Edited by: Ralph L. Holloway, Columbia University, USA
Reviewed by: Karol Osipowicz, Jefferson Neuroscience Hospital, USA; Jeffrey Bedwell, University of Central Florida, USA
*Correspondence: Donna L. Maney, O. Wayne Rollins Research Center, Emory University, 1510 Clifton Road NE, Room 2006, Mail stop 1940-001-1AC, Atlanta, GA 30322, USA. e-mail:
This is an open-access article distributed under the terms of the
Since the time of Darwin, biologists have wondered whether birdsong and music may serve similar purposes or have the same evolutionary precursors. Most attempts to compare song with music have focused on the qualities of the sounds themselves, such as melody and rhythm. Song is a signal, however, and as such its meaning is tied inextricably to the response of the receiver. Imaging studies in humans have revealed that hearing music induces neural responses in the mesolimbic reward pathway. In this study, we tested whether the homologous pathway responds in songbirds exposed to conspecific song. We played male song to laboratory-housed white-throated sparrows, and immunolabeled the immediate early gene product Egr-1 in each region of the reward pathway that has a clear or putative homologue in humans. We found that the responses, and how well they mirrored those of humans listening to music, depended on sex and endocrine state. In females with breeding-typical plasma levels of estradiol, all of the regions of the mesolimbic reward pathway that respond to music in humans responded to song. In males, we saw responses in the amygdala but not the nucleus accumbens – similar to the pattern reported in humans listening to unpleasant music. The shared responses in the evolutionarily ancient mesolimbic reward system suggest that birdsong and music engage the same neuroaffective mechanisms in the intended listeners.
Ornithologists and musicians alike have long contemplated whether the song of birds might somehow be classified as “music.” The question can be approached from a variety of angles, each of which produces a somewhat different answer. Researchers have asked, for example, whether birdsong and music share evolutionary precursors or functions (Darwin,
Birdsong, hereafter referred to as song, is a signal; it has a sender and a receiver. Ultimately, a signal’s effect on the receiver, not its structure, dictates its meaning and function (reviewed by Scott-Phillips,
Measuring behavioral responses is but one way to assess the effects of a signal on the receiver. Over the past decade, neuroimaging studies have identified at least 20 different brain regions that show altered BOLD or PET responses during music listening. Some of the most commonly reported responses, particularly to music that is pleasurable to the listener, are those of the mesolimbic reward system. This system consists of the ventral tegmental area (VTA) and its dopaminergic projections to several regions of the forebrain, for example the nucleus accumbens (nAc) in the ventral striatum. Release of dopamine in nAc occurs at precisely the time that intensely pleasurable autonomic responses, or “chills,” are experienced during music listening (Salimpoor et al.,
In this study we looked for neural responses to song in the avian homologues of music-responsive brain regions. Functional MRI can be used in songbirds listening to song (Van Meir et al.,
In this study, we used Egr-1 as a marker to map and quantify neural responses in the mesolimbic reward system in male and female white-throated sparrows (
The valence of song may be affected also by endocrine state. In
All research was conducted in accordance with National Institutes of Health (NIH) principles of animal care, federal, and state laws, and university guidelines. Twenty-three white-throated sparrows of each sex were captured in mist nets during fall migration and housed initially in mixed-sex aviaries at the animal care facility at Emory University. The sex of the animals was confirmed
Before the start of each experiment, birds were moved to individual cages (15″ × 15″ × 17″) inside walk-in sound-attenuating booths (Industrial Acoustics, Bronx, NY, USA). On the day of transfer, each bird received one subcutaneous silastic capsule (ID 1.47 mm, OD 1.96 mm, Dow Corning, Midland, MI, USA) sealed at both ends with A-100-S Type A medical adhesive (Factor 2, Lakeside, AZ, USA). Females received 12 mm capsules that were either empty (
On the afternoon prior to stimulus presentation, each bird was isolated by placing its cage inside an empty sound-attenuating booth equipped with microphone, speaker, and video camera. The stimulus playback began at 1 h after lights-on the following morning and was delivered
White-throated sparrow songs obtained from the Borror Laboratory of Bioacoustics birdsong database were converted to AIFF format and background noise was removed. The recordings were edited so that a song was heard every 15 s, which mimics a natural song rate. Sequences of songs were then spliced together so that the identity of the singer changed to a novel male every 3 min. Presenting a variety of songs helps overcome habituation to the stimulus (Stripling et al.,
For each of the 14 recordings of males singing, the frequency of each whistle (note) in one song was measured using AudioXplorer (Arizona Software, San Francisco, CA, USA). Songs usually contained five distinct frequencies. For each song, eight sinusoidal tones were generated at these frequencies and arranged in a random order 200 ms apart, resulting in a tone sequence that matched the song in duration, the average number of onsets and offsets, and total sound energy at each frequency. Tone sequences were spliced together as for the song stimuli, with 15 s of silence between each sequence, in an order determined by a balanced Latin Square.
