Connectivity between nidopallium caudolateral and visual pathways in color perception of zebra finches

Researchers demonstrated an elegant ability for red discrimination in zebra finches. It is interested to understand whether red activates exhibit much stronger response than other colors in neural network levels. To reveal the question, local field potentials (LFPs) was recorded and analyzed in two visual pathways, the thalamofugal and the tectofugal pathways, of zebra finches. Human studies demonstrate visual associated telencephalons communicate with higher order brain areas such as prefrontal cortex. The present study determined whether a comparable transmission occurs in zebra finches. Telencephalic regions of the thalamofugal (the visual Wulst) and the tectofugal pathway (the entopallium) with their higher order telencephalon, nidopallium caudolateral (NCL) were simultaneously recorded. LFPs of relay nuclei (the nucleus rotundus, ROT) of tectofugal pathway were also acquired. We demonstrated that LFP powers in the tectofugal pathway were higher than those in the thalamofugal pathway when illuminating blue lights. In addition, the LFP synchronization was stronger between the entopallium and NCL. LFPs also revealed a higher Granger causality from the direction of entopallium to NCL and from ROT to entopallium. These results suggest that zebra finches’ tectofugal pathway predominately processing color information from ROT to NCL, relayed by entopallium, and blue could trigger the strongest response.


Results
Tectofugal pathway mainly mediates brain responses to colors. We are interested in the brain structures to interpret the meaning of color stimulation. We recorded the LFPs from the relay nuclei of the thalamofugal and tectofugal pathways in brains. To study the effects of different colors on activating thalamofugal and tectofugal pathways, we used red, green, and blue (RGB) codes from 0 to 1 with 0.25 increment to create 15 colors (for details, Table S1; see "Materials and Methods" section). Fifteen colors were flashed to the subjects' right eye (Figs. 1, S1-S3) and simultaneously recorded the LFPs at the relay nuclei of the thalamofugal and tectofugal pathways, owing to the fact that male zebra finches usually use right eye to observe females during courtship 28 . We averaged the LFPs across the same color stimuli with the same background color (black background color (Figs. 1 and S1)). Although the subjects were recorded under general anesthesia, the evoked potentials still emerged between 0 and 500 ms after color flashing when black background was used as baseline  www.nature.com/scientificreports/ (Fig. 1). The LFP traces of evoked potentials were dissimilar among the four recording targets. As displayed in Fig. 1, the evoked potentials in NCL were slower and had indistinct negative potentials (the valley of NCL peaked after 250 ms) when compared with those of ENTO (the valley peaked at approximately 250 ms), VW (the valley peaked at approximately 250 ms), and ROT (nearly no valley occurred). Therefore, the LFPs demonstrated no contamination from volume conduction. Volume conduction is an electric current transmitting between nearby brain areas and usually shows no phase delay in LFPs between sources and the nearby areas, as indicated that the recorded waveforms are similar between different brain areas and may mask the actual LFP signals. The different shape and response time of evoked potentials among regions in our LFP recordings suggest that volume conduction did not cause artificial synchronization and the acquired LFPs were real signals. Therefore, it is worth to further analyze the functional connectivity. It has been demonstrated that gamma oscillations are triggered by vision through the tectofugal pathway in pigeons 42 ; therefore, we investigated the changes of LFP powers in different spectra after receiving the different 15-color stimuli before further analyzing the functional connectivity. The LFP spectrograms stimulated by 15 different colors depicted that the strongest responses were elicited after receiving blue to green lights (Figs. 2A, S2, color No. 6: RGB color (0, 1, 1), and color No. 7: RGB color (0.25, 1, 0.75)). However, if the white background color was used as the baseline, no distinct response was found among these color stimuli (Fig. S3). The spectrograms of color stimuli displayed a predominant power between middle and high frequencies from the onset of the trigger to half a second after the trigger ( Fig. 