recognize and reproduce.

Moreover, the precuneus activation observed with happy music may have been associated with aspects of vocalization and recognition of musical melody as well as attention. The precuneus has been implicated in general episodic memory processes [Krause et al., 1999], memory retrieval [Cabeza and Nyberg, 2000] as well as mental imagery [Fletcher et al., 1995]. The increased activation of this area in the present context could be a reflection of the attempt of retrieving the episodic memories related to the musical pieces. Interestingly, Platel and colleagues [Platel et al., 2003] reported precuneus activation during a music task probing episodic memory retrieval.

We also observed recruitment of left parahippocampal gyrus activation with happy musical excerpts. The parahippocampal gyrus has been engaged by unpleasant musical stimuli [Blood et al., 1999; Koelsch et al., 2006] as well as during both happy and sad emotional states [Damasio et al., 2000; Jatzko et al., 2006; Lane et al., 1997; Tabert et al., 2001; Winston et al., 2002]. Moreover, previous studies have reported a parahippocampal engagement in processing and retrieval of spatial scenes and navigation (e.g. [Epstein and Kanwisher, 1998; Hayes et al., 2004; Mayes et al., 2004; Rosenbaum et al., 2004]) and processing of highly context‐dependent information [Bar and Aminoff, 2003]. Given this evidence, it seems likely that the parahippocampal gyrus is involved in both the processing of emotional aspects of music as well as information related to the retrieval of spatial scenes. However, further research is needed to support such proposition.

The presentation of sad emotional music relative to neutral music, was associated with recruitment of the right‐sided hippocampus/amygdala region as predicted. Activation has been reported in previous neuroimaging research of emotion in healthy people in association with positive and negative stimulus material and through different channels of sensory processing [Phan et al., 2002; Zald, 2003]. A consistent finding in mood disorders is an engagement of the amygdala in the resting state, as well as during cognitive or affective challenges [Drevets, 2000]. Of particular importance for the present study are findings by Koelsch et al. [ 2006] who found left amygdala activation in response to unpleasant music, which underscores the specific role of this region for negative emotion processing. Increased hippocampus/amygdala activation during sad music processing could be an aspect of appraising and evaluating emotional stimuli, especially in response to sad material [Richardson et al., 2004]. There is evidence of an enhanced amygdala response to emotionally salient stimuli occurring before conscious stimulus appraisal [Phelps, 2004; Whalen et al., 1998]. The lateral nucleus of the amygdala serves as an interface for different sensory modalities and receives auditory projections from a thalamic cell group involved in auditory transmission, i.e. the medial geniculate body [LeDoux, 1992; Phelps and LeDoux, 2005]. Here, rapid processing of emotionally salient acoustic information can be made through amygdala mediation. The amygdala/hippocampus activation observed here could represent enhanced perception and attention assigned to emotionally arousing stimuli in order to determine the appropriate reaction in response to the perceived stimuli. The behavioral ratings would support the argument that volunteers perceived the musical pieces as sad. Previous neuroimaging studies investigating affect and music, however, found hippocampus activation in response to both pleasant [Brown et al., 2004] and unpleasant musical pieces [Koelsch et al., 2006]. Furthermore, Blood and Zatorre [ 2001] reported negative correlations between intensity of chills and activation in the hippocampus/amygdala region.

The hippocampal formation, which also receives input from the mesolimbic dopaminergic system, plays a vital role in episodic memory consolidation and has been found in association with reward processes [Wittmann et al., 2005]. The exact role of hippocampal activation during listening to emotional music deserves further investigation. However, a tentative explanation for the present finding could lie in the specific nature of sad music as well as the interconnections that the hippocampus and amygdala are holding with the mesolimbic dopaminergic system.

Anecdotal evidence and reports from the “strong experience of music project” [Gabrielsson, 2002] describe sad music often as extremely beautiful and capable of inducing euphoric reactions. It could therefore be that the neural network underlying sad emotions induced through music is different from the neural activation seen in sadness induced through other sensory modalities or experienced in depression.

