In species with long-term social relationships, the ability to recognize individuals after extended separation, and the ability to discriminate between former social affiliates that have died and those that have left the group but may return, are likely to be beneficial. Few studies, however, have investigated whether animals can make these discriminations. We presented acorn woodpeckers (Melanerpes formicivorus), a group-living, cooperatively breeding bird, with playbacks of current group members, former group members still living nearby, former group members that had died or left the study area, and familiar non-group members. Subjects responded more quickly to the calls of non-group members than to the calls of current group members or former group members still living in the study area but did not discriminate between non-group members and former group members that had died or disappeared. This suggests that acorn woodpeckers can vocally recognize both current group members and former group members that have dispersed to nearby groups, and that they either forget former group members that no longer live in the vicinity, or classify them differently from former group members that still live nearby. This study suggests an important role for vocal recognition in maintaining valuable relationships with social affiliates post-dispersal.

MATERIAL AND METHODS

Study site and population monitoring

All data were collected on wild acorn woodpeckers at Hastings Natural History Reservation in central coastal California, USA (36.387ºN, 121.551ºW). This population has been the subject of a long-term study since 1968 (MacRoberts and MacRoberts 1976; Koenig 1981b), and >95% of the individuals are color-banded. Most individuals are banded in the nest at 21 days of age, and unbanded adults immigrating into the population are captured and banded whenever possible. As of 2019, there are approximately 50 social groups within the study area, and each group is censused approximately every 8–10 weeks using spotting scopes to re-sight color banded individuals.

Experiment 1: vocal discrimination of current, former, and non-group members

We conducted Experiment 1 14 Apr–19 Jun 2017 and 2 May–14 Jun 2018. Subjects were 7 females and 7 males from 8 social groups, and all but one were of breeder status. We presented each subject with 3 different playback stimuli on different days: (1) a call of a current group member; (2) a call of a former group member that had died, left the group, or remained on the natal territory after the subject had dispersed 1.1–6.4 years prior to the experiment (median = 2.8 years); and (3) a call of an unrelated individual from a nearby territory that had never lived in the same group as the subject. The distance between the territory centroids of the subject and the non-group member caller was 40–862 m (median=167 m), and in 9 of 14 non-group member trials the caller and subject shared a territorial boundary. As acorn woodpeckers make frequent forays to other territories with a mean foray distance of 500–600 m, subjects were likely familiar with all of the non-group member callers (Barve et al. 2020).

The order of the presentation was balanced according to a Latin square design (Table 1). The 3 stimuli played to a given subject were always recorded from callers of a single sex, such that each subject received either 3 male or 3 female calls. The sex of the subject matched the sex of the callers in 7 of 14 cases. Successive playbacks to the same group or to groups < 250 m from each other were spaced by 6.0 ± 4.8 days on average (minimum 3 days to same group, 2 days to groups closer than 250 m).

Testing the difference in response to callers living in the study area (hereafter “nearby”) vs. callers that had died or left the study area (hereafter “absent”) was not an a priori goal of Experiment 1. However, because of the difficulty of obtaining playback-quality recordings from known individuals, we used the call of an individual that was no longer observed on the study area at the time of the experiment as the former group member stimulus for two female and four male subjects (Table 1). None of the former group members that were classified as “absent” had been observed on the study site for at least a year prior to the experiment and none were seen at least a year post-experiment (as of Sep 2019). For two of the males that received a former group member stimulus from an absent caller, the non-group member stimulus was also from an absent caller (not seen for 7 months prior to the experiment). Among the subjects that received the call of a nearby former group member, the distance between the territory centroids of the subject and the caller was ­­­121–1587 m (median = 228 m), and in 5 of 8 nearby former group member trials, the caller and subject shared a territorial boundary

Experiment 2: vocal discrimination of nearby and absent former group members

We conducted Experiment 2 6 Apr–12 Jul 2019, using 5 female and 6 male subjects from 8 different groups (Table 2). Six individuals were used as subjects in both Experiment 1 and Experiment 2, and among the 11 subjects used in Experiment 2, all but 2 were members of a social group that was exposed to playbacks in Experiment 1. Experiment 2 was designed as a follow-up to Experiment 1 to investigate whether acorn woodpeckers could vocally discriminate between nearby and absent former group members. We presented each subject with the call of a former group member living on a neighboring territory within the study area (nearby), the call of a former group member that had not been observed in the study area for 1.1–7.3 years before the experiment (median absence =3.0 years), and the call of an unrelated individual from a nearby territory that had never lived in the same group as the subject.

The interterritorial distance of subjects and nearby former group members was 102–257 m (median = 132 m), and all nearby former group members shared a territorial boundary with the subject. The interterritorial distance of subjects and non-group members was 102–734 m (median = 257 m), and 5 of 11 non-group members shared a territorial boundary with the subject. Thus, as in Experiment 1, subjects were likely familiar with all nearby former group member and non-group member callers. As of Sep 2019, none of the absent former group members had been observed on the study area since their last sighting 1.1–7.3 years before the experiment.

We ensured that the amount of time since the subject and caller last lived together did not statistically differ between nearby former group member and absent former group member playback stimuli (Paired t-test, t₁₀=-1.2, P=0.25). As in Experiment 1, order of presentation was balanced according to a Latin square design (Table 2), and the 3 playback stimuli presented to a given subject were recorded from 3 callers of a single sex, which matched the sex of the subject in 6 of 11 cases. Successive playbacks to the same group or to groups <250 m apart were separated by 6.6 ± 6.5 days (minimum 2 days).