2A dashed line box). The powers between the middle and high frequency bands were increased (as the arrows indicated in Fig. 2A); therefore, we referred to rat's frequency bands and divided the responsive spectrograms into three frequency bands: the low frequency (< 20 Hz, including the delta, theta and alpha waves), the middle frequency (20-60 Hz, mainly containing slow gamma waves), and the high frequency (60-100 Hz, mainly containing fast gamma waves) 43 Table S3A). These results suggested that the NCL-ROT pathway plays the major role in processing color information. We further created 7 rainbow colors with the same radiation power (Table S1), since the previous 15 colors created by the color codes did not control the radiation power. After adjusting these rainbow colors to an identical radiation power, the colors were a little bit pale ( Fig. 3A; color index on the left; also see "Materials and Methods" section, Table S1). The radiation-power-controlled rainbow colors still triggered evoked potentials (Fig. S4, LFP traces). In addition, the rainbow colors still generated strongest power in ENTO and significantly stronger than VW (Fig. S5A, analyzed by one-way repeated measure ANOVA compared between the total rainbow colors among 4 areas, Bonferroni-adjusted significance tests for pairwise comparisons. See supplementary  Table S3C for detailed means ± SEMs, F values, p values, and corrected p values). Since there were some missing values from VW recording because of the broken recording wires and one-way repeated measure ANOVA excludes the missing trials, we analyzed the statistical differences by one-way ANOVA. The statistical analysis still demonstrates significant higher power in ENTO than VW (see supplementary Table S3B for detail statistic values). The blue colors triggered the highest potentials, especially in ENTO and NCL (Figs. 3B-D, and S5B, analyzed by one-way repeated measures ANOVA, Bonferroni-adjusted significance tests for pairwise comparisons; see supplementary Tables S4 and S3D for detailed means ± SEMs, F values, p values, and corrected p values). The powers in ROT and VW are still relatively low ( Fig. 3A-D), which is consistent with the findings using 15-color stimuli. We were interested in knowing whether blue trigged highest response in ENTO than VW, therefore, we analyzed z-scored power after the blue stimulation. The data elucidate a significant stronger power in ENTO than VW (Fig. S5B, One-way repeated ANOVA, see supplementary Table S3D for detailed means ± SEMs, F values, p values, and corrected p values).
Phase synchronization between the NCL and the tectofugal pathway is stronger. We analyzed the phase synchronizations between ROT and NCL (ROT-NCL), ENTO and NCL (ENTO-NCL), and VW and NCL (VW-NCL) by the weighted phase lag index (WPLI) to determine whether the higher-level cognitiverelated brain region, the NCL, involves in processing color stimuli. The WPLI is an index which is insensitive to volume conduction 44 and suitable to analyze the levels of synchronization within a small size region of brain. We calculated WPLI within 500 ms after different color stimuli ( Fig. 4A-C) and we also determined WPLI within 500 ms by all of rainbow color stimuli ( Fig. 4D-F). The phase synchronization of the ROT-NCL, ENTO-NCL, and VW-NCL were not significantly higher compared with other colors when stimulated with blue ( Fig. 4A D. -0.5 0 0.5 -0.5 0 0.5 -0.5 0 0.5 -0.5 0 0.5  Table S6A for detailed statistic values). One-way repeated measures ANOVA were also used and demonstrated higher ENTO-NCL synchronization than VW-NCL in middle and high frequency (Fig. S5C, Table S6B for detailed statistic values). Again, blue was extracted and further analyzed (Fig. S5D, Table S6C for detailed statistic values; One-way repeated measures ANOVA). The results demonstrated significant stronger phase synchronization in ENTO-NCL than that of VW-NCL in the high frequency. These results imply that the color signal transmission was mediated by the tectofugal pathway to NCL. To further confirm the transmission pathway and connectivity, we injected the non-trans-synaptic retrograded tracer, fluorogold, into NCL to confirm the anatomical afferent projections to NCL ( Fig. S6A-C). The brain histology showed that the bilateral ENTOs exhibit prominent fluorescent signals after injecting fluorogold bilaterally to NCL. Projections from the striatum were also noticed. In contrast, there was a mild retrograded signal in the caudal part of VWs (Fig. S6B).
