The experiment was conducted according to the European Union Directive 2010/63/EU for animal experimentation and was approved by the local authority (Landesamt für Natur, Umwelt und Verbraucherschutz North-Rhine Westphalia, Germany). Fifteen male Long-Evans rats (Charles River, Italy) in total were obtained in a batch of 12 experimental animals (PND 40, Mw_eight = 320 g, at the starting day of the experiment) and 3 Juvenile rats [PND 28, M_weight = 290 g, at the starting day of the Social-Sucrose Preference Test (SSPT)], serving as social stimulus/reward. Experimental rats were housed in groups of N = 3 rats in standard Type IV Macrolon cages under a reversed 12:12 h light-dark cycle. The housing room was kept at a constant temperature of 22°C and a humidity of 60%. Throughout the experiment, all rats received standard laboratory rodent food, ad libitum, except for the Sucrose Discrimination Test (SDT) phase in which all experimental animals were limited in their food intake (food per rat per day: 22 g on weekdays and 25 g on weekends).

We used an eight-arm radial maze as previously adapted by Schönfeld et al. (2020), detached four arms to arrive at a cross/plus-maze setup (Figure 1A). The maze consisted of a central platform (36 cm diameter; so-called neutral zone in our design) and four arms (14 cm wide and 60 cm long) that extended from the central platform in an octagon platform. Each of the four arms was consistently associated with one single reward type: 3 arms with three different levels of a sucrose solution reward (see Figure 1A) and one arm with a social stimulus. To circumvent any spatial bias, we divided our subjects into two groups (A and B, per group = 6) with a different allocation of reward positions for each group. Notably, during any test day in the experiment, only 2 out of 4 arms were open at a time to provide a head-to-head preference test between two rewards. On the arm of the maze assigned to the social reward, an unfamiliar Juvenile rat could be placed in a fixed cylindrical restrainer built from metal bars and compact plastic for its floor and ceiling (Height: 25.5 cm, Diameter: 17 cm, Ugo Basile Sociability Cage). The restrainer was fixed on the maze at the end of the Juvenile’s arm, and the Juvenile could move around in this restrainer, and social contact through the openings between the bars was possible. On the arms allocated to non-social reward (i.e., different sucrose concentrations 2, 5, and 10%), sucrose solution was provided to the experimental animal in a cube plastic dish (8 × 8 cm) mounted at the end of each arm. Additionally, in order to facilitate spatial learning of the reward conditions in each arm over days, we included sandpapers (17 × 13 cm) in the entrance of each arm that the rats’ whiskers touch when entering the arms. The sandpapers had varying grades (Group A:2% [P800], 5% [P400], 10% [P150], and Juvenile [P1200], Group B: 2% [P150], 5% [P1200], 10% [P800], and Juvenile [P400]), following the findings of Guić-Robles et al. (1989) These authors have demonstrated that rats’ whiskers can discriminate between sandpapers with 200 and 25 grains/cm². To record the ultrasonic vocalizations (USVs), four ultrasonic microphones (Condenser Microphone CM16/CMPA, Avisoft Bioacoustics, Glienecke, Germany) were positioned via a microphone stand to approximately 20 cm on the right side of each reward dish and the restrainer (see Figure 1A).

