Budgerigar Preferences Toward Universal Auditory Cues Of High and Low Arousal – A “Better Safe than Sorry” Strategy?

doi:10.21203/rs.3.rs-8122301/v1

Assessing others’ emotional states including arousal is crucial to survival and social interaction, with vocalizations being a particularly salient channel for conveying emotion across species. Here, we investigated whether arousal cues that are thought to be universal across vertebrates translate to perceptual preferences. Budgerigars (Melopsittacus undulatus) are an ideal model to study heterospecific arousal perception, exhibiting proficient spectral processing and distinct arousal-related vocalisations. We investigated budgerigar preferences using a place preference paradigm, measuring time spent with high- and low-aroused versions of synthetic sound sequences and vertebrate vocalizations. We expected a preference of low arousal over high arousal because high arousal is associated with stress. However, when testing the birds with synthetic sounds that contain universal cues to low and high arousal, our results show that while females preferred low arousal, males preferred high arousal. When we tested the birds with heterospecific vocalizations, both sexes showed equal avoidance of both conditions. Emotion recognition is thought to exploit cues shared across vertebrate species. Our results suggest that in addition to potential universal cues, the ecological niche, listener sex, and sound source type act as factors influencing emotion perception. We suggest a model for processing acoustic cues to emotion and provide testable predictions.

Budgerigar

Emotion Recognition

Cross-Species

Vocalization

Music

Perceiving, interpreting and responding to emotional states is crucial to survival and social interaction across many species. Understanding cross-species emotion communication, particularly through acoustic cues, such as in avian (Riters et al., 2022; Filippi, 2016) and mammalian communication systems (Briefer, 2012; Taylor et al., 2016), including music (Singh & Mehr, 2023; Nguyen et al., 2023) and speech (Frick, 1985; Scherer, 1995), is an area of ongoing research. Emotions are transient affective reactions to salient stimuli and facilitate adequate adaptive responses that are often conceptualized along the dimensions of arousal (cognitive and bodily activation) and valence (positive vs negative affective tone; Briefer 2012). Vocalization, employed to our best knowledge by all bird and mammal species, may be a particularly salient channel for conveying emotion, especially arousal.

The idea of shared codes for emotion across communication systems is long-standing (Darwin, 1872) and continues to draw attention to date (Briefer, 2012; Filippi et al., 2017; Greenall et al., 2022). Vocalizations have been found to be reliable indicators of the caller’s arousal across mammalian and avian species, with similar vocal features found throughout vertebrates (Briefer, 2012; Filippi et al., 2017). This phenomenon is likely due to shared mechanisms of arousal, which affect the properties of the sound-producing structures and downstream filtering structures (Taylor & Reby, 2010; Elemans et al., 2015; Briefer, 2020). Consequently, the vocal communication of arousal was suggested to be an evolutionarily conserved trait in terrestrial animals, including mammals and birds (Briefer, 2012, 2020; Filippi et al., 2017).

Recent studies have suggested that listeners can infer the emotional state of a caller across species (Mason et al., 2024; Debracque et al., 2023; Greenall et al., 2022; Maigrot et al., 2022; Thévenet et al., 2023; Merkies et al., 2022). For example, human listeners are able to discriminate low-arousal and high-arousal from calls of representatives of all vertebrate species groups (Filippi et al., 2017). Congdon et al. (2019) showed that black-capped chickadees and humans can learn to discriminate conspecific high- and low-arousal calls and transfer these discrimination abilities to heterospecific vocalizations. However, the universality of acoustic cues used for (heterospecific) emotion perception is under debate (Thévenet et al., 2023) and an investigation of other species is needed to draw broader conclusions. While discrimination studies demonstrate basic perceptual and learning differences across sounds varying in arousal, they do not directly assess whether heterospecific high- or low aroused sounds convey emotional information informing behavioral responses.

Therefore, a closer investigation of heterospecific arousal perception is needed to understand emotion perception across the animal kingdom, including its potential benefits such as avoiding threatening situations. Measuring untrained responses to stimuli might offer insight on the perceived emotional content of signals with low and high arousal. In this study, we extend this line of research by investigating budgerigars’ (Melopsittacus undulatus) preferences towards sounds encoding high and low arousal using a place preference paradigm, a method which measures time spent with stimulus types as a proxy of preference. Budgerigars, a small parrot species, offer an ideal model, exhibiting proficient spectral processing skills (Lohr & Dooling, 1998) and using distinct arousal-related vocalisations (Brockway, 1964). We hypothesized that budgerigars would show greater aversion of locations associated with highly aroused sounds than low aroused ones, reflecting adaptive responses to acoustic cues of threat. We carried out two place preference experiments, using synthetic sound sequences to assess the contributions of specific acoustic features (Experiment 1) and using vertebrate sounds to assess responses to biologically more plausible stimuli (Experiment 2).

Participants

We conducted two place preference experiments each with 12 adult budgerigars (6 male / 6 female) housed at the budgerigar lab at the Acoustics Research Institute of the Austrian Academy of Sciences in Vienna. This sample size of 12 individuals was used because a previous study on budgerigars in the same setup found that this sample size was adequate to find significant differences between conditions (Bellmann et al., 2025a, 2025b). All birds except for the females ‘Pistachio’ (Exp. I) and ‘Summer’ (Exp. II) participated in both experiments, which were six months apart. The birds were housed in two colonies that were located in separate rooms (L and R), which provided acoustic and visual separation at all times. This is important, because acoustically isolated birds produce acoustically distinct vocalizations (dialects, Farabaugh et al., 1994). From colony L, four birds took part (2 male, 2 female; females: Dino, Gretchen; males: Zeilinger, Ice), from colony R, eight birds took part (4 male, 4 female; females: Summer, Storm, Lemon, Famine; males: Fourier, Moonbeam, Hades, Hilbert).

Apparatus and Procedure

We used a place preference aviary with three perches, each equipped with an infrared beam and an adjacent speaker. Disrupting an infrared beam on the perches for over 300 ms (i.e. landings) triggered the playback of a sound from the adjacent speaker. One speaker-perch combination played high-arousal (“HA”) sounds, one low-arousal (“LA”) sounds, and the third one remained silent (“SIL”) as a control. Stimuli positions were counterbalanced across sessions and birds. A fourth speaker located under the testing aviary played sounds from the subject’s home aviary to minimize stress from isolation. Food and water were provided at the bottom of the test aviary ad libitum.