To determine the hemispheric integration of these pathways, we injected fluorogold into the left NCL and incubated for additional two weeks. We found that most of the signals were noticed in the mesopallium, and there was few fluorogold signal in the left cranial part of VW (Fig. S6C). This retrograde staining result implies that the afferent projection to NCL is predominant from ENTO, but not from VW.  www.nature.com/scientificreports/ Color information is processed in the direction from tectofugal pathway to NCL. We further assessed the functional connectivity between the two pathways and NCL to reveal the directional interactions when the subjects were stimulated by different colors (Fig. 5A). The overall Granger causality for both directions are significantly higher in the ENTO ⇔ NCL (Fig. 5B,C, Table S7A for detailed statistic values and Bonferroni post-hoc tests of one-way ANOVA). One-way repeated ANOVA demonstrated the significant strongest Granger causality in the ENTO ⇔ NCL as well if we pooled all the results of all rainbow colors (Fig. S7A, Table S7B for detailed statistic values). This phenomenon is still significant if we only analyzed blue stimuli (Fig. S7B, Table S7C for detailed statistic values). The increase of Granger causality was observed in a direction from ENTO www.nature.com/scientificreports/ to NCL during blue stimuli (Fig. 5D, two-tailed paired t-test, see Table S8B for detailed degree of freedom, t values and p values). Similar finding could also be found after indigo (Fig. 5D, Table S8B) and violet stimuli (Fig. 5D, Table S8B), but not after other color stimuli. These results suggest that the LFPs recorded from ENTO are leading the LFPs in NCL, because the direction from ENTO to NCL demonstrated the highest Granger causality and implied the color information is processed from the tectofugal pathway to NCL. Finally, it is also interesting to determine whether ROT transfers information to ENTO, because ROT is a relay nucleus of tectofugal pathway. Our data demonstrated that the direction from ROT to ENTO had a higher Granger causality than that of the direction from ENTO to ROT after the stimulation of rainbow colors (Fig. S7C, Table S7D for detailed statistic values) or blue color (Fig. S7D, Table S7E for detailed statistic values).

Discussion
ENTO and VW are respectively the telencephalic areas of tectofugal pathway and thalamofugal pathway. However, it is still unclear whether higher order avian brain areas, such as NCL, communicate with them or not. We hypothesized that the telencephalons of visual pathways communicate with higher order brain areas in zebra finches and the ENTO has stronger communication with the NCL than VW, when stimulated by colors. We initially were expecting prominent responses in ENTO when zebra finches see colors. The present study demonstrated that the ENTO generates stronger power than those of ROT and VW when the right eye was stimulated by colors. In addition, the NCL also represented strong power. When we further analyzed the functional connectivity between ENTO and NCL, we found that color stimuli enhanced synchronization between ENTO and NCL in the direction from ENTO to NCL. Interestingly, the synchronization between VW and NCL were negatively synchronized, suggesting these two areas communicated poorly with each other 37 . Moreover, blue is the color that evoked the strongest power in the direction from ENTO to NCL; the Granger causality also demonstrated the LFPs acquired from ROT leaded the consequent LFPs of ENTO. However, the different colors did not result in different phase synchronization between NCL and other three regions. We were also interested in learning what colors zebra finches are sensitive to. Indeed, we hypothesized that their visual pathway is sensitive to red, but we cannot ignore the fact that blue is important for finding sky or water and green may be important for searching plants. Surprisingly our results partially support that blue is relatively important to zebra finches. These results also raise some interesting questions about the physiological or ecological advantages of "blue" and also the physiological functions of the tectofugal pathway, the thalamofugal pathway, and NCL for zebra finches. The LFP experiments are mainly done in rodent models; therefore, we firstly applied findings from rodents to explain our current results. Then, we discussed some potential physiological meanings for zebra finches.
The LFP spectrograms. We employed the results of LFPs to determine our hypothesis. The LFP is a summation of various potentials such as membrane potential and action potentials. The synaptic potential is often the main source of LFP 38 . The LFP contains multi-dimensional information including frequency, amplitude, phase, and time. Therefore, the present study analyzed several aspects of the LFPs. In the spectrograms, we noticed that blue evoked a strongest power within 0-500 ms after the stimuli, especially in ENTO and NCL. Strong LFP power (or amplitude) usually suggests that the target areas are in a very active status 38 . In addition, the color stimuli evoked approximately three frequency bands. The LFP frequency is an unique way for communicating between brain regions and each frequency band may encode different information and has its own physiological functions 39 . We analyzed the frequency bands to explore the potential communication of color information between ENTO and NCL. Our data demonstrated that the middle and high frequency bands revealed the power differences between ENTO and VW. In general, low frequency oscillations are propagated farther, given that the cell membrane is a low pass filter 37 . In the other hand, high frequency oscillations provide a more temporally organized transmission than low frequency oscillations 40,41 . We cannot conclude that the color information is not transmitted between VW and NCL, since we still acquired significant potentials in VW. However, the evidence from phase synchronization and Granger causality support that the ENTO-NCL is more sensitive to colors.

