All the methodological approaches used in this study were approved by the Ethics Committee of the IPEE.RAS (permit No. 31 dated 30.04.2019). The methods were noninvasive and did not require sedation or anesthesia of the animals.

To record the acoustic signals, the experimental animals were placed individually in a partially isolated pool (5*9*3.3 m) separated by a metal partition from the main pool. The acoustic connection between the pools was diminished but maintained. In this study, a synchronous two-channel recording was used to distinguish the acoustic activity of the experimental individual and the dolphins in the main pool. Video monitoring was also conducted simultaneously.

In the experimental pool, a recording hydrophone with a built-in preamplifier (B&K 8106) was placed and connected to a Nexus 2690 amplifier (B&K) and a USB-6251 data acquisition board (National Instruments). A control hydrophone (B&K 8104) was placed in the adjacent main housing pool and connected to the board through a custom-made amplifier based on the AD822AN chip. The hydrophones were positioned 50 cm from the pool wall, at a depth of approximately half a meter from the water surface.

The recording system was also based on a custom-made program based on LabView. The sound recording parameters were as follows: sampling rate of 126,000 S/s and a bit rate of 16 bits, with a maximum available recording frequency of 63 kHz.

Each experimental animal underwent one session of acoustic signal recording during individual isolation. Each session lasted 60 minutes. During the experimental session, the dolphins were allowed to swim freely in the isolation pool. They showed no signs of stress, as they were accustomed to staying alone in the pool during veterinary procedures.

Sound spectrogram analysis was conducted via Adobe Audition 1.5 and visualized in Python 3.9 (scipy library) with the following parameters: a fast Fourier transform block size of 512 points, a Hamming window function, a sampling rate of 126 kHz, 100% overlap along the frequency axis, and 50% overlap along the time axis. This configuration provided a frequency resolution of approximately 246 Hz and a temporal resolution (window length) of approximately 4.06 ms (with a time step between successive frames of 2.03 ms due to 50% overlap).

All the signals were initially classified by visual inspection of spectrograms into the three main categories: whistles, clicks and burst pulses. Signal classification was fulfilled by four experienced specialists, and they obtained similar results. All whistles were grouped by the contour of fundamental frequency into two categories: whistles with a flat whistle contour (flat whistles) and whistles with a frequency-modulated whistle contour (frequency-modulated whistles). Our identification of flat whistles was based on the categorization of fundamental frequency contours suggested by Hickey et al., 2009. Their classification was also used by Akkaya et al., 2023 for signals without inflection points, defined as “changes in the slope of the contour shape from positive to negative to negative to positive”. Signals of the three main categories were also differentiated from their combinations, as dolphins produced different types of signals, mainly whistles and clicks, simultaneously.

Finally, the signals were categorized into six types for subsequent analysis (see the examples of signals in Figure 2): clicks (Type 1), burst pulses (Type 2), flat whistles (Type 3), combined signals with a flat whistle component (Type 4), frequency-modulated whistles (Type 5) and combined signals with a frequency-modulated whistle component (Type 6).

For comparative analysis of signals with a flat whistle component, 30 signals from each dolphin were randomly selected. The following whistle parameters were analyzed: the fundamental frequency (ff, Hz) and the duration of the whistle component (L, seconds). The fundamental frequency was determined from the power spectrum in several points of the signal.

For comparative analysis of frequency-modulated whistles, 50 signals of Type 5 and Type 6 were randomly selected for each dolphin. Whistles and whistle components of combined signals were grouped in the samples of each animal due to the low number of uncombined whistles of Dolphin_1. All frequency-modulated whistles including the whistle components (Type 5 and Type 6) of the three dolphins were compared in terms of several parameters: duration (s), maximum frequency (Fmax, Hz), minimum frequency (Fmin, Hz), and frequency modulation coefficient (Q=Fmax/Fmin).

To test the possible prototype acoustic model of the shared signal type, the frequency of the trainer’s whistle (bridge signal) for Dolphin_3 (the only adult animal in the group) was compared to the frequencies of the whistle components of the signals of Type 4 produced by the experimental individuals. Dolphin_ 3’s bridge signal was chosen because infants do not participate in training procedures and thus do not have their own bridge signals.

In this study, the responses to rhythmic click trains in Dolphin_1 and Dolphin_2 were determined (Figures 3A, B).

In the normal-hearing subject (Dolphin_2), the click trains evoked a rhythmic AEP of the same frequency rate as the click rate (Figure 3A). The AEP amplitude was level dependent: the higher the stimulus SPL was, the higher the response, up to SPL of 120 dB re. 1 µPa. Fourier transforms of these waveforms revealed a definite peak at a frequency of 1 kHz (Figure 4A). Within a stimulus SPL range from 80 to 120 dB re. 1 µPa, this peak featured a dependence on the stimulus level: the higher the SPL was, the higher the peak amplitude was, up to 4.3 µV at 120 dB. At an SPL of 80 dB, the peak amplitude decreased to that of the background noise level. Stimulus SPLs greater than 125 dB were not tested to avoid overloading the hearing system of the normal-hearing subject.