Each bird completed three sessions. We recorded the time spent on each perch per landing, the duration of stimulus playback for each perch, the number of landings per perch. Following the 2-hour sessions, the birds were returned to their home aviary. For a detailed description of the apparatus and procedure including, e.g., technical specifications of the speakers, duration of fade-out ramps, see supplementary Materials S0 or Bellmann et al. (2025b).

Data Analysis

Data from the perches was used to determine the time birds spent with the three conditions. We were only interested in the time spent with the three experimental conditions; the time spent not on any of the perches (e.g., transitions between the perches, feeding, flying in the cage) was thus excluded from further analysis. Sessions where birds did not listen to all sounds at least once were excluded from analysis both because they cannot reflect an informed decision about which sounds the bird wanted to hear and because these sessions had the tendency to skew the overall distribution across sound conditions.

We analyzed our data using R (4.4.1, 2024-06-14) and RStudio (RStudio 2025.05.0+496), employing generalized linear mixed-effects models (glmmTMB package, Brooks et al., 2017; McGillycuddy et al., 2025). All models included subjects as a random effect. The fixed effects (reference levels underscored) included condition (HA/LA/Silent), sex (Female/Male), colony (L/R), and session (1/2/3) as well as their interactions. If a model with interactions failed to converge, we first simplified towards an additive structure. If issues persisted, we singled out and removed the factor which least impaired the model performance.

Sex was included as a factor because budgerigars have sexual dimorphisms in the auditory domain that are known to influence auditory preferences (Tu et al., 2011; Hoeschele & Bowling, 2016). Colony was included as a factor because budgerigars are known to have vocal phenotypes that can differ between groups (Farabaugh et al., 1994). Our birds came from two colonies that were kept entirely (including acoustically) isolated from one another, which could influence auditory preferences.

Model fit was assessed using the ‘simulateResiduals’ function in the DHARMa package (Hartig, 2024). Statistical significance was assessed through likelihood-ratio tests conducted using the ‘anova’ function in the car package. Follow-up contrasts were conducted using the ‘emmeans’ package (1.10.4; Lenth & Piaskowski, 2025) and the Tukey method to account for the increased risk of type I error resulting from multiple comparisons. The adjusted p-values were calculated to determine significant differences between conditions, with a significance level of α = 0.05.

Experiment 1 - Musical Sounds, Budgerigars

Participants

Two of the birds (Fourier, and Gretchen; 1M and 1F respectively) had explored all options by the second session, but not in the first session. These birds may have been unaware of all conditions in session 1. Based on our exclusion criterion, we excluded the session 1 data of these birds from further analysis. Data for another bird, Famine, had to be removed entirely because she did not explore all options until Session 3, during which a technical malfunction had occurred.

Stimuli

We modeled HA and LA versions of harmonic tones that were varied in pitch and timbre. Across mammalian and avian species, pitch- and timbre-related acoustic features correlate with arousal. Principal pitch cues to high arousal are increases in the f0 and of the f0 range (Filippi et al., 2017; Briefer, 2012, 2020; Perez et al., 2012; Geberzahn & Aubin, 2014). Important timbral cues to high arousal are increases of spectral noise (i.e., amount of noise in the signal, noise occluding the harmonic structure; Briefer, 2012, 2020; Blesdoe & Blumstein, 2014), increases of the spectral centroid (SC, Filippi et al., 2017; Briefer, 2012, 2020), and faster attack (Eerola et al., 2012; McAdams et al., 2017).

All tones consisted of sinewave stacks – a fundamental frequency and the first 3 odd harmonics (f3 = 3*f0; f5 = 5*f0; f7 = 7*f0) which were 3 dB quieter than the f0 – plus manipulations.
For the pitch manipulations, we created seven 8-note sequences ascending or descending in f0 for the HA and LA condition, respectively, by using a set of 14 tones (Table 1) from a harmonic series tuned to a fundamental of approximately D3 (146.67 Hz), starting at the second harmonic (440 Hz). As shown in Table 2, for the HA condition, the sequence moved seven steps upward on the scale starting from tones 1-7 and ending on tones 8-14; For the LA condition, the sequence moved seven steps downward starting from the tones 8-14 and ending on tones 1-7. All tones lie well within the hearing range of budgerigars (Farabaugh et al., 1998; Dooling & Saunders, 1975; Jackson et al., 1999).

For the timbre manipulations, we changed the tones in noisiness, SC, and attack using the software Audacity (v.3.7.1), adding either -3 dB unfiltered white noise and 15 ms onsets in the HA conditions; or -12 dB white noise with 200 ms onsets, which was low-pass filtered (3-kHz cutoff, -20 dB) in the LA condition.

Table 1

Tones used for the sequences

Tone no	1	2	3	4	5	6	7	8	9	10	11	12	13	14
f0 (Hz)	440	586	733	879	1025	1172	1318	1464	1610	1757	1903	2049	2196	2342

Note. ‘f0’ = fundamental frequency. ‘Hz’ = Hertz.

Table 2

Ascending (a, numbers 01-07) and descending (b, numbers 08-14) sequences

(a)															(b)
01 →	440	586	733	879	1025	1172	1318	1464							← 08
02 →		586	733	879	1025	1172	1318	1464	1610						← 09
03 →			733	879	1025	1172	1318	1464	1610	1757					← 10
04 →				879	1025	1172	1318	1464	1610	1757	1903				← 11
05 →					1025	1172	1318	1464	1610	1757	1903	2049			← 12
06 →						1172	1318	1464	1610	1757	1903	2049	2196		← 13
07 →							1318	1464	1610	1757	1903	2049	2196	2342	← 14

Note. Numbers in the leftmost and rightmost column denote the number of the sequence, all other numbers denote the fundamental frequencies (in Hz) of the tones. Sequences 1-7 (a) are read from left to right and represent the HA condition, sequences 8-14 (b) are read from right to left and represent the LA condition.