An observation that integrates both happy and sad emotional states is the left sided activation during happy musical stimuli in the ventral and dorsal striatum and the right sided activation observed during sad emotional stimuli in hippocampus/amygdala, relative to neutral music.

However, when neural activity during sad and happy emotional states were compared only the superior temporal gyrus (happy greater than sad) showed a significant increase in activation. This finding is somewhat puzzling since sad and musical pieces differ not only in terms of affective valence, but also in terms of structural features (tempo, mode). It is not clear at present why this structure responded more strongly to happy than sad stimuli, but could be to do with differences in the music structure between the two types of stimuli. More notably, an fMRI study comparing major and minor classical musical pieces performed on a MIDI piano found medial and superior frontal and posterior cingulate activation during minor, i.e., sad musical pieces, whereby major, i.e. happy musical pieces, failed to reveal differences [Khalfa et al., 2005].

In previous studies of affect in music, no direct comparisons between happy and sad pieces were possible, since control conditions have involved dissonant musical stimuli [Blood et al., 1999], scrambled musical pieces [Koelsch et al., 2006; Menon and Levitin, 2005] or rest [Brown et al., 2004]. Applications comparing musically induced sad and happy states are needed to determine the extent of overlap or difference between those states.

Somewhat surprisingly, neutral stimuli were associated with increased BOLD responses in the left auditory association area and left posterior insula. Previous studies found a lateralization for auditory functions with the right hemisphere traditionally involved in processing emotional information of sound and speech, whereas the left hemisphere is concerned with the structural analyses of auditory information (for a review please refer to [Tervaniemi and Hugdahl, 2003]). As the selected stimuli had little affective content the activation observed in this area could be a reflection of an analytical listening process.

LIMITATIONS

As the primary interest was on the emotional response to music, pieces were solely selected on their likely ability to induce sad, happy, and neutral states. Therefore, other musical features such as tempo, mode, pitch, or timbre were not taken into account. It is possible that the sad musical pieces could have had a more relaxing effect on the listener, while the happy pieces could have induced arousal states. These factors should be controlled in future studies by measuring skin conductance and collecting behavioral arousal ratings. Another limitation was that the strength of emotional valence was unbalanced between happy and sad musical pieces, suggesting stronger emotional valance for happy than sad pieces. This finding could support the notion that happy music in general triggers stronger reward responses compared to sad emotional states. However, areas that have previously been observed during sad emotional processing, such as amygdala, were significantly activated during sad emotional music, which suggests a sufficiently strong emotional valence for sad pieces. Future studies might want to alternate happy, neutral, and sad pieces throughout the experiment, allowing easier within‐subject comparisons, rather than having to perform between‐group comparisons on a relatively small sample. However, the present design was chosen to investigate and establish whether order of presentation has an impact on the induction of affective states and to facilitate the temporary mood induction by the musical pieces.

An in‐depth debriefing after the scanning procedure should also be considered in future studies employing popular pieces of music. This might be of importance as autobiographical memory retrieval could have an impact on the processing of musical pieces.

Additionally, the study might have profited from a procedure to enforce attendance to the musical stimuli. However, given the pattern of activation and the fact that music is a very pervasive stimulus it can be argued that participants in this study were in fact attentive toward the stimuli.

CONCLUSIONS

The present study investigated neural correlates of emotions induced by happy and sad emotional musical stimuli. Objectively validated happy, sad, and neutral musical pieces were determined in an initial behavioral study. The happy musical stimuli were associated with increased activation in the bilateral ventral and left dorsal striatum, left ACC and left parahippocampal gyrus. Sad musical stimuli led to increased activation in right medial temporal structures. These activations were observed without employing region‐of‐interest analyses frequently used in previous studies, which further underlines the strength of the present data. These findings may be integrated with recent research investigating musical emotions with familiar musical pieces.