Behavioral testing on the SSPT included three phases (see Figure 1B). In all phases of this study, experimental animals started the trials from the neutral zone facing not toward targeted arms in given condition. In the first habituation phase, all four arms were open and unbaited, and each experimental animal explored the maze for 10 min. This phase aimed to find out whether animals were inherently biased toward selecting one specific reward zone or sandpaper (see Figures 2A,B). The second phase of training was the Sucrose Discrimination Test (SDT), which was implemented to verify that the experimental animals could indeed distinguish among the three selected sucrose concentrations (2, 5, and 10%). Food deprived animals were tested on the SDT phase over 9 days in three repetitions of three different conditions. In each condition, only two arms were open, and rats chose to allocate their time between rewards on the maze in the following order of conditions: 2% vs. 5%, 2% vs. 10%, and 5% vs. 10%. Notably, each animal was tested in only one condition each day. Each test trial took 10 min; during this time, experimental animals could move freely in the two open arms and drink up to 20 ml sucrose solution per plastic dish at the end of each arm. Both dishes were filled with fresh sucrose solution for each new trial/experimental animal. After passing the SDT phase (Figure 2C), the experiment was continued to the SSPT phase. In this phase, over each trial with a duration of 10 min, the experimental animal could similarly move freely between two open arms: either to explore the arm baited with sucrose, or to investigate the Juvenile rat in the restrainer at the end of the Juvenile arm. Animals were tested once per day in three conditions (Juvenile vs. 2%, Juvenile vs. 5%, and Juvenile vs. 10%) spread out over the three SSPT testing days (see Figure 1B). To keep baseline motivation equal for both types of reward (social vs. non-social), food deprivation was stopped after the final SDT test day, and animals were allowed to recover weight over 2 days before starting the SSPT. For the remainder of the experiment, animals were kept ad libitum. Rats usually spend more time exploring novel conspecifics than familiar ones (Smith et al., 2015, 2017), suggesting that the value of social interaction dynamically decreases over days with increasing familiarity with the conspecific. To keep the novelty, and, hence, the value of investigation of the social stimulus similar across testing sessions, three different Juvenile rats were used in all three conditions of SSPT for each experimental animal. The order of the identities of these Juveniles was counterbalanced across experimental animals to exclude identity effects. All USVs from all trials over the two phases (SDT and SSPT) were recorded for the full 10-min trial duration, with the sampling rate set at 250 kHz.

For the recorded videos from all sessions, Ethovision (EthoVision XT version 11.5, Noldus) was used to track the animals’ position. Tracking settings were optimized separately for each different phase of the study (Habituation, SDT, SSPT). In the habituation phase, each arm was divided into two zones (Sandpaper zone and Reward zone) to check for any inherent bias for the different reward zones and sandpaper zones. For the SDT and SSPT phases, we used the time that the animals spent in the reward zones (see reward zones; Figure 1A). The time spent in the neutral zone was excluded from the analysis.

Acoustic analysis of the USVs was executed using the software Avisoft-SASLab Pro (Version 5.2, Avisoft Bioacoustics, Berlin, Germany). Spectrograms were generated with a fast Fourier transform (FFT)-length of 512 points and an overlap of 75% (Flat Top window, 100% frame size). Correspondingly, spectrograms had a frequency resolution of 390 Hz and a time resolution of 0.64 ms. In the setup, we recorded the USVs through 4 microphones, providing a four-channel spectrogram recording. The amplitude of the USVs differed depending on the distance between the animal and the different microphones (Supplementary Figure A). The microphone channel that recorded the largest amplitude was selected for labeling for each USV in the spectrograms. This channel differed between the conditions and minutes of the trial. The labeling phase was conducted by two trained, independent scorers who labeled and classified each USV based on its sonographic features (as in Wright et al., 2010). Notably, in the SSPT phases, calls could be emitted by both the experimental animal and the Juvenile social stimulus. In these analyses, we did not attempt to tease apart the source of these vocalizations but instead rely on within-subject comparisons of experimental animals to quantify differences.