All tones had a duration of 500 ms including a linear 50-ms fade at the end. Each stimulus was made from 8 concatenated tones with no silences, generating sequences that were 4-seconds in length. After the last tone of a sequence, the next sequence started after a delay of 1 second. All sounds were RMS normalized and were played at 70 dB SPL, measured at the middle of the perch (digital sound level meter, AZ Instrument, AZ 8922).

Figure 1 shows the amount of time birds spent in total (A) and per trial (B) with each of the three conditions (HA/LA/SIL). A trial is every instance in which a perch was contacted such that the infrared barrier was disrupted for longer than 300 ms (i.e., every perch landing that resulted in sound playback). The mean number of perch switches per session was 100.29 times and the mean overall time not spent on any perch was 10515.13 s (48.68 % of session times). The mean number of initiated trials, time per trial, and the overall time is shown in Table 3.

Table 3

Mean number of trials initiated, time per trial, and overall time birds spent with each stimulus in Experiment 1

cond.	mean number of trials			Mean time per trial			Mean overall time
	f&m	f	m	f&m	f	m	f&m	f	m
HA	94	79	105	93	63	111	3181	1788	4399
LA	89	55	118	90	145	68	2924	2838	2999
SIL	103	79	123	141	152	135	5494	4614	6209

Note: Cond = condition (HA = High arousal; LA = Low arousal; SIL = Silent); f = female, m = male. All times are in seconds. N trials are summed, across all sessions.

In Figure 2, the relative amount of time spent with the different conditions is shown for each individual.

Figure 1

Total time (A) and mean time per trial (B) with the different conditions (dots; HA = High arousal, LA = Low arousal) and standard error (bars) for both sexes together or separated by sexes (A* and B*) during Experiment 1. All times are in seconds.

Figure 2

Experiment 1: Raw relative time spent on the different perches per bird

Regarding the overall time, the model (formula: time ~ condition + sex + colony + session + (1|subject)) revealed that females (EMM = 2643, SE = 443, 95% CI [1902, 3672]) had a lower total time than males (EMM = 4213, SE = 628, 95% CI [3146, 5642]; χ²(1) = 4.42, p = .0355) and showed significant effects of condition (χ²(2) = 7.35, p = .0254). The effects of session (χ²(2) = 0.2, p = .9068) and colony (χ²(1) = 2.61, p = .106) were not significant. Post-hoc pairwise comparisons showed that HA was avoided compared to SIL (p = .0358), while there was no significant difference between the time spent with LA compared to SIL (p = .0762), and no difference between HA and LA (p = .9847).

Regarding the times per trial, the model (formula: time ~ condition:sex + session + colony + (condition|subject)) revealed significant effects of colony (χ²(2) = 18.13, p < .0001), with the left colony (EMM = 66.7, SE = 24.9, 95% CI [32,139]) spending less time per trial than birds from the right colony (EMM = 385.4, SE = 117.4, 95% CI [212, 700]). There was a significant effect of sex (χ²(1) = 7.08, p = .0078), with the time per trial for females (EMM = 98, SE = 38.9, 95% CI [45, 213]) being lower than for males (EMM = 262, SE = 95.2, 95% CI [129, 534]. The effect of condition (χ²(2) = 7.99, p = .0184) also was significant, while session (χ²(1) = 0.03, p = .8574) was not. There was a significant interaction effect between condition and sex (χ²(2) = 13.84, p < .001). Post-hoc pairwise comparisons between sexes per condition showed that males spent significantly more time than females with HA (p = .0078) and SIL (p = .0361). There was no significant sex difference for LA (p = .975). There was an opposite pattern regarding HA vs LA preferences: Males spent significantly more time with HA than with LA (p = .0368), while females, conversely, spent significantly less time with HA than with LA (p = .016). All other within-sex comparisons (SIL contrasted with LA or HA) showed no significant differences (all p > .05).

Regarding the number of trials initiated, the model (formula: number of trials ~ condition:sex + session + (condition|subject)) did not reveal any significant effects or interactions (all p > .05).

Finally, regarding the relationship between the number of trials and time per trial, Spearman correlations showed that there was a strong negative correlation between the number of trials and the time per trial (rs = –.6822, p < .0001) indicating that birds who switched more often spent less time per visit, while the overall time was similar irrespective of the number of switches (rs = –.1112, p = .538).

Experiment 2 - Animal sounds

This experiment investigated preferences for locations associated with either sounds with high or low arousal, but used vertebrate vocalisations to assess responses to biologically more plausible stimuli compared to the previous experiment.

Stimuli

We used 180 animal vocalizations from 9 species used in Filippi et al. (2017) whose arousal level was known and also shown to be consistently identified by humans (Filippi et al., 2017) and by a songbird species (Congdon et al., 2019). These stimuli included 20 calls of each of the following species: hourglass tree frog (Dendropsophus ebraccatus), American alligator (Alligator mississippiensis), common raven (Corvus corax), black-capped chickadee (Poecile atricapillus), pig (Sus scrofa), African bush elephant (Loxodonta africana), giant panda (Ailuropoda melanoleuca), Barbary macaque (Macaca sylvanus), and humans (Homo sapiens). For each species, 10 vocalizations were recorded in a HA context, and 10 measurably different vocalisations were recorded in an LA context (Filippi et al., 2017). All sounds had previously been RMS normalized and had a 5 ms amplitude ramp at the onset and offset to avoid transients (see Filippi et al., 2017 for more information).

Results

Figure 3 and Table 4 show the time spent in total (A) and per trial (B) for the three conditions (HA / LA / SIL). The mean number of switches per session was 136.44; the mean overall time not on any condition was 5560.615 s.

Table 4

Mean number of trials, time per trial, and overall time birds spent with each stimulus in Experiment 2

cond.	mean number of trials			Mean time per trial			Mean overall time
	f&m	f	m	f&m	f	m	f&m	f	m
HA	1504	489	1015	24	40	16	3037	3219	2783
LA	1614	534	1080	23	33	18	3124	2942	3306
SIL	1830	647	1183	57	84	42	8686	9014	8358

Note: Cond = condition (HA = High arousal; LA = Low arousal; SIL = Silent); f = female, m = male. All times are in seconds. The number of trials are across all sessions.

