Subjects
We observed two captive bottlenose dolphin groups at the Shimonoseki Marine Science Museum and Kagoshima City Aquarium. Shimonoseki Marine Science Museum housed five adult females and one mother–calf male pair during study term. In Kagoshima City Aquarium, we observed five adult females and two calves in total. The observed subjects were sometimes changed (Table S2). More detailed information about the facilities is provided by Yamamoto et al.24,30. Our study complied with the Ethical Guidelines for the Conduct of Research Animals by Zoo and Aquariums issued by the World Association on Zoos and Aquariums (WAZA) and the Code of Ethics issued by the Japanese Association of Zoos and Aquariums (JAZA). This study was approved by the Shimonoseki Marine Science Museum and Kagoshima City Aquarium.
Data collection
All dolphins at the Shimonoseki Marine Science Museum were observed between 0830 and 1730 from July 2012 to May 2013 (51 days, approximately 229 h). The dolphins were observed at the Kagoshima City Aquarium between 0900 and 1800 from July 2012 to February 2015 (87 days, approximately 558 h). The observations were adjourned for approximately 30 min during feeding on five or six occasions in both the aquariums. The behavioral data collected were of dolphins that were older than one year.
Definitions of behavior and classification followed that of our previous studies24,30. Aggressive behavior included chasing, biting and hitting, and excluded playful interactions. Former opponents were classified as the aggressor who was the last attacker, and the victim was classified as the other former opponent. We recorded whether counter aggression occurred in the aggression and whether the aggressor or victim was a calf. Individuals under the age of six were considered calves, because dolphins tend to remain with their mother until they are three to six years old21. We recorded individuals who were closest with the former opponents at the end of aggression. Vocalizations and bubble streams during aggressions were recorded at Shimonoseki Marine Science Museum. We recorded whether dolphins produced whistles or burst-pulses during the aggressions, and whether aggressors and victims emitted bubble streams when whistles or burst-pulses were recorded. Whistles and burst-pulses were identified by the first author listening to the vocalization. If extraneous noises were great, we removed the data of aggression for this recording.
Post-conflict (PC) observations were conducted for 10 min immediately after the original aggression39. When renewed aggression between the same former opponents occurred within 1 min after the original aggression, PC observation was cancelled and a new observation was started after the end of aggression. We made only one PC observation of each pair per period between feeding times. If the initiator of post-conflict affiliation between former opponents and bystanders was not clear, the PC was removed for each analysis. During PC, we recorded affiliations initiated by bystanders toward aggressors or victims, affiliation between former opponents and affiliation between group members (not former opponents), and whether the bystander was a calf. In analyzing whether the occurrence of reconciliation affected the occurrence of bystander affiliation, recording of the focal bystander affiliation was extended for the period of reconciliation (for example, if reconciliation continued for 4 min, the period that observation of bystander affiliation was set at 14 min after the aggressions) because former opponents were unable to receive affiliation from bystanders during reconciliation. For the recording of post-conflict affiliation between group members, the target member was a randomly selected individual, who (1) did not initiate bystander affiliation to former opponents, (2) was not the aggressor or victim, and (3) was close to former opponents at the end of aggressions. We recorded whether the targeted member initiated affiliation toward group members during the aggression and PC. We included flipper-rubbing, contact swimming, and synchronized swimming as affiliations35,40,41.
Post-conflict attack by aggressors or victims to group members was recorded in the period after bystander affiliation and unaffiliated PC. When bystander affiliation occurred, we recorded the occurrence of PC attack for 10 min after bystander affiliation. We excluded PCs in which there were more than two types of post-conflict affiliation to show the effect of each bystander affiliation. In unaffiliated PCs, the recording period was set as one of three types for each analysis: (1) to compare with MC, the attack was recorded for 10 min after the original aggression; (2) to compare to the period after bystander affiliation toward aggressors, we recorded the attack for 2–12 min after the aggression; (3) to compare to bystander affiliation toward victims, the attack was recorded for 4–14 min after the original aggression. As PC attack was reportedly likely to reduce over time in previous studies on primates4,5, we delayed the start time of unaffiliated PC at average times from the end of aggression to the end of bystander affiliation.
Matched-control (MC) observation was set at 10 min from the same time the PC began on the next possible observational day of the corresponding PC. When any aggression occurred within 10 min before scheduled MC, the MC observation was cancelled and restarted after aggressions. If MC observation did not start within 40 min from the time that the corresponding PC started, the MC observation was carried out on the next day. During MC, we recorded whether former opponents attacked group members, and whether the focal member initiated affiliation toward group members.