The labeling phase consisted of two steps: calibration and final labeling. During the first step, two scorers became familiar (under the supervision of the expert scorers) with sonographic features of each of the 50 kHz USV subtypes (and 22 KHz) according to the classification suggested by Wright et al.(2010; for an overview of the different USV subtypes considered in this study, Figure 3F). They initially labeled USVs together to reach a consensus labeling scheme. After this calibration step, they separately labeled the same 400 USVs and, subsequently, compared their labeling match. In total, inter-rater reliability was high (Cohen’s kappa = 0.95), such that 94.3% of 50 kHz USV’s subtypes were labeled with the same category by both scorers. Due to technical problems, the USV files of the condition 2% vs. 5% and some animals (1,10,11,12) from the SSPT task were lost. Therefore, for all USV related statistical analyses, we only applied the USVs from 8 animals for both tasks. Thirty-two trials from SDTs’ phase, including 2 days (2 and 3) for conditions (2% vs. 10% and 5% vs. 10%), were labeled. For the USVs from the SSPT phase, the recordings from all three test days (N = 24 recordings in total) were labeled. Both scorers tagged half of all USVs from the same conditions (every odd minute of each trial).

When labels were assigned in Avisoft, through the self-written code in python, we exported the USV raw data (Avisoft SAS-Lab Pro’s output) to generate a time series of vocalization labels with a temporal resolution of 25 Hz, synchronized to the video stream and position data (Ethovision output). Thus, each 0.040 ms sample had a one-hot encoded binary label, corresponding to the presence/absence of each of the 50 kHz subtypes, 22 kHz or background/noise. We first looked at the summed frames spent vocalizing, including all rats, to establish inclusion/exclusion criteria. The 22-kHz USVs accounted for 23.3% of all samples with USVs, counted in ms spent vocalizing. This high proportion of 22 kHz frames is mainly caused by the naturally longer length of a 22 kHz USV compared to the length of a 50 kHz call. As the main goal of this experiment only covers the 50 kHz calls, no further analysis was conducted on the 22 kHz calls. Figure 3E shows the inter-individual variation in USV production, warranting a within-subjects approach that includes normalization to correct these inter-individual differences in calculating group contrasts (see below). During the labeling phase, 3.9% of all call frames could not be clearly labeled in any of the 14 categories of 50 kHz subtypes. These USVs with varying sonographic features were called Unclear (Un, and Supplementary Figure B) and excluded from USVs within-between analyses. After labeling all 50 kHz USVs, six subtypes (Step-Down, Inverted-U, Step-Up, Multi-Step, Downward Ramp, and Upward Ramp) were excluded because of their small incidence (<2% of all call frames [an arbitrary cut-off]). The selected call subtypes were thus: Trill, Flat, Complex, Composite, short, Flat-Trill-combination, Split, and Trill-with-Jump.

Our initial analysis focused on a Combined vocalization score (CVS), including all 15 subtypes (Including Un and excluding only the 22 kHz) per session to look for overall differences in vocalization rates between conditions. Here, we first summed up all frames the rats vocalized for each of the 15 subtypes in a certain zone and then divided that score by the time the animal spent in that zone, thus normalizing the vocalization time to the occupation time per zone, creating a normalized vocalization rate. As inter-individual differences resulted in a skewed distribution of normalized vocalization rates, we performed a log transformation on these CVSs to reduced skewness and facilitate visualization. To investigate if the number of vocalizations differed depending on the reward type (social vs. non-social) or sucrose concentration, we applied a two-way repeated-measures ANOVA for each task (SDT and SSPT) separately. Here, we considered the condition with two levels for SDT (conditions: 2 vs. 10% and 5 vs. 10%), three levels for SSPT (Juvenile vs. 2%, Juvenile vs. 5%, and Juvenile vs. 10%), and two levels reward zone (SDT: higher/lower sucrose and SSPT: Juvenile/Sucrose) as IVs, and the log of the CVS of each task as DV.

To zoom in to differences between subtypes, we performed a similar analysis pipeline per subtype: after excluding the 22 kHz, Un, and infrequent call subtypes (excluded calls), for the remaining eight categories, we again normalized the subtype-specific vocalization rate to the spatial occupancy per zone to calculate a subtype vocalization score (SVS). This SVS was thus calculated by summing the number of frames the rat vocalized a specific subtype (1 frame = 0.040 ms) in a given zone and dividing it by the time the animal spent in that zone.