Figure 4 shows the relative amount of time spent with the different conditions for each individual.

Figure 3

Total time (A) and mean time per trial (B) with the different conditions (dots; HA = High arousal, LA = Low arousal) and standard error (bars) for both sexes together or separated by sexes (A* and B*) during Experiment 2. All times are in seconds.

Figure 4

Experiment 2: Relative time spent on the different perches per bird

Regarding the total time, the model (time ~ condition:sex:colony + poly(session,2) + (1|subject)) showed significant main effects of condition (χ²(2) = 7.0224, p = .0299) and showed that females (EMM = 1460, SE = 260, 95% CI [1030, 2070]) spent more time on the perches than males (EMM = 1271, SE = 234, 95% CI [886, 1822]; χ²(1) = 7.1315, p = .0076). There were also significant interaction effects of condition:sex (χ²(2) = 9.7324, p = .0077), sex:colony (χ²(1) = 5.5813, p = .0182) and condition:sex:colony χ²(2) = 10.2504, p = .0059). All other effects or interactions were not significant. Post-hoc comparisons with pairwise contrasts showed that there was no difference between HA and LA, while the time spent with SIL was significantly higher compared to both HA and LA (both p < .0001). Pairwise contrasts between conditions separated by sex showed the same overall pattern, with no differences between HA and LA for both females and males, and significantly more time with SIL compared to both HA and LA for both sexes. Females spent more total time with HA than males (p = .0478), while there were no differences for LA and SIL. For the interactions involving colony and sex, we were not able to conduct post hoc comparisons because in some cases, this would have meant conducting comparisons between as little as two individuals.

Regarding the time per trial, the model (time ~ condition:sex:session+(1|subject)) revealed significant main effects of condition (χ²(2) = 9.1148, p = .0105), the interaction of sex:condition (χ²(2) = 8.0364, p = .0179) and of condition:sex:session (χ²(4) = 13.3560, p = .0097). All other effects or interactions were not significant. Post-hoc comparisons with pairwise contrasts among conditions showed that there were no differences between HA and LA, and this remained the case when conducting comparisons separated by sex. SIL was consistently preferred compared to both HA and LA for both sexes together, for females, and for males (all p < .01). Comparisons factoring in session showed that females in session 1 avoided LA compared to SIL (p = .0072), but not HA vs SIL (p = .315). Across all other session/sex combinations, HA and LA were either both significantly different to SIL or both not different to SIL. There were no differences between sexes for any condition and session (all p > .05), except for session 3, where females spent more time with SIL than males (p = .0221). There were no differences in HA between sessions in both sexes (all p > .05). However, males’ time with LA in was greater in session 1 compared to both session 2 (p = .0004), and in session 3 (p = .0036), but there was no difference between session 2 vs session 3 (p = .6287).

Regarding the number of trials, the model (number of trials ~ condition + sex + session + colony + (1|subject)) showed an effect of session (χ²(2) = 13.57, p = .0011), while the effects of condition, sex, and colony were not significant. Post-hoc comparisons between sessions showed that the number of trials was higher for session 2 compared to session 1, (p = .0007), while there were no other significant differences (session 2 vs session 3, p = .0892; session 1 vs session 3, p = .2702).

Regarding the relationship between the number of trials and time per trial, Spearman correlations showed that birds who spent more time per trial switched less often (rs = -0.5582, p = .0004), while the overall time was similar irrespective of the number of switches (rs = .0474, p = .7838).

Lastly, to assess whether the birds’ responses towards the conditions were consistent across the two experiments (i.e., to determine whether synthetic vs natural made a significant difference for a birds’ behavior in the apparatus), we first compared the relative total times per condition and subject using paired Wilcoxon tests. The results suggested that the behavior was consistent across experiments in terms of time per condition on a subject level (HA, Exp. 1 vs HA, Exp. 2; LA, Exp. 1 vs LA, Exp. 2; etc); There were no significant differences between the experiments in the relative time a subject spent on each condition. A Wilcoxon test directly comparing the ratio between HA and LA per subject also found no differences across experiments (p = .2785). Next, we compared the number of trials per condition using paired Wilcoxon tests, which tested for the number of trials in Experiment 1 being less than in Experiment 2, as indicated by Tables 3 and 4. The results showed that the number of trials was significantly lower in Experiment 1 than in Experiment 2, holding for HA (p = .0098), LA (p = .0141), and SIL (p = .0029).

This study investigated budgerigars’ preferences for locations associated with either high aroused sounds, low aroused sounds, or silence, using time spent with the conditions as a proxy for preference. To understand the potential contributions of specific acoustic features and the type of sound source to place preference, we conducted one experiment using controlled sounds from synthetic sources and a second experiment with audio recordings from heterospecific vertebrate individuals.

While we expected a preference of low arousal over high arousal regardless of the type of sound source or the listener’s sex, our results show that both of these factors had a marked impact on the observed preferences. Specifically, with synthetic sounds in the first experiment, males spent more time per trial with high arousal than with low arousal and females spent less time with high arousal – i.e., preferences were inverted by sex. Using animal vocalisations, the second experiment revealed an overall avoidance of both sound types by both sexes.

Overall our hypothesis that across source types and sexes, budgerigars would show greater aversion of high arousal, reflecting adaptive responses to acoustic cues of threat, was not clearly supported. Exploiting cues can inform an animal about the state of its environment, opening up behavioral options that may increase survival. But how an individual species uses those cues to reach the goal of keeping safe may depend on multiple factors. In the following sections, we propose that a combination of factors – that is, the birds’ sex, its ecological niche, and its lifetime experience – might be better suited to explain how individuals react to arousal cues, rather than assuming a fixed response to arousal cues; always avoiding high arousal and attending to low arousal.

The sex of the listener did influence the appreciation of high arousal versus low arousal. With synthetic sounds, it became apparent that males – contrary to our prediction – preferred high over low arousal sounds. Females, conversely, preferred low arousal over high arousal, as predicted. One interpretation of this pattern is that males might have been more aggressive or thrill-seeking, while females showed a more avoidant approach. However, attributing this finding solely to a simple dichotomy of male aggression versus female defensiveness presents considerable challenges, as the published literature does not provide a definitive framework for this pattern and explanatory guidance remains limited.