Sixty min following the onset of the stimulus presentation, birds were deeply anaesthetized with isoflurane (Abbott Laboratories, North Chicago, IL, USA) and decapitated. Ovaries were inspected to confirm a regressed state. Brains were harvested, fixed, and sectioned at 50 μm as previously described (Maney et al.,
Examples of Egr-1 labeling are shown in Figure
Human region | Avian region | Reference |
---|---|---|
Nucleus accumbens (nAc) | Nucleus accumbens (nAc) | Balint and Csillag ( |
Ventral tegmental area (VTA) | Ventral tegmental area (VTA) | Bottjer ( |
Caudate nucleus | Lateral striatum (LSt) | Karten ( |
Hippocampus (Hp) | Hippocampus (Hp) | Erichsen et al. ( |
Medial amygdala (MeA) | Nucleus taeniae of the amygdala (TnA) | Cheng et al. ( |
Prefrontal cortex (PFC) | Caudolateral nidopallium (NCL) | Mogensen and Divac ( |
Egr-1-ir was quantified in each ROI by a blind observer using the thresholding feature in ImageJ (NIH). Briefly, the labeled nuclei with an optical density higher than a threshold value were counted within the entire ROI, obtained by tracing its borders in ImageJ (VTA) or within a circular area of approximately 0.2 mm2 placed within the ROI (all other regions). In most cases we used the threshold automatically set by ImageJ, which is based on contrast. Rarely, because of higher background staining, automatic thresholding caused obvious errors in the selection of labeled nuclei; in those cases the threshold was set manually. Our manual thresholding procedure has been shown to have high interrater reliability and low variability (Matragrano et al.,
Data from males and females were analyzed separately. The values for cells per unit area in each region were square root transformed to normalize their distribution, then entered into a two-way MANOVA (SPSS) with treatment (hormone or blank) and stimulus (song or tones) as the independent variables. One missing data point for the VTA in the male dataset was generated in SPSS using the series mean. Significant main effects or interactions were followed by two-way ANOVAs for each region. When a main effect of stimulus or an interaction between stimulus and treatment was found, pairwise t-
Egr-1 induction in the ROIs in females is plotted in Figure
Effects of stimulus and treatment |
Pairwise comparisons, |
||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Stimulus |
Treatment |
Stimulus × treatment |
song vs. tones |
E2 vs. blank |
|||||||||
Region | η2 | η2 | η2 | E2 | Blank | Song | Tones | ||||||
nAc | 4.855 | 0.130 | 2.894 | 0.105 | 0.078 | 9.652 | 0.259 | 0.467 | 0.302 | ||||
VTA | 12.313 | 0.311 | 6.099 | 0.154 | 2.043 | 0.169 | 0.052 | 0.232 |
0.471 | ||||
LSt (caudate) | 4.801 | 0.117 | 14.222 | 0.347 | 2.743 | 0.114 | 0.067 | 0.712 |
0.078 | ||||
Hp | 8.626 | 0.182 | 11.121 | 0.234 | 7.880 | 0.166 | 0.943 | 0.751 | |||||
TnA (MeA) | 20.359 | 0.363 | 0.023 | 0.880 | <0.001 | 14.913 | 0.266 | 0.602 | |||||
NCL (PFC) | 4.398 | 0.159 | 1.135 | 0.300 | 0.037 | 5.475 | 0.176 | 0.948 | 0.450 |
In the females that performed CSD (E2-treated birds hearing song,
Egr-1 induction in the ROIs in males is plotted in Figure
Effects of stimulus and treatment |
Pairwise comparisons, |
||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Stimulus |
Treatment |
Stimulus × treatment |
song vs. tones |
E2 vs. blank |
|||||||||
Region | η2 | η2 | η2 | E2 | Blank | Song | Tones | ||||||
nAc | 0.015 | 0.904 | 0.001 | 3.445 | 0.079 | 0.151 | 0.220 | 0.645 | 0.010 | – | – | – | – |
VTA | 0.158 | 0.696 | 0.006 | 8.528 | 0.307 | 0.066 | 0.800 | 0.002 | – | – | – | – | |
LSt (caudate) | 0.038 | 0.847 | 0.002 | 0.005 | 0.947 | <0.001 | 0.016 | 0.900 | 0.001 | – | – | – | – |
Hp | 1.784 | 0.197 | 0.067 | 5.577 | 0.210 | 0.315 | 0.581 | 0.012 | – | – | – | – | |
TnA (MeA) | 18.598 | 0.440 | 3.275 | 0.086 | 0.077 | 1.381 | 0.254 | 0.033 | 0.052 | 0.678 | |||
NCL (PFC) | 2.667 | 0.119 | 0.122 | 0.002 | 0.963 | <0.001 | 0.154 | 0.700 | 0.007 | – | – | – | – |
The number of songs given by the males during the stimulus presentation was correlated with Egr-1 expression only in the VTA (Spearman’s
In this study, we showed evidence of neural responses in the reward pathway of songbirds listening to conspecific song. These responses were significantly greater than those to behaviorally irrelevant control sounds in all of our regions of interest only in females with breeding-typical levels of E2. In non-breeding females treated with placebo, the response to song was not different from the response to control sounds. These results are consistent with studies in other species showing behavioral evidence of the incentive salience of song for receptive females (Eriksson and Wallin,
In males, we found a main effect of T-treatment on Egr-1 expression in several regions of interest, which is consistent with other findings that gonadal steroids alone, independent of sensory stimuli, can induce IEG activity (Charlier et al.,