As a within-subjects normalization step, from these SVSs, we calculated a delta SVS score to show the differences in vocalization rate between zones for a given subtype. The delta SVS score was calculated as follows: i) SVS score in the low sucrose zone subtracted from the SVS score in the high sucrose zone for SDT and ii) SVS score in the non-social reward zone subtracted from the SVS score in the social reward zone for SSPT. We used this deltaSVS to compare normalized vocalization rates between subtypes in a given condition (between-subtype analyses) and within a subtype, between conditions (within-subtype analyses).

In the between-subtype analysis, with these dSVS, we ran a Kruskal Wallis test per condition for the SDT and SSPT data, with the subtype as the IV and the dSVS score as DV for each condition.

In the within-subtype analysis, we performed a Wilcoxon Signed-Rank test for the SDT sessions, comparing the vocalization of the given subtype in two conditions [2% vs. 5% and 5% vs. 10%]) and a Friedman test for each subtype across the three SSPT conditions (Juvenile vs. 2%, Juvenile vs. 5% and Juvenile vs. 10%). For all statistical analyses, the significance level was set at p < 0.05, and all the post hoc tests p-values are Bonferroni- corrected for multiple comparisons.

To exploit the continuous range of sucrose solutions used in the SSPT, to look for a linear association between vocalizations and sucrose solution, we conducted two mixed linear models, one on total calls (CVS) and one on subtype-specific SVS. Both models entered Animals as random effects, Conditions (2% vs. Juvenile, 5% vs. Juvenile, and 10% vs. Juvenile) as fixed effects, and CVS/dSVS as the dependent variable.

All statistical analyses were carried out using SPSS Statistics (version 24; IBM, United States) and R 3.5.1 (R Core Team, 2018). We applied the following libraries in R: the tidyverse (Wickham, 2017), the haven psycho, the readxl (Wickham et al., 2019b), the tidyr (Wickham et al., 2019a), the tibble (Wickham et al., 2019a), the sjplot (Lüdecke, 2020) the ggstatsplot (Patil, 2018) and the rockchalk (Johnson and Grothendieck, 2019). Moreover, visualizations of some figures (Figure 2D and Supplementary Figure D) were made using Jupyter Notebook (Kluyver et al., 2016) through the packages matplotlib (Hunter, 2007), pandas (McKinney, 2010), and seaborn (Waskom, 2021). Remaining figures were created by Inkscape (version 0.92.1, Inkscape project, 2020). In order to run the synchronization of USV and Animals’ positions we used the packages fileinput (Sinha, 2017) numpy (Harris et al., 2020).

A between-group comparison did not find evidence for a difference in spatial/reward preference based on the maze layout for groups A and B (Supplementary Figures CA,CB). Similarly, an analysis of the habituation period did not find any evidence for a preference for a specific zone of reward [F(3, 33) = 1.35, p > 0.05; Figure 2A] or sandpaper zone [F(3, 33) = 1.6, p > 0.05; Figure 2B].

SDT. To determine whether experimental animals could indeed discriminate between different sucrose concentrations (i.e., 2, 5, and 10%), we conducted a two-way repeated-measures ANOVA with task condition and task repetition day as within-subject factors and percentage of the higher sucrose reward as DV. We found no significant main effect of task condition, suggesting that animals did not significantly differ in their preference for the sweeter sucrose solution across sessions with different levels of sucrose concentrations. We did observe a significant main effect of day [F(2, 22) = 15.2, p < 0.001, η_p² = 0.581]. Post hoc analysis revealed that animals preferred the higher-percentage sucrose solution significantly more in all conditions on day three (M = 81.7, SE = 2.8) compared to day two (M = 69.1, SE = 2.7, p < 0.05, d = 4.6) and day one (M = 63.3, SE = 2.7, p < 0.001, d = 6.8). The data thus showed that animals develop a clearer preference for the sweeter sucrose solution over days (Figure 2C), probably as a consequence of learning. There was no significant interaction effect.