Specifically, what has been found is mixed: In the visual domain, female budgerigars were found to be more aggressive towards conspecific intruders than males (Boiten et al., 2023), while in the auditory domain, male budgerigars were more inclined toward atypical stimuli (Hoeschele & Bowling, 2016), female budgerigars show a more avoidant reaction to alarm calls than males (Nicolas et al., 2004) and males show lower responsiveness to mate calls after separation (Eda-Fujiwara et al., 2016). More generally, there is no conclusive evidence for more pronounced novelty- and thrill-seeking behaviour based on sex in another parrot (Monk parakeet, Kerman et al., 2018), but also not across animal species more generally (Harrison et al., 2022).

So although the dimorphism observed in our first experiment brings to mind previous preference experiments in budgerigars, which identified a similar opposite pattern between females and males in the rhythmic domain (Hoeschele & Bowling, 2016), it is unclear what it might be grounded in. Some sex-dependent auditory preferences were suspected to be grounded in sexual behavior, such as the rhythm preferences due to rhythmic courtship displays performed by males (Hoeschele & Bowling, 2016), and spectral preferences in females due to warble song imitation by males as another courtship display (Moravec et al., 2010).

While we could not determine here which functional background accounts for the sex differences we observed in Experiment 1, we suggest that our findings might serve as a starting point for future experiments. Specifically, our results show that there are sex differences in the evaluation of spectral characteristics related to arousal that remain to be explored in more detail, including the underlying functional causes. However – turning to the other potential factors that contribute to preferences –, the sex differences between HA and LA appear to have vanished as the birds were presented with animal sounds. The initial sex differences might have been overridden by other information, such as cues to potentially dangerous animals. In addition, budgerigars may use a strategy different from animals less prone to predation in the approach to potentially threatening heterospecifics.

Budgerigars generally show high levels of neophobia, the fear of novel stimuli (ManyBirds Project et al., 2025). Baseline neophobia has a strong relationship with the ecological niche of the species, for example the position in the food chain: Animals in lower trophic positions, such as budgerigars, show higher baseline neophobia compared to those in higher positions (Crane & Ferrari, 2017). Other factors recently highlighted are the migratory pattern (migrating vs non-migrating) and a specialist vs generalist diet (ManyBirds Project et al., 2025). Because of these links between risk and neophobia, the latter is suggested to serve as an adaptive threat-aversion strategy (Brown et al., 2013; Thapa et al., 2024). Some of the budgerigars that were used in the current experiments had previous experience with experiments on octave equivalence (Wagner et al., 2020), rapid harmonic tone sequences (Bellmann et al., 2025b), and roughness preferences (Bellmann et al., in preparation), all of which used synthetic harmonic sounds, similar to the sounds in Experiment 1. As a result, our birds may be particularly familiar with – and not show neophobic responses to – synthetic stimuli. Since budgerigars occupy a rather vulnerable position in the food chain, a more reactive response to potential threats such as the comparably novel animal sounds in Experiment 2 (natural sounds from other vertebrates) could reflect adaptations to their ecological niche in the form of a ‘better safe than sorry’ strategy (Haftorn, 2000). Understanding the communication of emotion further thus may profit from research comparing neophilic and neophobic species’ responses to high aroused and low aroused sounds. We thus suggest investigating emotion perception across a wider range of animals with varying levels of neophobia as a next step for further research.

The captivity of the tested birds likely is another factor that had an impact on the stark avoidance of the animal sounds. Animals reared in captivity are especially neophobic compared to wild animals (Crane & Ferrari, 2017). Neophobia is a plastic trait that partially (ManyBirds Project et al., 2025) depends on experience (Elvidge et al., 2016). Animals with initially high baseline neophobia, especially in the wild, balance the risk of predation with the need for resource acquisition and exploration (Elvidge et al., 2016). Our results, collected from individuals reared in captivity, may thus reflect the high-neophobia end of responses to potential threats in budgerigars, with explorative or neophobic behavior depending on the level of potential threat (potentially higher for heterospecifics than for artificial melodies). Comparing wild and captive/captivity-reared budgerigars may thus be a next step to clarify heterospecific emotion perception. Specifically, we would expect wild budgerigars, while still exhibiting avoidant behaviour toward heterospecifics, to show a more nuanced assessment of heterospecific vocalizations than captive birds.

In addition, a recent study identified further factors shaping budgerigars’ behavioral responses to threat – though in a non-acoustic domain (Zhao et al., 2025). They showed that personality, sex, morphology, and social factors are all involved in threat responses. Social factors included social rank, the presence or absence of flockmates, and the familiarity of flockmates, all of which were decisive in the latency of approaches to food and the frequency of environmental scans under threat. In our study, birds were tested in isolation, which may be a particularly risk-avoiding situation. Taking into account social factors (e.g., rank, familiarity) in risk assessments in the acoustic domain would be an exciting next step. Taken together, evidence points to risk assessment in budgerigars as a complex process whose behavioural outcome depends on the interplay of a variety of biologically ingrained, individual as well as social factors.

Certainly, responding to vocal emotion is not as simple as ‘high arousal yields less attraction than low arousal’ – But also not as simple as ‘always better safe than sorry’. While there is evidence that arousal is recognized universally (Congdon et al., 2019; Debracque et al., 2023; Filippi et al., 2017; Greenall et al., 2022; Maigrot et al., 2022; Mason et al., 2024; Merkies et al., 2022; Thévenet et al., 2023), this doesn't mean that these cues to emotion are the only information present in signals that lead to attraction/repulsion. Subsumed under the term ‘aesthetic cognition’ (See Westphal-Fitch & Fitch, 2018 for a proposal), non-human animal behavior may be also driven by aesthetic judgements and the capacity to make choices based thereon, which includes perceiving and creating beauty (French, 2022; Welsch, 2004; Westphal-Fitch & Fitch, 2018). Appealing stimuli were suggested to play an important role in non-human animal behavior for various functional reasons, such as nest material selection (Guillette et al., 2016), mate choice (MacLaren, 2024), but potentially also for not immediately reproductive goals such as perceiving pleasure for its own sake (French, 2022).