In the hearing-impaired subject (Dolphin_1), the 1-kHz spectrum component did not exceed the background noise at stimulus SPLs ranging from 125 to 140 dB (Figure 4B). For this subject, stimulus SPLs lower than 125 dB were not tested because of the absence of a response at higher SPLs.

The dependences of the 1 kHz spectral peak on the stimulus intensity for Dolphin_2 and Dolphin_1 were as follows (Figure 5). For Dolphin_2, the response amplitude steadily decreased with reduced stimulus intensity, whereas for Dolphin_1, it remained subthreshold and was almost unaffected by increased stimulus intensity.

Each of the three dolphins produced the same three types of signals: whistles, clicks and burst pulses. The data analysis revealed an essential portion of combined signals, i.e., whistles produced simultaneously with clicks, in signalization of the deaf Dolphin_1, his paternal sister (Dolphin_2), and his mother (Dolphin_3). Furthermore, most of the signals of Dolphin_1 were combined. In addition, the whistle components of the combined signals, as well as the whistles, were either flat or frequency modulated. Thus, initially, all the signals of the three dolphins were classified into six types (see the examples of signals in Figure 2): clicks (Type 1), burst pulses (Type 2), flat whistles (Type 3), combined signals with a flat whistle component (Type 4), frequency-modulated whistles (Type 5) and combined signals with a frequency-modulated whistle component (Type 6).

Finally, the rates of these 6 types of signals in the signalization of the animals were compared (Figure 6). Overall, 1084 signals for Dolphin_1, 794 signals for Dolphin_2 and 773 signals for Dolphin_3 were analyzed.

According to the data presented in Figure 6, Dolphin_1 produced the same six types of signals as Dolphin_2 and Dolphin_3. On the whole, the series of clicks (Type 1) prevailed in the signalization of the three dolphins. The rate of burst pulses (Type 2) in the overall number of signals was the lowest for Dolphin_1 (6.4%) and the highest for Dolphin_2 (15.6%). The most prominent feature of the deaf dolphin’s signals was the very low percentage of whistles not combined with clicks, either flat (Type 3) or frequency-modulated (Type 5), approximately 0.2% and 2%, respectively. Among the signals other than clicks (Type 1) and burst pulses (Type 2), combined signals containing a frequency-modulated whistle component (Type 6) constituted the main portion of the signals of Dolphin_1, approximately 24%, compared with 1.8% in Dolphin_2 and 8.3% in Dolphin_3.

The ratio of whistles (Type 3, Type 5) to combined whistles (Type 4, Type 6) in the three dolphins was also examined (Table 1).

The data presented in Table 1 demonstrate the substantial distinction of Dolphin_1 from Dolphin_2 and Dolphin_3. In contrast to the two dolphins with normal hearing, more than 90% of the whistles of Dolphin_1 were whistle components of combined signals. Uncombined whistles constituted only 3% of all his flat whistles and flat whistle components of Types 3 and 4 (overall, 71 signals) and 7% of all frequency-modulated whistles and frequency-modulated whistle components of Types 5 and 6 (overall, 279 signals). Dolphin_1 did not produce any stereotyped whistle corresponding to the individually distinctive signature whistles well studied in bottlenose dolphins.

The ratios of Type 3 and Type 4 signals were similar for Dolphin_2 and Dolphin_3. Moreover, the percentage of Type 5 signals was greater for Dolphin_2 (87% of 107 signals) than for Dolphin_1 (7% of 279 signals) and Dolphin_3 (56% of 146 signals).

During the hour-long sessions of individual isolation, all three dolphins produced similar combined signals with a flat whistle component (Figure 7).

The fundamental frequency (ff, Hz) and the duration of the whistle component (L, seconds) were analyzed for 90 signals. The results of the analysis are presented in Figure 8; Table 2.

The results of the MANOVA revealed a significant effect of individual identity on the acoustic parameters (Wilks’ λ = 0.71, F4, 170 = 8.66, p<0.001): the deaf animal significantly differed from any other individual in whistle duration (Tukey’s HSD post-hoc test, Dolphin_1 vs Dolphin_2 p=0.012 Cohen’s d=0.82 corresponds to large effect; Dolphin_1 vs Dolphin_3 p<0.0001 Cohen’s d=1.40 corresponds to very large effect), while the dolphins without hearing disorder produced whistles of similar duration (p=0.123 Cohen’s d=0.45 corresponds to medium effect). For fundamental frequency, Dolphin_1 differed only from Dolphin_2 (p=0.03 Cohen’s d=0.64 corresponds to large effect).