SSPT. To assess whether animals expressed a significant preference between social and non-social rewards (with three different sucrose concentrations) in the social-sucrose preference test (SSPT), we conducted a one-way repeated-measures ANOVA on the percentage of time spent with the social reward (Juvenile zone). The results showed that preferences for the Juvenile differed significantly between conditions [F(2, 22) = 52.2, p < 0.001, η_p² = 0.826]. Post hoc tests revealed that the animals’ preference for the Juvenile increased significantly from the condition Juvenile vs. 10% (juv. pref: M = 19%, SD = 10%) condition to the Juvenile vs. 5% (juv. pref: M = 55%, SD = 15%, p < 0.001, d = 12.2) condition. There was a further but non-significant increase in Juvenile preference when reducing the sucrose concentration to 2%; in this condition, Juvenile preference was also significantly higher than in the Juvenile vs. 10% condition (juv. pref: M = 61%, SD = 13%, p < 0.001, d = 9.4). Three one-sample t-tests vs. indifference (50%) showed that animals preferred the social reward in Juvenile vs. 2% [M = 61.5, SD = 13, t(11) = 3.06, p < 0.05], were indifferent between Juvenile vs. 5% [M = 54.7, SD = 15, t(11) = 1.08, p > 0.05] and preferred the sucrose reward in Juvenile vs. 10% [M = 80.6, SD = 10, t(11) = 9.9, p < 0.001]. These results show clearly that animals indeed traded off interacting with a Juvenile to the consumption of sucrose and also that a preference for interacting with the Juvenile when sucrose levels were low (2%) could be reversed when confronted with a more preferred 10% sucrose solution (Figure 2D). These between-condition differences could be due to a change in time (%) spent at the sucrose reward, the social reward, or both. To quantify this, we investigated if the absolute time animals spent in each reward zone differed between different conditions. A repeated-measures ANOVA on the absolute time animals spent on social reward showed a significant effect of conditions [F(2, 22) = 33.2, p < 0.001, η_p² = 0.751]. Post hoc tests revealed that the absolute time that animals spent in the Juvenile zone in the condition of Juvenile vs. 10% (M = 97.7, SD = 55) was significantly less than in the condition Juvenile vs. 5% (M = 250, SD = 76, p < 0.001, d = 2.2) and the condition Juvenile vs. 2% (M = 259, SD = 64, p < 0.001, d = 2.7). There was no significant difference between the condition Juvenile vs. 2% and Juvenile vs. 5%. A second repeated-measures ANOVA on the absolute time animals spent with non-social rewards also showed a significant effect of the condition [F(2, 22) = 74.7, p < 0.001, η_p² = 0.872]. Here, post hoc tests revealed that the absolute time that animals spent in the sucrose zone in the condition Juvenile vs. 10% (M = 408, SD = 19) was significantly more than the condition Juvenile vs. 5% (M = 205, SD = 20, p < 0.001, d = 10.4) and the condition Juvenile vs. 2% (M = 159, SD = 15, p < 0.001, d = 14.5). No significant difference was found between the conditions Juvenile vs. 2% and Juvenile vs. 5%. As a follow-up analysis, a paired sample t-test per condition revealed that in Juvenile vs. 2%, the Juvenile side (M = 259, SD = 64) was significantly (p < 0.05, d = 1.7) preferred over the sucrose side (M = 159, SD = 53). In the condition Juvenile vs. 5%, animals were indifferent between the reward types (Juvenile: M = 250, SD = 76; 5% sucrose: M = 205, SD = 71). In contrast, in the condition Juvenile vs. 10%, the sucrose side (M = 408, SD = 67) was preferred significantly (p < 0.001, d = 5.6) over the Juvenile (M = 97, SD = 55) (see Figure 2E).