Based on the insights produced by this investigation as well as their discussion with the consulted literature, we would like to conclude with a revision of our original hypothesis. As portrayed in Fig. 5, we suggest that processing auditory cues of emotion is led by a risk assessment of the stimulus in the light of its meaning for immediate survival (tied to a combination of stimulus features, familiarity, lifetime experience; guided by baseline neophilia and trophic position) and is, if an immediate threat to survival is not indicated, further guided by seeking opportunities for learning (attraction to stimulus features, potentially for vocal learning or foraging, serving later reproduction).

Figure 5

Uses of Auditory Emotion-Cues in Animals: A basic model

Based on this hierarchical, multifactorial conception of emotion processing, we predict that

The higher the trophic position and/or baseline neophilia in a species are, the more pronounced their behavioral differences across levels of perceived emotion are.

The more familiar a type of stimulus (e.g., synthetic harmonic tones, familiar species, flock-specific calls) to an individual is, the more pronounced differences are.

Sex differences – coming secondary to decisions related to survival – are more pronounced in species with sexually dimorphic acoustic courtship behaviors where certain cues may be more relevant to one sex than the other.

In social animals, the rank and social context might further contribute to preferences, with higher ranking individuals being less vigilant and showing less differentiation between high and low arousal than lower ranking ones when exposed to sounds together (Zhao et al., 2025).

Further testing of other species, also considering valence, which was not studied here given the difficulty in studying and pinpointing clear valence cues in non-human animals (Briefer, 2018), including budgerigars, could potentially paint an interesting picture about how universal cues of emotion are used across species. We invite other researchers to collaborate to further understand emotion processing across the animal kingdom by testing the hypotheses put forth above in a vast range of species. Future studies should consider a variety of taxonomic and social factors, including: Varying levels of neophobia and positions in the food chain (Ad prediction 1); A wider range of familiar and unfamiliar stimuli (Ad prediction 2); Species with vs without sexually dimorphic courtship behavior in the acoustic domain (Ad prediction 3); and social factors such as rank and isolation vs group settings during stimulus presentation (Ad prediction 4).

Ethics statement

The place preference paradigm used in this study is a noninvasive experimental technique and was carried out according Austrian animal protection and housing laws and were approved by the ethical board of the behavioural research group in the faculty of Life Sciences at the University of Vienna (approval number 2015-005). The participation of individual birds in the experiments was approved by an animal keeper at our Institute.

Competing Interests

The authors have no competing interests to declare that are relevant to the content of this article.

Funding Information

OTB is a recipient of a DOC Fellowship of the Austrian Academy of Sciences at the Acoustics Research Institute (ARI) under Grant No. 27276. MW was supported by a FEMtech grant (“BudgieConDis”) from the Austrian Research Promotion Agency (FFG). This work was supported by the Vienna Science and Technology Fund (WWTF) project ANIML (LS23-014) to MH.

Author Contribution

OTB : Conceptualization, Investigation, Resources, Visualization, Writing - Original Draft, Writing - Review & Editing, Project administration; MW: Investigation, Resources, Validation, Visualization, Writing - Review & Editing; MH : Conceptualization, Supervision, Writing - Review & Editing, Methodology

Acknowledgement

We would like to thank Barbara Götz and Barbara Timmler for taking care of the animals, Barbara Götz for helping with the experiments, and we would like to thank our non-human participants.

Data Availability

All data and analysis scripts can be found under the following OSF link: [https://osf.io/mqthr/overview](https:/osf.io/mqthr/overview)