Phase synchronization and directional connectivity. The phase of LFP also encodes information
for propagating between brain regions 37 . Synchronizations help the communication between different brain regions 45 . Although we found no color preference of phase synchronization, the ENTO-NCL still revealed strong positive synchronization, whereas VW-NCL showed weak or negative synchronization. Even though the blue evoked the highest activities in the ENTO and NCL, the WPLI cannot reflect the difference of amplitudes because WPLI only measures the phases of LFPs. We postulated that the phase synchronization encoded information of stimuli, but not colors. In the other word, the communications between ENTO and NCL increased, which paved a way for color information but the color itself was encoded by amplitudes. The weak communication between VW-NCL implys that the tectofugal pathway is the main path in response to color stimulation. The directional connectivity also suggests a strong information transmits from ENTO to NCL when zebra finch received blue stimulation.
Sensitivity to blue stimulation. We demonstrated that the male zebra finch is sensitive to blue lights. The retina studies have shown that several species of birds are sensitive to both yellow and UV lights, but some species show their maximal sensitivity at the green spectrum 46 . Therefore, the spectral sensitivity in birds may vary among species. Bennett et. al. reported that UV vision dominantly contributes to the mate-selection in zebra finches 47 . Our data could not rule out the possibility of sensitivity for UV light in the zebra finch, but we believe our results shed some light on zebra finches' color preference. Additionally, a recent study also demonstrates the zebra finches' retina is sensitive to blue as well 48 .
Scientific Reports | (2020) 10:19382 | https://doi.org/10.1038/s41598-020-76542-z www.nature.com/scientificreports/ Our results also showed that the color-evoked potentials were significantly detected when we used black as baseline, rather than white color, between different colors. It is possible that white baseline, which contains all spectra of the visible lights, may saturate the visual response and let the visual pathways do not respond to the subsequent colors. There is also a behavior report indicates that the contrast of the background affects color discrimination in zebra finches 49 . Although the background used in the present study are not parallel to the back or white baseline, it is likely that the responses to colors are modulated by recent or adjacent colors. To minimize the confounding factor, we stimulated the subjects with colors in random orders. The intensities of white or other colors may also affect the brain activities. A study regarding the spectral sensitivity in avian retina reports that increasing intensity of certain color shifts the maximum electroretinal potential toward the shorter wavelengths 46 . Therefore, we manually adjusted the intensities of different colors to be the same and acquired their evoked potentials. The limitation of this study is the light source, since we did not use a light with narrow wavelength band such as laser. We simply used a laptop, which generates colors based on a human's vision system (i.e. trichromatic colors). Even though this study did not stimulate the visual pathway with a precisely narrow wavelength band and no invisible light (for humans) was generated by the laptop, the results still shed light on their brain activities when stimulated by trichromatic colors. We expected more complicated responses will be observed when stimulated by mixing the UV light with trichromatic colors 47 .
In addition, the color perception in zebra finches is a complicated issue. Our report raises several new questions needed to be discuss. For instances, what the limitation is for using a monitor which generates RGB colors to stimulate the animals with tetrachromatic vision. Do colors influence the zebra finches' perceptive brightness and cause the strongest evoked LFPs from blue? Although the functions of the avian brain are species-specific, we reviewed some studies from other bird species and tried to reveal the potential answers of whether the violet cones (or ultraviolet cones in finches 48 ) interact with the S-, M-, and L-cones in the tetrachromatic vision of bird. For humans, the purple from the monitor stimulates both red-and blue-sensitive cone cells and humans interpret it as purple 50 . The purple from the laptop monitor is a mixing of red and blue, not a pure short wavelength light. Therefore, we should roughly interpret the rainbow colors as: red, red + green (which generate orange and yellow), green, blue, blue + red (which generate indigo and violet) ( Table S2 for the detail combination of RGB). Recently, a study from hummingbirds constructs an avian tetrahedral color space 51 and proves their ability to discriminate UV. Our study just implied that the stimulation on the zebra finch S-cones evoked a strong response, but still cannot rule out the possibility that UV can generate strongest visual responses in their brain. On the other hand, despite we adjusted the rainbow colors to have the same radiation power, it is still unclear how blue color generates strongest responses in the brain. In human, we feel yellow is brighter than red, green, blue, even they are displayed by the same radiation power. We proposed that the mechanism of feeling different brightness among colors in human is not the cause of strong blue response in the zebra finch. The sensitivity spectrum of human's green-and red-sensitive cone cells is highly overlapped 52 . Therefore, yellow is able to stimulate more cones than other colors. But the sensitivities of zebra finches' cone cells are evenly distributed across spectrum 48 . Thus, we think the strong blue response is not a phenomenon of cross-reaction between different types of cone cells. Indeed, study from the oil droplets of zebra finches implies that zebra finches have higher cone spectral sensitivities for blue and UV than red and green 48 . Our data obtained from the brain activities further support this result from the retina's study.
