Northeastern Naturalist 580 M. Kerstens, A.M. Grade, and P.S. Warren 22001199 NORTHEASTERN NATURALIST 2V6(o3l). :2568,0 N–5o9. 23 Is “Pishing” Tantamount to Mobbing? Black-capped Chickadees Respond Similarly to Human Pishing and Conspecific Mobbing Calls in Rural and Suburban For ests Mark Kerstens1, Aaron M. Grade2,*, and Paige S. Warren1 Abstract - Poecile atricapillus (Black-capped Chickadee) mobbing calls elicit mobbing events in which birds drive away a predator. Human birders simulate these calls by “pishing” (vocalizing mobbing calls) to coerce birds into view. We investigated whether pishing is as salient as natural mobbing calls, and whether birds respond differently in suburban forest fragments than in intact forests. Using experimental playbacks, we broadcast mobbing calls and pishing calls in suburban forest patches and intact forests and measured Chickadee response. We found that Chickadees had significantly stronger responses to mobbing calls than to alarm calls used as a positive control, but pishing call response was not significantly different than mobbing or control alarm call responses. We also saw no difference in number of Chickadees responding between suburban fragments and intact forests, but we did see a difference in areas with denser vegetation. These findings show that pishing may be less urgent than Chickadee mobbing calls but may still contribute to stress and energetic demands on birds. Introduction Alarm calls and mobbing calls are used by many animal species to convey the presence of predators and other threats. Predators may be an imminent danger due to their proximate behavior, such as an aerial predator in an attack dive, or they may present a background level of predation risk, through their mere presence in an area. Alarm calls typically signal imminent threats by predators, and are characterized by short, high frequency, and difficult to localize calls that are used when immediate evasive actions need to be taken. Mobbing calls, by contrast, are characterized by lengthy and loud calls that are easier to localize, used when a predator threat is less imminent, and function to recruit other nearby individuals to join in mobbing efforts (Magrath et al. 2007, Templeton et al. 2005). Animals often flee for cover or freeze in response to alarm calls (Bedneko and Lima 1998, Leavesley and Magrath 2005, Rajala et al. 2003), whereas mobbing calls typically elicit a mobbing response in which several birds attack and call at a potential predator to confuse or drive it off (Courter and Ritchison 2010, Desrochers et al. 2002, Ficken and Popp 2018). Mobbing calls cause individuals of many species to approach the targeted predator in a group attempt to drive the predator away from the area. 1Department of Environmental Conservation, University of Massachusetts, Amherst, MA 01003. 2Program in Organismic and Evolutionary Biology, University of Massachusetts, Amherst, MA 01003. *Corresponding author - agrade@umass.edu. Manuscript Editor: Susan Smith Pagano Northeastern Naturalist Vol. 26, No. 3 M. Kerstens, A.M. Grade, and P.S. Warren 2019 581 Birders (i.e., hobbyists that birdwatch) often take advantage of birds’ responses to acoustic signals to lure birds into the open for improved observation. They do this by playing conspecific calls from a speaker or by using a method dubbed “pishing”—making a vocal sound that imitates that of mobbing calls in the family Paridae (hereafter, Parids). Some researchers have raised concerns over the impacts of these “active” birding methods (Kronenberg 2014, Steven et al. 2011). Such disturbances are likely to be stressful for birds, and might have negative consequences during breeding seasons or in times of resource scarcity, such as during winter (Gabrielson and Smith 1995). Poecile atricapillus (L.) (Black-Capped Chickadee, hereafter Chickadee), is a resident Parid that forages in mixed species flocks outside of the breeding season (Sieving et al. 2010). Chickadees are considered a nuclear species of mixed species flocks in North America, and they are often the first species to initiate mobbing in response to a predator threat (Langham et al. 2006). Interestingly, human pishing has a similar acoustic structure to the highly conserved variant of the “chick-a-dee” mobbing call that is vocalized by numerous Parid species, indicating that birds responding to pishing may be drawn to the call because of its structural resemblance to Parid mobbing calls (Langham et al. 2006). The vocal communication system of Parids is highly complex and multifaceted. When “chick-a-dee” calls are used to broadcast an assessment of the predator risk environment, they can convey both the degree of (graded) and category of (functionally referential) risk that the signaler has detected (Sieving et al. 2010). In the context of predator threats, the structure of this call can be used as a proxy for the “degree of risk” that an individual Chickadee perceives in the environment (Baker and Becker 2018, Templeton and Greene 2007, Templeton et al. 2005). When used to assess predation risk, the ratio between the “d” notes and the “chick” notes in the “chick-a-dee” call (i.e., d/chick) is a particularly informative indicator of risk level. A higher ratio indicates that