We recorded the occurrence of synchronized swimming every 10 min to elucidate the affiliation relationship between two individuals. However, the data obtained 10 min after aggressions were excluded from this calculation. The affiliative index was calculated as (XAB/YAB), where XAB is the number of periods in which synchronized swimming between individuals A and B occurred, and YAB is the number of periods in which both individuals A and B were present. The affiliative index in Kagoshima City Aquarium was separately calculated before and after the birth of the second infant because the behavior of the adults differed significantly. The 10-min observational periods totaled 972 (all pairs, respectively) in the Shimonoseki Aquarium, and the average was 823 (min: 243 periods, max: 1173 periods) during the earlier period and 604 (min: 111 periods, max: 1240 periods) during the later period in the Kagoshima City Aquarium. Behavioral and vocal data were collected using both observations and videos (Handycam HDR–CX; Sony, Tokyo, Japan), which were recorded by the first author from underwater windows.
Statistical analyses
We investigated whether attacks by former opponents to group members increased after the original aggression using the GLMM. The occurrence of attacks toward group members was set as the dependent variable with a binomial error structure and logit link function. The predictor variable was PC or MC. The identities of the aggressor, victim and aquarium were included as random effects. To explore whether bystander affiliation reduced PC attacks on group members, we compared the occurrence of PC attacks in unaffiliated PCs and after bystander affiliations using GLMM with binomial family and logit link functions. The identities of aggressor, victim and aquarium were included as random effects.
Effects of factors on the occurrence of bystander affiliations were investigated using the GLMM with a binomial family (logit link function). The dependent variable was the occurrence of bystander affiliation. The predictor variables were the occurrence of PC affiliation between former opponents before bystander affiliation, the occurrence of bystander affiliation before the other bystander affiliation (i.e., when we investigate bystander affiliation to aggressors, we recorded whether bystander affiliation to the victim occurred before bystander affiliation to the aggressor), counter aggression, the duration of aggression, and whether the aggressor or victim was a calf. The identities of the aggressor and victim, as well as the aquarium, were set as the random effects.
We investigated the effect of affiliative relationships between victims and bystanders on the occurrence of bystander affiliation toward aggressors. The number of bystander affiliations toward aggressors for each victim and potential bystander pair was the dependent variable with a Poisson error structure (log link function) using the GLMM. The affiliative index between the bystander and victim was set as the predictor variable. The opportunity to offer bystander affiliation by each bystander for each aggressor was included as an offset variable. An individual had a chance to offer bystander affiliation to the former opponent when the individual did not fight with the former opponent. For example, in a group composed of individuals A, B, C, and D, individual B had a chance to initiate bystander affiliation toward aggressor A when aggressor A fights with the victim C or D. The number of this chance was used as the opportunity to offer bystander affiliation. As the random effect, we set the identities of the victim, bystander and aquarium. To show the effect of the affiliative relationship between aggressors and bystanders on the occurrence of bystander affiliation toward victims, we ran the GLMM with a Poisson structure (log link function) on the number of bystander affiliations toward victims for each aggressor and potential bystander pair, with the affiliative index between aggressors and bystanders as the predictor variable. The identities of the aggressor and bystander, as well as the aquarium, were set as the random effects.
To explore whether the occurrence of bystander affiliation toward aggressors was affected by affiliative relationship, physical closeness, and whether the bystander was a calf, we ran a GLMM with a Poisson family (log link function) on the number of bystander affiliations toward aggressors for each aggressor and potential bystander pair, with the affiliative index between aggressors and bystanders, the ratio of physical closeness, and whether the bystander was a calf as the predictor variables. The physical closeness ratio was calculated as (the number of times each bystander was closest to former opponents at the end of aggressions)/(the opportunity to offer bystander affiliation). The opportunity to offer bystander affiliation to the aggressor was included as an offset variable. The identities of the aggressor, bystander and aquarium were set as the random effects. Similarly, the number of bystander affiliation toward victims for each victim and bystander pair was set as the dependent variable with a Poisson error structure (log link function) using a GLMM. The affiliative index between victims and bystanders, the ratio of physical closeness with former opponents and whether the bystander was a calf were set as the predictor variables. The opportunity to offer bystander affiliation toward victim was included as the offset variable. The identities of victim, bystander and aquarium were set as the random effects. We ran GLMM in the lme4 package, and compared the full and null models using analysis of variance (ANOVA; car package) to test whether the overall model was improved compared to the null model.
To investigate whether group members was activated to initiate affiliation because they were aroused by aggression, we compared the proportion of attracted and dispersed of PC-MC pair for each former opponent using Wilcoxon matched-pairs signed-ranks exact test (the exactRankTests package). All statistical analyses were conducted in R v. 3.6.0 (R Core Team 2019).
Source: Ecology - nature.com