Playback stimuli

In both experiments, the calls used as playback stimuli were waka calls, an individually distinctive, affiliative call typically produced when members of the same group approach one another after a short period of separation (MacRoberts and MacRoberts 1976; Yao 2008). All playback stimuli were recorded at Hastings Reservation 19 Mar 2015–26 May 2017 using a Sennheiser ME67 or ME62 microphone (Wedemark, Germany) and a Marantz PMD661 (Kanagawa, Japan), Fostex FR-2 (Akishima City, Tokyo, Japan), or Roland R26 (Hamamatsu, Shizuoka, Japan) digital recorder (48 kHz, 16 or 24 bits). Prior to constructing the playback stimuli, the calls were high-pass filtered (200 Hz cut-off, 6 dB roll off) and normalized to -3 dB in Audacity® 2.1.1, and any calls originally recorded at 24 bits were converted to 16 bits.

In Experiment 1, the playback stimuli consisted of 60 sec of background noise with a 10-sec fade-in, followed by a single waka call, followed by 30 sec of background noise, followed by the same waka call, followed by a final 10 sec of background noise with a fade-out applied to all 10 sec. Repeating the call increased the likelihood that the subjects would respond to the playback, and the 30-sec interval between calls followed previously published protocols (Yao 2008; Pardo et al. 2018). While natural waka calls are most commonly produced singly, they are sometimes repeated at an interval close to 30 sec (M. Pardo, unpublished data). The playback stimuli for Experiment 2 were constructed in the same way except that the initial period of background noise only lasted 30 sec. We made this change to reduce the chance that the subject would fly away before the call began.

Playback protocol

Playback trials for both experiments followed a similar protocol to Pardo et al. (2018). In brief, we placed a Yamaha PDX 11 loudspeaker (Hamamatsu, Shizuoka, Japan) characterized at 100.1 ± 1.3 dB re 20 µPa at 1 m in a tree 1–1.5 m off the ground and 40 m away from a tree near the center of the group’s territory (“center tree”). This volume was at the upper end of the range of natural waka calls produced by a captive adult male acorn woodpecker (Pardo et al. 2018). The speaker was always placed in the same location during successive trials to a given group. Once the subject was located in the center tree, an observer began filming the subject using either a Canon PowerShot SX510 digital camera (Ota City, Tokyo, Japan) or a Sony Handycam DCR-SX45 Camcorder (Minato, Tokyo, Japan), and immediately played the appropriate playback file.

Measuring response to playback

Based on video and audio recordings of each playback trial, we measured the following aspects of the focal bird’s response: latency to the first “reaction” (defined as vocalizing, flying up to a higher vantage point, or flying toward the speaker), latency to the first “positive” flight (defined as flying up to a higher vantage point or toward the speaker), latency to the first approach to the speaker, latency to the closest approach to the speaker, distance of the first approach to the speaker, and distance of the closest approach. For the latency variables, if the focal bird did not exhibit the behavior of interest within 3 min after the start of the playback, latency was assigned the maximum possible value of 180 sec and marked as censored. Distances were estimated by eye to the nearest 5 m using landmarks of known distance to the speaker, measured before the experiment with a 50-m tape. The authors involved in scoring the videos were blind to the experimental condition in each trial until all scoring was complete.

Statistical analyses

We conducted statistical analyses in R 3.6.1 (R Core Team 2019). Because of limited sample sizes, we pooled the results of both experiments. For Experiment 1, we rescored each former group member trial as either nearby former group member or absent former group member, according to whether the caller was still living in the study area at the time of the experiment. Thus, in the pooled dataset there were four treatment categories (current group member, nearby former group member, absent former group member, and non-group member), and each subject only received 3 of these 4 treatments within a given experiment. We used 13 unique call exemplars from 13 different callers as playback stimuli for the current group member category (n=14), 17 unique exemplars from 13 different callers for the nearby former group member category (n=19), 12 unique exemplars from 8 different callers for the absent former group member category (n=17), and 19 unique exemplars from 17 different callers for the non-group member category (n=25).

We limited our model set to latency to react, latency to positive flight, and distance of first approach as response variables, as all other responses measured were highly correlated (Pearson’s r>0.75) with at least one of these. We analyzed latency to react and latency to positive flight using Cox regression in the R package coxme (Therneau 2018) to account for the fact that some of the latencies were censored. We analyzed distance of first approach using linear mixed models in the R packages lme4 (Bates et al. 2015) and lmerTest (Kuznetsova et al. 2017). For each response variable, we ran a model with treatment, subject sex, caller sex, and the interaction of subject sex by caller sex as fixed effects, and individual ID as a random effect. As the goal of the experiment was to determine whether subjects could discriminate current and former group members from non-group members, we compared the responses to non-group members with the responses to each remaining treatment category using Dunnett’s method in package emmeans to adjust for multiple comparisons (Lenth 2018). We also conducted pairwise comparisons for the interaction of subject sex*caller sex using Tukey’s method in emmeans.

To determine if time since the subject last lived with a former group member affected the subject’s response to that former group member’s call, we ran an additional model for each response variable using only the present and absent former group member trials. These models included time since together as a fixed effect and individual ID as a random effect.

Definitions of all the variables are included in a separate file