Although present study did not explain why the brain of zebra finch is more active to blue, their nature habitat may hint the potential reasons. Zebra finch is a diurnal animal and lives in relative arid areas 53 . We hypothesized that water and sky are key factors for zebra finch surviving, so they need to be spotted as fast as possible. During the experiment we also observed that zebra finches became quite and standstill if we turn off the room light. We think, as the room light was off, zebra finches were searching for sky for flying toward. Besides, researchers discover that zebra finches have blue and UV light-dependent magnetic compass, suggesting that blue light is critical for their navigation 16 . Tectofugal pathway, thalamofugal pathway and NCL. Our findings indicated that the color-evoked potentials were stronger in the tectofugal pathway than the thalamofugal pathway. Moreover, the synchronization between the relayed nuclei of the tectofugal pathway and NCL was also stronger. These results are similar to some studies using pigeons as subjects, in which lesion of ROTs impairs the color discrimination 9 and some color sensitive units are also found in ROT 54 . In addition to ROT, about 30% of tectal units are able to respond to certain wavelengths 17 . Although we discovered that ENTO (a downstream of ROT) represents stronger activities than those of ROT, we cannot exclude the roles of ROT for processing color information. In our result, we demonstrated that NCL responses were highly correlated with ENTO but the correlation is relatively low between the NCL and ROT. We think ROT did not reveal as strong response as ENTO because ENTO needs more intensive communication with NCL. This hypothesis is supported by the synchronization and directional connectivity results of NCL-ENTO. With regard to the thalamofugal pathway, color-sensitive units have been discovered in the ventral lateral geniculate nucleus, which consists of inputs from both retinas and VW 55 . Bredenkotter and Bischof used 1-ms flash to evoke and record the field potentials of VW and ENTO, and found the amplitudes recorded from the contralateral hemispheres are similar 56 . A lesion study demonstrates that the VW in zebra finch involves in spatial information processing and ENTO analyzes the pattern of objects 57 . The zebra finches' VW even perceive vision mediated by earth magnetic field orientation 58 . It is still unclear whether the VW modulates other cognitive functions related to vision, but researchers demonstrate the important role of VW in imprinting for chicken 59 . These pieces of evidence may support our hypothesis that the tectofugal pathway is much critical than the thalamofugal pathway in regard to the color information processing, since colors are also important cues for discriminating objects. It would be also of interest to simultaneously record from the ventral Scientific Reports | (2020) 10:19382 | https://doi.org/10.1038/s41598-020-76542-z www.nature.com/scientificreports/ lateral geniculate nucleus. However, because of the limitations of channel and skull space, we selected only to record the relay nucleus (ROT) of tectofugal pathway.

Conclusions
Our result suggests that the communication between nidopallium caudolateral and tectofugal pathways is crucial for color discrimination. Moreover, ENTO and NCL are more active when the eyes are stimulated by blue and the visual information was transmitted from the direction of ENTO to NCL.

Materials and methods
Animals. In the experiments, the male zebra finches (n = 9, 5 to 8-month-old) were acquired from the commercial bird breeders (San-Xing Bird Store, Taipei, Taiwan). The birds were housed in home cages individually. The temperature of the environment temperature was controlled at 23 ± 1 °C, and the light-dark cycle was maintained under nature light (AM 7:00 light, PM 7:00 dark; summer). Food and water were available ad libitum.
All procedures performed in this study were approved by the National Taiwan University Animal Care and Use Committee, approval ID: NTU-106-EL-026. All methods described in this paper were performed in accordance with the guidelines and regulations of National Taiwan University Animal Care and Use Committee.