A Linear Discriminant Analysis (LDA) was performed to assess individuality based on the two acoustic parameters of combined signals with a flat whistle component. The whistle duration contributes most strongly to the first discrimination function (LD1), whereas fundamental frequency is mainly associated with the second discriminant function (LD2) (Table 3) (Figure 9).

The discriminant analysis revealed that whistle duration plays the primary role in separating individuals. The first discriminant function (LD1) accounted for 78.1% of the total between-group variance, while the second function (LD2) explained 21.9%.

The Mahalanobis distance matrix (Table 4) shows that Dolphin_1 formed the most distinct group, being farther from the other two individuals (1.019 and 1.303), whereas Dolphin_2 and Dolphin_3 were more similar to each other (distance = 0.793).

The percentage of correct classifications was 56.7% for Dolphin_1, 41.4% for Dolphin_2 and 56.7% for Dolphin_3. Thus, signals with a flat whistle component were shorter for the deaf individual, than for the other two dolphins.

The coefficients of variation indicated low and nearly identical variability across individuals for the fundamental frequency. In contrast, the coefficients of variation were higher for whistle duration, with Dolphin_1 showing slightly lower variability compared to the other two individuals (Table 5). Generally, the variability of the acoustic parameters was comparable across individuals, with only minor differences observed for whistle duration.

The testing of the possible acoustic model, which served as a prototype for the shared signal, revealed that the frequency of the trainer’s whistle (bridge signal) for adult Dolphin_3 was 7384 Hz. As shown in Table 2, this frequency closely matches the median frequency of the whistle component of the dolphins’ shared signal.

The duration (s), maximum frequency (Fmax, Hz), minimum frequency (Fmin, Hz), and frequency modulation coefficient (Q = Fmax/Fmin) were analyzed for 150 frequency-modulated whistles, including whistle components (Types 5 and 6) (Table 6; Figure 10).

Because of a high correlation between minimum frequency and the modulation coefficient Q, the minimum whistle frequency was excluded from the MANOVA and LDA analyses. To achieve normality, logarithmic and square-root transformations were applied to the data. The results of the MANOVA revealed significant differences among the individuals (Wilks’ λ = 0.197, F6, 282 = 58.93, p<0.001), confirming strong inter-individual variability in signal structure. All animals differed from each other for all three parameters (Tukey’s HSD post-hoc test, p<0.0001, for Q, Dolphin_2 vs Dolphin_3 p=0.0002). The effect sizes were large to very large for all comparisons (Cohen’s d = 0.7–2.6), indicating strong differences in whistle duration, Q, and maximum frequency among individuals.

A Linear Discriminant Analysis (LDA) was performed to assess individual distinctiveness based on the three acoustic parameters of frequency-modulated whistles. Whistle duration and Q contribute mostly to the first discrimination function (LD1), whereas maximum frequency was mainly associated with the second discrimination function (Table 7; Figure 11).

The first discriminant function (LD1) accounted for 89.3% of the total between-group variance, while the second function (LD2) explained 10.7%.

Thus, Dolphin_1 was characterized by shorter and less stable signals with relatively lower modulation coefficient. Dolphin_1 was acoustically most distinct from Dolphin_3 and moderately separated from Dolphin_2 (Table 8).

The percent of correct classification was 76% for Dolphin_1, 78% for Dolphin_2 and 86% for Dolphin_3, respectively. The coefficients of variation demonstrated that Dolphin_1 exhibited the highest variability in signal duration, while maintaining comparatively stable frequency-related parameters (Table 9).

Using click trains as the stimuli, the hearing thresholds of Dolphin_2 were within the normal range and were assessed at 80 dB relative to 1 µPa. This estimate is higher than the tonal threshold measured by both behavioral and AEP methods, which may be as low as 50 dB at optimal sound frequencies. This difference is expected because of the short click duration (less than 0.5 ms), which is shorter than the duration of temporal summation in the auditory system of dolphins. Owing to the short duration, the click energy was rather low; therefore, the threshold was reached at a higher SPL. Additionally, it should be taken into consideration that the energy of the stimulation click was distributed along a frequency range as wide as approximately 100 kHz, so in each critical band of hearing, the stimulus energy could be rather low. Thus, in the normal-hearing dolphins, the obtained thresholds of 80 dB SPL may be assessed as normal for this species.

In contrast to Dolphin_2, which has normal hearing, the thresholds in the hearing-impaired subject Dolphin_1 could not be measured. With the stimulus SPLs as high as 140 dB, definite AEP responses exceeding the background noise were not recorded. This result indicates profound hearing loss. Therefore, Dolphin_1 was determined to be practically deaf.