Our overall goal in this study was to compare the neural responses in the mesolimbic reward system of songbirds listening to conspecific song with those of humans listening to music. We compare below the pattern of Egr-1 responses observed in this study with the published literature describing BOLD (fMRI) and rCBF (PET) responses in humans listening to music. As a caveat, is important to note that BOLD and PET responses are qualitatively different both from each other and from IEG responses, and we should not assume that an increase in one signal always accompanies an increase in the other. BOLD and IEG signals do overlap extensively, however, when both techniques are applied in the same animals (Lazovic et al.,
In the human literature, the brain region most commonly reported to respond to music is the ventral striatum, which includes the nAc. Responses in the human ventral striatum can be elicited by pleasant, liked, or happy music and are not elicited by unpleasant or sad music (Blood and Zatorre,
For males and non-breeding females, hearing song may be predictive of a fight. For females in breeding condition, however, hearing male song may precede rewarding activities such as courtship and copulation. We must therefore consider the possibility that song-induced Egr-1 responses in the reward pathway indicate the anticipation of reward, not reward
Many investigators have reported caudate responses in humans listening to pleasurable music (Blood and Zatorre,
Of all the Egr-1 responses observed in this study, the largest was in the Hp. This region was previously shown to respond to conspecific song in zebra finches of both sexes (Bailey et al.,
Perhaps because it is heavily interconnected with the Hp, the response of the amygdala to music often mirrors the Hp response (Blood and Zatorre,
Although the amygdala is a heterogeneous region containing several nuclei with distinct functions (reviewed by Swanson and Petrovich,
Although the MeA does appear to mediate the processing of emotional and possibly rewarding sensory stimuli, it seems unlikely that a response in this part of the pathway alone, as was seen here in T-treated males, indicates that the signal has incentive salience. Note that in humans, listening to sad or unpleasant music stimulates the amygdala and Hp without stimulating the nAc or caudate (Koelsch et al.,
The PFC responds in humans listening to music (Blood and Zatorre,
Our main goal in this study was to compare the pattern of neural responses in songbirds listening to song with those of humans listening to music. In doing so, we focused on the mesolimbic reward pathway. There are, however, additional regions of the human brain that respond to music and which have homologues in birds. To complete our comparison, we summarize those findings below.
In humans listening to music, BOLD responses in the nAc and the VTA were strongly correlated with responses in the hypothalamus (Menon and Levitin,
Although it is not traditionally considered part of the mesolimbic dopamine system, the PAG is interconnected with the VTA and plays an important role in opioid-mediated reward (reviewed by Le Merrer et al.,
Comparisons of music and song are usually limited to the bioacoustic analysis of the signals themselves. Here, we compared the responses of the receivers of the signal – the listeners. We found that song is similar to music in that they both induce responses in components of the mesolimbic reward pathway. In receptive females, every region of this pathway that has been reported to respond to music in humans, and that has a clear avian homologue, responded to song. In T-treated males, we detected responses only in the avian homologue of the MeA. This pattern, in other words a response in the amygdala without responses in the nAc or caudate, is typical of humans listening to unpleasant or fearful music (Baumgartner et al.,
At first, this result may seem to distinguish song from music, the pleasantness of which is not thought to vary along these parameters (cf. Panksepp,
Both song and music elicit responses not only in brain regions associated directly with reward, but also in interconnected regions that are thought to regulate emotion. The involvement of this circuit in music listening suggests that hearing music activates evolutionarily ancient neuroaffective mechanisms usually reserved for stimuli that, like song in birds, are critical for reproduction and survival. The adaptive value of bird song may be obvious; less obvious is the fact that music shares many of the same social functions. Like song, for example, music facilitates social contact, reduces conflict, communicates emotional state, and helps maintain interpersonal attachments (reviewed by Koelsch,
The musical quality of bird songs has provided the impetus for many comparisons between the two sounds (e.g., Darwin,
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank Ellen Cho, Chris Goode, Henry Lange, Meredith LeBlanc, and Madiha Raees for technical assistance. We are grateful to the Department of Biology at Emory University for the use of resources, and to Wayne Kuenzel, Scott Husband, and Lauren Riters for advice. Finally, we thank Darryl Neill and Kristin Wendland for comments on a previous version of the manuscript.