Surgery and electrophysiological data collection. After at least a 7-day accommodation in their home cages, the zebra finches were randomly selected for the recordings of evoke potentials. These finches stayed in an induction box, which provided pure oxygen for 10 min to raise their blood oxygen levels. Subsequently, they were intraperitoneally administrated butophanol (2 µg/g) and midazolam (2 µg/g) for analgesia and muscle relaxing. Anesthesia was induced by 2% isoflurane mixed with oxygen. Once they lost their reflex of deep pain, the subjects were fixed on the stereotaxic instrument and maintained anesthesia with 1.5% isoflurane 60  We further created 7 rainbow colors with the same radiation power (Table S1)  www.nature.com/scientificreports/ Evoked potentials (LFPs) and power estimation. The LFPs were Z-scored and averaged across the same color for measuring the evoked potentials, and the negative Z-scored power means it is below the average power. The power spectrograms were analyzed with the multitaper method from the open-source MATLAB toolbox Chronux 62 . Because the spectrograms need a period of samples and step a short time to create a dynamic of spectrum as function of time. We used 0.5-s windows with 0.05-s overlapping steps, set the time-bandwidth product at 3, and set the number of tapers at 5. We also extracted and averaged the values of stimuli between 0 and 500 ms for testing the statistic differences among stimuli or brain regions.
Phase synchronization across regions. The levels of synchronization between the ROT-NCL, ENTO-NCL, and VW-NCL were evaluated with a debiased estimator of the squared weighted phase lag index (WPLI) 44 . The codes can be download from the open source tool box, Fieldtrip https ://www.field tript oolbo x.org/downl oad.php 63 .
The WPLI analyzes an imaginary component of the spectrum across two LFPs, because it not only relates to the phase synchronization between two LFPs but is also insensitive to noise or contamination from volume conduction. Compared with classical coherence, this estimator analyzes the phase synchronization across brain regions and minimizes the effect of volume conduction contamination and sample size bias 44 . The WPLI normalized the two LFP with perfect synchronization to 1 and completed out of phase to -1. In order to access the WPLI in a similar manner with multitaper power estimation, we also broke the ± 1-s periods into 0.5-s windows with 0.05-s step and estimated the WPLIs. We extracted and averaged the values of each stimulation during 0-500 ms for testing the statistic differences among stimuli or brain regions.
Granger causality. The function connectivity between the ROT⇔NCL, ENTO⇔NCL, and VW⇔NCL were accessed by using Granger causality. We adapted the open-source MATLAB toolbox developed by Barnett and Seth 64 . It is also due to the input samples must contain a period of time to generated a Granger causality dynamic over time, which is similar to the method described in the power spectrogram and the WPLI, we calculated the Granger causality for the time domain every 0.5 s with 0.05-s steps for measuring the dynamics of Granger causality. We extracted and averaged the values of each stimulation during 0-500 ms for testing the statistic differences among stimuli or brain regions.
Histology and retrograde tracing of the afferents to the NCL. For tracing the afferent projections to NCL, a retrograde tracer, 4% fluorogold (Sigma Chemical), was microinjected into NCL in two birds. One subject was bilaterally (the NCL; AP, 1.0 mm; ML, ± 4.5 mm; DV, 4.0 mm relative to y point) injected with fluorogold using microinjection syringe pump at a speed of 1 µl/10 min. We administered 0.33 µl of fluorogold at each site and waited for 1 min, then moved the tip of the needle up for 100 µm and administered another 0.33 µl again and waited another 1 min, and repeated the procedure once again. Two weeks after injections, this bird was euthanized by isoflurane. The brain was removed and dissected into a 30-µm coronal section by a cryostat microtome. Some brain slices were stained with DAPI. We used an ultraviolet filter in the inverted microscope (IX83; Olympus, Tokyo, Japan) to detect the fluorescent reaction of fluorogold. Since the fluorescent reaction in VW was not high in this bird (Fig. S6B), we double confirmed this result with anther finch, which was microinjected with 4% fluorogold to the left NCL and waited for 4 weeks to let fluorogold travel a longer distance. The brain slices for confirming the locations of the implanted electrodes were also dissected into a 30-µm coronal section by the cryostat microtome. The pictures in Fig. S6E-H were taken under stereo microscope without staining. The coordinates were adopted from A stereotaxic atlas of the brain of the zebra finch 61 .

Results and statistics
All results in the figures are depicted as the means ± SEMs. The results of statistical analyses were done by SPSS (Version: 10.0.7, IBM, New York, USA). The stimulation trials were used as unit to be analyzed. We initially used one-way ANVOA to compared the differences between brain areas. In addition, we excluded the miss values of VW in one bird and retested differences between brain areas with one-way repeated measures ANOVA. Bonferroni post hoc comparison was used if the ANOVA test indicates a significant difference. For measuring the difference between the two causality directions, we used two-tailed paired t-tests.