Our data indicate that the deaf dolphin (Dolphin_1) produced the same types of signals as normal- hearing Dolphin_2 and Dolphin_3, but in different ratios. The dolphins emitted a large portion of the combined whistle-click signals, and their whistles and the whistle components of the combined signals had a flat or frequency-modulated contour of the fundamental frequency. All the signals were classified into six main types: clicks (Type 1), burst pulses (Type 2), flat whistles (Type 3), combined signals with a flat whistle component (Type 4), frequency-modulated whistles (Type 5), and combined signals with a frequency-modulated whistle component (Type 6). Compared with Dolphin_2 and Dolphin_3, Dolphin_1 produced the highest number of signals over the same period, the highest portion of clicks, the lowest portion of burst pulses (Type 2) and whistles not combined with clicks (Type 3 and Type 5).

The statistical analysis revealed distinctions among the deaf male and the normal-hearing females by several parameters. The duration of flat whistles was shorter and relatively stable in Dolphin_1 compared to Dolphin_2 and Dolphin_3. The frequency parameters of flat whistles were stable and did not differ in the animals. The frequency-modulated whistles of the deaf male were also shorter, but most variable in duration among the three dolphins. In contrast to variability in temporal parameters, the frequency-related characteristics of Dolphin_1’s signals remained relatively stable, and the modulation coefficient of his whistles was the lowest compared to the other two dolphins.

The very low number and rate of whistles that were emitted without clicks (Type 3 and Type 5) are of particular interest. The newborn dolphin observed by Killebrew et al (Killebrew et al., 2001). initially produced only burst pulses. He began to produce whistle-like signals only on the fifth day of life, and all of them were the elements of whistle-squawks. Thus, the very low number of pure whistles not accompanied by clicks could be a primitive infantile feature of Dolphin_1’s signalization. Furthermore, we could not select among his combined or uncombined frequency-modulated whistles any stereotyped whistle corresponding to individually distinctive signature whistles of bottlenose dolphins. According to the above data, signature whistles are not present at birth, are usually learned by calves by the age of 1 year, and their development strongly depends on the acoustic environment of infant dolphins (Caldwell and Caldwell, 1979; Caldwell et al., 1990; Sayigh et al., 1995, Sayigh et al., 2022).

Thus, our results provide some evidence of insufficient development of the generation system of tonal signals in Dolphin_1. We suppose that this could be a result of hearing deficits from birth or from the first months following the birth of this dolphin.

The intriguing aspect of this study was the discovery of a specific identical signal type among all three experimental dolphins—a flat whistle combined with simultaneous clicks characterized by a clear structure. In the hourly recordings, the dolphins emitted the shared signal at rates of 6,4% for Dolphin_1, 8,4% for Dolphin_2, and 4,1% for Dolphin_3.

The literature data (McCowan and Reiss, 1995; Tyack, 2008; Smolker and Pepper, 1999; Watwood et al.,2004; Jones et al., 2020) show that bottlenose dolphins in nature and in captivity can develop a shared group-specific whistle for communication with closely associated specimens. The development of a shared signal may be a result of convergence of parameters of group members’ signature whistles (Smolker and Pepper, 1999; Watwood et al., 2004) or a product of learning of a new shared whistle that differs from the individual signature whistles (Jones et al., 2020).

At the same time, owing to the fluent ability of bottlenose dolphins for vocal imitation of artificial sounds (Caldwell and Caldwell, 1972; Richards et al., 1984; Reiss and McCowan, 1993) calves born in captivity and exposed to artificial whistles can copy and incorporate them into their own signature whistles during the early postnatal period (Miksis et al., 2002).

Considering that the animals in our study were kept together and isolated in the same enclosure for an extended period of time (over a year after the birth of Dolphin_1), it seems quite possible that they developed a shared signal based on the trainer’s flat whistle (bridge signal). In search of an acoustic model for the group-specific signal, a comparison was made between the frequency of Dolphin_3’s trainer’s whistle (bridge signal) and the fundamental frequency of the whistle component of the shared signal. Dolphin_ 3’s bridge signal was chosen because infants do not participate in training procedures and thus do not have their own bridge signals. However, young dolphins can mimic natural and artificial sounds. The results of the acoustic analysis indicate that the fundamental frequency of the flat whistle component of the group-specific signal closely matches the frequency of Dolphin_3’s trainer’s whistle. Thus, the dolphins presumably copied the trainer’s whistle and incorporated it as a component of the shared combined signal.

It is impossible to establish by what means each of the animals has learned the shared signal—by copying the trainer’s whistle or the signals of the groupmates. Considering the specific combined character of this signal and the peculiarities of the deaf dolphin’s acoustic signalization, we may assume that Dolphin _1 was the first who used it. In any case, Dolphin_ 1 could hardly begin to produce the described above signal by chance. The presence of a shared group-specific signal in Dolphin_1’s vocal repertoire indicates that his auditory system most likely functioned for some time after birth.