As indicated in “Materials and Methods” section, the 50 kHz USVs produced by experimental animals in the SSPT were labeled and further categorized into subtypes. Descriptive statistics were generated for each of the subtypes included in our analyses, along with within-condition and between-condition comparisons. We found that rats emitted vocalizations in a total of N = 7,252 call frames (290 s, combined SDT, and SSPT, 2.4% of total recorded frames, Figure 3A). After exclusion of 22 kHz calls, based on prevalence, we selected eight subtypes: Trill (Tr), Flat (Fl), Complex (Cx), Trill-with-Jump (Tj), Short (Sh), Flat-Trill-combination (Ft), Split (Sp), and Composite (Ce) for further analysis (Figure 3F). Six subtypes (Step-Down, Step-Up, Upward Ramp, Multi-Step, Inverted-U, Downward Ramp) were excluded from analysis due to their limited occurrence (<2% of calling time, Figure 3D). From the selected subtypes, Tr (27.2%), Fl (24.4%), Cx (11.5%), and Ce (11.3%) were the most prevalent, while Sh (5.5%), Ft (4%), Sp (3.4%), and Tj (2.2%) were least prevalent in both tasks (Figure 3D). Notably, we found Un calls (3% in the SDT task, Figure 3B) and (6% in the SSPT task, Figure 3C. For more details about Un calls, see section “Materials and Methods”).

SDT. In total, throughout the SDT, 2155 call frames were found in which the rats were vocalizing, and after exclusion of 22 kHz calls, from the eight selected subtypes, Fl (48%), Tr (13%), Cx (10%), Sp (6%), Sh (5%), and Ce (5%) were most prevalent while, Ft (2%), and Tj (0.06%), were least prevalent in SDTs’ conditions (Figure 3B). SSPT. In total, in the SSPT, 5097 call frames were found in which the rats were vocalizing, and after exclusion of 22 kHz calls, from these eight selected subtypes, Tr (33%), Fl (14%), Ce (14%), Cx (12%) were most prevalent while Sh (6%), Ft (5%), Tj (3%), and Sp (2%) were least prevalent (Figure 3C) in SSPTs’ conditions.

To determine if the number of frames that the rat vocalized was affected by sucrose concentration or type of rewards in the different conditions, we conducted a two-way repeated-measures ANOVA on the Combined vocalization score (CVS; the number of frames vocalized relative to the time spent in the visited zone, see section “Materials and Methods”) with condition and reward zone as factors, separate for SDT and SSPT.

SDT. The SDT analyses found a significant effect of condition on the CVS [F(1, 7) = 14.9, p < 0.01, η_p² = 0.680]. The main effect showed that the CVS was significantly higher in the condition 2% vs. 10% (M = 0.310, SE = 0.075) than in the condition 5% vs. 10% (M = 0.128, SE = 0.054; Figure 4A). The factor reward zone also had a significant effect on the CVS [F(1, 7) = 14.3, p < 0.01, η_p² = 0.672; Figure 4B]. The main effect showed that the CVS was, surprisingly, higher (p < 0.01) in the lower sucrose concentration zone (M = 0.268, SE = 0.065) compared to the higher sucrose concentration zone (M = 0.171, SE = 0.058). There was also a significant interaction effect of conditions and reward zones [F(1, 7) = 5.9, p < 0.05, η_p² = 0.459; Figures 4C,D). Post hoc comparisons showed that CVS was higher in lower-reward zones only for the condition 2% vs. 10%. In the zone of lower sucrose concentration (M = 0.407, SE = 0.093) the animals had a higher CVS (p < 0.05) than the condition 5% vs. 10% (M = 0.213, SE = 0.064, see Figures 4A,B). SSPT. For the SSPT task, we again performed a two-way within-subjects repeated-measures ANOVA. There was no significant effect of condition (Figure 4E), but we found a significant effect of reward type [F(1, 7) = 13.6, p < 0.01, η_p² = 0.658, Figure 4F]. Post hoc comparisons showed that the CVS was significantly higher in the Juvenile zone (M = 0.544, SE = 0.075) than in the sucrose zone (M = 0.313, SE = 0.067; p < 0.01). Furthermore, there was a significant interaction between condition and reward types [F(2, 14) = 5.1, p < 0.05, η_p² = 0.426, Figures 4E–I]. Post hoc comparisons showed that animals’ CVS in the Juvenile vs. 10% condition was significantly higher (p < 0.01) in the Juvenile zone (M = 0.685, SE = 0.121) compared to the sucrose zone (M = 0.297, SE = 0.064). No significant differences in CVS between reward zones were found for the Juvenile vs. 2% (p = 0.06) and Juvenile vs. 5% conditions (p = 0.07).