Bellmann OT, Witt M, Wagner B, Hoeschele M (2025a), in press Humans but not Budgerigars (Melopsittacus undulatus) Consistently Prefer Sounds with Clear Harmonic Spectra. Music Perception
Bellmann OT, Hoeschele M, Witt M, Shair Ali E, Chu PC, Bianco R (2025b) Parrots and humans prefer pitch sequences of intermediate complexity. https://doi.org/10.1101/2025.06.22.660906. bioRxiv, 2025-06
Bellmann OT, Witt MH, Frank J, Hoeschele M (in preparation). Humans and Budgerigars Prefer Sounds with Low Roughness
Blesdoe EK, Blumstein DT (2014) What is the sound of fear? Behavioral responses of white-crowned sparrows Zonotrichia leucophrys to synthesized nonlinear acoustic phenomena. Curr Zool 60(4):534–541. https://doi.org/10.1093/czoolo/60.4.534
Boiten G, van Iersel R, Pinxten R, Eens M (2023) Females are more aggressive than males towards same-and opposite-sex intruders in the blue tit (Cyanistes caeruleus). Animals 13(4):585. https://doi.org/10.3390/ani13040585
Briefer EF (2012) Vocal expression of emotions in mammals: mechanisms of production and evidence. J Zool 288(1):1–20. https://doi.org/10.1111/j.1469-7998.2012.00920.x
Briefer EF (2018) Vocal contagion of emotions in non-human animals. Proceedings of the Royal Society B: Biological Sciences, 285(1873), 20172783. https://doi.org/10.1098/rspb.2017.2783
Briefer EF (2020) Coding for ‘Dynamic’ information: Vocal expression of emotional arousal and valence in non-human animals. In T. Aubin, N. Mathevon, N. (eds), Coding strategies in vertebrate acoustic communication (pp. 137–162). Springer. https://doi.org/10.1007/978-3-030-39200-0_6
Brockway BF (1964) Ethological studies of the budgerigar (Melopsittacus undulatus): Non-reproductive behavior. Behaviour 22(3–4):193–222. https://doi.org/10.1163/156853964X00021
Brooks ME, Kristensen K, van Benthem KJ, Magnusson A, Berg CW, Nielsen A, Skaug HJ, Maechler M, Bolker BM (2017) glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. R J 9(2):378–400. 10.32614/RJ-2017-066
Brown GE, Ferrari MC, Elvidge CK, Ramnarine I, Chivers DP (2013) Phenotypically plastic neophobia: a response to variable predation risk. Proceedings of the Royal Society B: Biological Sciences, 280(1756), 20122712. https://doi.org/10.1098/rspb.2012.2712
Congdon JV, Hahn AH, Filippi P, Campbell KA, Hoang J, Scully EN, Bowling DL, Reber SA, Sturdy CB (2019) Hear them roar: A comparison of black-capped chickadee (Poecile atricapillus) and human (Homo sapiens) perception of arousal in vocalizations across all classes of terrestrial vertebrates. J Comp Psychol 133(4):520–541. https://doi.org/10.1037/com0000187
Crane AL, Ferrari MCO (2017) Patterns of predator neophobia: a meta-analytic review. Proceedings. Biological sciences, 284(1861), 20170583. https://doi.org/10.1098/rspb.2017.0583
Darwin C (1872) The Expression of the Emotions in Man and Animals. London: John Murray. https://hdl.handle.net/11858/00-001M-0000-002B-06AC-B
Debracque C, Slocombe KE, Clay Z, Grandjean D, Gruber T (2023) Humans recognize affective cues in primate vocalizations: acoustic and phylogenetic perspectives. Sci Rep 13(1):10900. https://doi.org/10.1038/s41598-023-37558-3
Dooling RJ, Saunders JC (1975) Auditory intensity discrimination in the parakeet (M elopsittacus undulatus). J Acoust Soc Am 58(6):1308–1310. https://doi.org/10.1121/1.380813
Eda-Fujiwara H, Satoh R, Hata Y, Yamasaki M, Watanabe A, Zandbergen MA, Bolhuis JJ (2016) Sex differences in behavioural and neural responsiveness to mate calls in a parrot. Sci Rep 6(1):18481. https://doi.org/10.1038/srep18481
Eerola T, Ferrer R, Alluri V (2012) Timbre and affect dimensions: Evidence from affect and similarity ratings and acoustic correlates of isolated instrument sounds. Music Perception: Interdisciplinary J 30(1):49–70. https://doi.org/10.1525/mp.2012.30.1.49
Elemans CP, Rasmussen JH, Herbst CT, Düring DN, Zollinger SA, Brumm H (2015) … vec, J. G. Universal mechanisms of sound production and control in birds and mammals. Nature communications, 6(1), 8978. https://doi.org/10.1038%2Fncomms9978
Elvidge CK, Chuard PJ, Brown GE (2016) Local predation risk shapes spatial and foraging neophobia patterns in Trinidadian guppies. Curr Zool 62(5):457–462. https://doi.org/10.1093/cz/zow013
Farabaugh SM, Dent ML, Dooling RJ (1998) Hearing and vocalizations of wild-caught Australian budgerigars (Melopsittacus undulatus). J Comp Psychol 112(1):74–81. https://doi.org/10.1037/0735-7036.112.1.74
Farabaugh SM, Linzenbold A, Dooling RJ (1994) Vocal plasticity in budgerigars (Melopsittacus undulatus): evidence for social factors in the learning of contact calls. J Comp Psychol 108(1):81–92. https://doi.org/10.1037/0735-7036.108.1.81
Filippi P (2016) Emotional and interactional prosody across animal communication systems: a comparative approach to the emergence of language. Front Psychol 7:1393. https://doi.org/10.3389/fpsyg.2016.01393
Filippi P, Congdon JV, Hoang J, Bowling DL, Reber SA, Pašukonis A, Güntürkün O (2017) Humans recognize emotional arousal in vocalizations across all classes of terrestrial vertebrates: Evidence for acoustic universals. Proceedings of the Royal Society B: Biological Sciences, 284(1859), 20170990. https://doi.org/10.1098/rspb.2017.0990
French F (2022) Expanding aesthetics. Front Veterinary Sci 9:855087. https://doi.org/10.3389/fvets.2022.855087
Frick RW (1985) Communicating emotion: The role of prosodic features. Psychol Bull 97(3):412–429. https://doi.org/10.1037/0033-2909.97.3.412
Geberzahn N, Aubin T (2014) How a songbird with a continuous singing style modulates its song when territorially challenged. Behav Ecol Sociobiol 68:1–12. https://doi.org/10.1007/s00265-013-1616-4
Greenall JS, Cornu L, Maigrot AL, De La Torre MP, Briefer EF (2022) Age, empathy, familiarity, domestication and call features enhance human perception of animal emotion expressions. Royal Soc Open Sci 9(12):221138. https://doi.org/10.1098/rsos.221138
Guillette LM, Scott AC, Healy SD (2016) Social learning in nest-building birds: a role for familiarity. Proceedings of the Royal Society B: Biological Sciences, 283(1827), 20152685. https://doi.org/10.1098/rspb.2015.2685
Haftorn S (2000) Contexts and possible functions of alarm calling in the willow tit, Parus montanus; the principle of'better safe than sorry'. Behaviour, pp 437–449. https://www.jstor.org/stable/4535716
Harrison LM, Noble DW, Jennions MD (2022) A meta-analysis of sex differences in animal personality: no evidence for the greater male variability hypothesis. Biol Rev 97(2):679–707. https://doi.org/10.1111/brv.12818