These results already indicate an interesting finding: while behavioral preferences shifted toward the sucrose reward zone with higher sucrose concentration, the vocalization rate showed the opposite trend, with increasing vocalizations recorded in the juvenile zone with increasing sucrose concentrations. We next investigated whether this pattern was present for specific subtypes and if there were differences between subtypes.

Comparing USV subtypes between and within conditions.

As one of the main questions of this study, we were interested in finding out if the different sucrose concentrations or different reward types were associated with a different vocalization palette across the 50 kHz USV subtypes. Here, we used the delta Subtype Vocalization Score (dSVS; see section “Materials and Methods”), indexing the relative difference in vocalization rates between reward zones in a given session for these analyses, as it accounts for normalization of inter-individual differences in absolute call rates.

SDT. We conducted a Kruskal Wallis test separately for each condition (2% vs. 10% and 5% vs. 10%) by taking the eight subtypes observed in the SDT as a factor and their dSVS as the dependent variable (DV). We found no significant difference in the dSVS between subtypes for any condition (Supplementary Figure D).SSPT. We similarly conducted a Kruskal Wallis test for each condition (Juvenile vs. 2%, Juvenile vs. 5%, and Juvenile vs. 10%). In the condition Juvenile vs. 5%, we found a significant difference [H (7) = 16.6, p < 0.05]. Post hoc pairwise comparisons showed a significant difference between dSVS of the subtypes Tr (median = 0.3) and Fl (median = −0.04), (Mann-Whitney U-test, p < 0.01) and dSVS of subtypes Tr and Sp (median = 0, p < 0.05; Figure 5).

SDT. this analysis was conducted to determine whether dSVS for a given subtype differed between conditions. The Wilcoxon Signed-Rank test results showed that the dSVS score of Tr was lower in condition 2% vs. 10% (median = −0.4) than in condition 5% vs. 10% (median = 0), Z = 2.1, p < 0.05). There was no other significant difference within any subtypes between conditions (Figure 6).

For the within-subtype analysis of call rates in the SSPT, we exploited the continuous nature of the sucrose concentration in a mixed linear model, estimating the relationship between sucrose concentration (in %) and dSVS with individual animals modeled as random effects. We first modeled the total call rate (all calls combined) using the Combined vocalization score (delta CVS; see section “Materials and Methods”). The mixed linear model showed a linear association between the delta CVS and the sucrose level (beta = 0.034, 95% CI [0.01–0.06], t(15) = 3.27, p < 0.01, R2 fixed effect = 0.208). This suggests that the difference in total vocalization time in the Juvenile over the Sucrose zone significantly increased with higher levels of sucrose concentration (see Figure 7A and Supplementary Table 1). We then modeled the sucrose concentration to delta SVS relationship in linear mixed models separately for each subtype. The models showed a significant association for the subtypes Tr (beta = 0.18, 95% CI [0.05–0.031], p < 0.05) and Ce (beta = 0.07, 95% CI [0.01–0.013], p < 0.05). This means that, for these two subtypes, the difference in the number of frames vocalized in the Juvenile over the Sucrose zone significantly increased with higher levels of sucrose concentrations (see Figure 7B and Supplementary Tables 2A,B for more individual model statistics).