Hartig F (2024) DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed) Regression Models. R package version 0.4.7. https://github.com/florianhartig/dharma
Hoeschele M, Bowling DL (2016) Sex differences in rhythmic preferences in the budgerigar (Melopsittacus undulatus): a comparative study with humans. Front Psychol 7:1543. https://doi.org/10.3389/fpsyg.2016.01543
Jackson LL, Heffner RS, Heffner HE (1999) Free-field audiogram of the Japanese macaque (Macaca fuscata). J Acoust Soc Am 106(5):3017–3023. https://doi.org/10.1121/1.428121
Kerman K, Sieving KE, Mary CS, Avery ML (2018) Social conformity affects experimental measurement of boldness in male but not female monk parakeets (Myiopsitta monachus). Behaviour 155(13–15):1025–1050. https://doi.org/10.1163/1568539X-00003519
Lenth R, Piaskowski J (2025) emmeans: Estimated Marginal Means, aka Least-Squares Means. https://CRAN.R-project.org/package=emmeans
Lohr B, Dooling RJ (1998) Detection of changes in timbre and harmonicity in complex sounds by zebra finches (Taeniopygia guttata) and budgerigars (Melopsittacus undulatus). J Comp Psychol 112(1):36–47. https://doi.org/10.1037/0735-7036.112.1.36
MacLaren RD (2024) Aesthetic preference for artificially selected color variant affects mate choice copying behavior in female Poecilia latipinna. PLoS ONE 19(3):e0298171. https://doi.org/10.1371/journal.pone.0298171
Maigrot AL, Hillmann E, Briefer EF (2022) Cross-species discrimination of vocal expression of emotional valence by Equidae and Suidae. BMC Biol 20(1):1–14. https://doi.org/10.1186/s12915-022-01311-5
Majdak P, Goupell MJ, Laback B 3-D localization of virtual sound sources: Effects of visual environment, pointing method, and training., Attention (2010) perception, & psychophysics, 72(2), 454–469. https://doi.org/10.3758/APP.72.2.454
ManyBirds Project, Miller R, Šlipogor V, Caspar KR, Lois-Milevicich J, Soulsbury C, Zanutto JA (2025) A large-scale study across the avian clade identifies ecological drivers of neophobia. PLoS Biol 23(10):e3003394. https://doi.org/10.1371/journal.pbio.3003394
Mason MA, Semple S, Marshall HH, McElligott AG (2024) Goats discriminate emotional valence in the human voice. Anim Behav. https://doi.org/10.1016/j.anbehav.2023.12.008
McAdams S, Douglas C, Vempala NN (2017) Perception and modeling of affective qualities of musical instrument sounds across pitch registers. Front Psychol 8:153. https://doi.org/10.3389/fpsyg.2017.00153
McGillycuddy M, Warton DI, Popovic G, Bolker BM (2025) Parsimoniously Fitting Large Multivariate Random Effects in glmmTMB. J Stat Softw 112(1):1–19. 10.18637/jss.v112.i01
Merkies K, Crouchman E, Belliveau H (2022) Human ability to determine affective states in domestic horse whinnies. Anthrozoös 35(3):483–494. https://doi.org/10.1080/08927936.2021.1999605
Moravec ML, Striedter GF, Burley NT (2010) Virtual parrots’ confirm mating preferences of female budgerigars. Ethology 116(10):961–971. https://doi.org/10.1111/j.1439-0310.2010.01809.x
Nguyen T, Flaten E, Trainor LJ, Novembre G (2023) Early social communication through music: State of the art and future perspectives. Dev Cogn Neurosci 63:101279. https://doi.org/10.1016/j.dcn.2023.101279
Nicolas G, Fraigneau C, Aubin T (2004) Variation in the behavioral responses of Budgerigars Melopsittacus undulatus to an alarm call in relation to sex and season. An Acad Bras Cienc 76:359–364. https://doi.org/10.1590/S0001-37652004000200025
Perez EC, Elie JE, Soulage CO, Soula HA, Mathevon N, Vignal C (2012) The acoustic expression of stress in a songbird: does corticosterone drive isolation-induced modifications of zebra finch calls? Horm Behav 61(4):573–581. https://doi.org/10.1016/j.yhbeh.2012.02.004
Riters LV, Polzin BJ, Maksimoski AN, Stevenson SA, Alger SJ (2022) Birdsong and the neural regulation of positive emotion. Front Psychol 13:903857. https://doi.org/10.3389/fpsyg.2022.903857
Scherer KR (1995) Expression of emotion in voice and music. J Voice 9(3):235–248. https://doi.org/10.1016/S0892-1997(05)80231-0
Singh M, Mehr SA (2023) Universality, domain-specificity and development of psychological responses to music. Nat Reviews Psychol 1–14. https://doi.org/10.1038/s44159-023-00182-z
Taylor AM, Reby D (2010) The contribution of source–filter theory to mammal vocal communication research. J Zool 280(3):221–236. https://doi.org/10.1111/j.1469-7998.2009.00661.x
Taylor AM, Charlton BD, Reby D (2016) Vocal production by terrestrial mammals: Source, filter, and function. In R. Suthers, W. Fitch, R. Fay, A. Popper (eds), Vertebrate sound production and acoustic communication, (pp. 229–259). Springer. https://doi.org/10.1007/978-3-319-27721-9_8
Thapa H, Salahinejad A, Crane AL, Ghobeishavi A, Ferrari MC (2024) Background predation risk induces anxiety-like behaviour and predator neophobia in zebrafish. Anim Cogn 27(1):69. https://doi.org/10.1007/s10071-024-01908-z
Thévenet J, Papet L, Coureaud G, Boyer N, Levréro F, Grimault N, Mathevon N (2023) Crocodile perception of distress in hominid baby cries. Proceedings of the Royal Society B, 290(2004), 20230201. https://doi.org/10.1098/rspb.2023.0201
Tu HW, Osmanski MS, Dooling RJ (2011) Learned vocalizations in budgerigars (Melopsittacus undulatus): the relationship between contact calls and warble song. J Acoust Soc Am 129(4):2289–2297. https://doi.org/10.1121/1.3557035
Wagner B, Bowling DL, Hoeschele M (2020) Is consonance attractive to budgerigars? No evidence from a place preference study. Anim Cogn 23(5):973–987. https://doi.org/10.1007/s10071-020-01404-0
Welsch W (2004) Animal aesthetics. Contemporary Aesthetics. (Journal Archive) 2(1):15. http://hdl.handle.net/2027/spo.7523862.0002.015
Westphal-Fitch G, Fitch WT (2018) Bioaesthetics: The evolution of aesthetic cognition in humans and other animals. Prog Brain Res 237:3–24. https://doi.org/10.1016/bs.pbr.2018.03.003
Zhao X, Su J, Lu M, Yu J, Wang H (2025) Which factors can influence budgerigars’ feeding behavior under predation risk: individual and/or social factors? Curr Zool zoaf069. https://doi.org/10.1093/cz/zoaf069

No competing interests reported.

SupplementaryMaterials.docx

Budgerigar Preferences Toward Universal Auditory Cues Of High and Low Arousal – A “Better Safe than Sorry” Strategy?

Status:

Version 1

Abstract

Figures

Introduction

General Methods

Results

Discussion

Uses of Auditory Emotion-Cues in Animals: A basic model

Declarations