the individual is communicating that the environment is “riskier” (i.e., more d notes, less chick notes), whereas a smaller ratio indicates a “less risky” environment (Templeton et al. 2005). Once the risk level passes the threshold from a mobbing action to an evasive action from an imminent predator, Chickadees switch to other note types, such as the high-pitched and rapid “hi-seet” notes that are utilized as an urgent alarm call (Grade and Sieving 2016, Sieving et al. 2010). If Chickadees perceive human pishing as Chickadee mobbing calls, they should respond to both sounds with equivalent d-to-chick ratios. The surrounding human development context may also play a role in birds’ responses to mobbing calls. Urbanization has a wide range of effects on animals, including a heightened perception of predation risk due to higher predator densities (McKinney 2002). In addition, the urban environment may present a riskier landscape, such as more edge habitat and areas for aerial predators to ambush prey (Bowers and Breland. 2016, Sieving et al. 2004). Due to the heightened perception of both background- and proximate-level risk, we can expect weaker mobbing responses in suburban forest fragments. Birds will typically join in mobbing predators only if their risk of predation is relatively low, such as during the day, when Northeastern Naturalist 582 M. Kerstens, A.M. Grade, and P.S. Warren 2019 Vol. 26, No. 3 an aerial predator is perched and not actively diving, and in an area with a lot of potential cover (Desrochers et al. 2002). In a habitat patch with higher predator densities (i.e., a “riskier” location for background-level predation risk), birds may be less likely to mob, or the mobbing strength (e.g., intensity, distance of approach to predator) may be tempered (Forsman and Mo 2001). Alternatively, the increased disturbances of human activity brought on by urbanization (e.g., dog walking, vehicle traffic) might lead to habituation to potential threats and result in a reduction in aggressive mobbing behavior. Several studies have found changes in foraging, predator avoidance, and nesting behaviors with urbanization or heightened human activity, but to our knowledge, the impacts of urbanization on mobbing behavior have not been explored (Grade and Sieving 2016, Sol et al. 2013). To date, only one study of which we are aware has examined responses of birds to pishing (Langham et al. 2006). Langham et al. (2006) found that mobbing responses in southern California by resident Parids (Baeolophus inornatus [Gambel] [Oak Titmouse]) were equally strong to Parid mobbing calls and pishing playbacks, intermediate to Parid mobbing calls from other regions, and weakest to non-Parid alarm calls. The authors argued that these findings provided support for the popular conjecture that pishing mimics the mobbing calls of Parids, and therefore elicits mobbing responses. Further tests of responses to pishing from other Parid species in other regions and comparisons to other non-mobbing alarm calls would provide greater support for this conjecture. The focus of this study was to determine whether pishing elicits the same response strength as Parid mobbing calls, and whether birds respond differently to these mobbing calls in suburban versus intact forest areas. We performed a playback study during the post-breeding (fall and winter) season in forests of western Massachusetts at sites that varied in the degree of suburban development in the surrounding landscape. Answering these questions provides insight into the salience of pishing versus natural calls in suburban settings. We tested the following predictions: We predicted that the Chickadee mobbing call and pishing playbacks would both garner a stronger response (i.e., more individuals approaching the playback) than playbacks of alarm calls made by Cardinalis cardinalis (L.) (Northern Cardinal, hereafter, Cardinal). We used Cardinal alarm calls as a control because pishing is supposed to mimic the Chickadee mobbing call, which is a stronger driver of Chickadee behavior and communicated risk than Cardinal alarm calls (Langham et al. 2006). Given that habituation to threats may outweigh the elevated predator densities in suburban environments, we also expected responses to all playbacks to be stronger in suburban patches than in intact forests. Birds often respond more vigorously to mobbing elicitation in “safer” habitats than more “risky” habitats, and the behavior of animals often indicates that suburban habitats are “less risky” than intact forested habitats (Bowers and Breland 2016, Lerman et al. 2012,). We predicted that Chickadees would communicate higher risk (i.e., higher d/chick ratio) when exposed to Chickadee mobbing calls and pishing playbacks than to Cardinal alarm calls. We also expected higher risk ratios in intact forests than suburban patches. Northeastern Naturalist Vol. 26, No. 3 M. Kerstens, A.M. Grade, and P.S. Warren 2019 583 Methods Field-site description We conducted a playback experiment in temperate forests and suburban forest patches in western Massachusetts, between September 2015 and January 2016. The forests were characterized by mixed deciduous and coniferous vegetation, hills, and mountains. We identified 2 kinds of sites: suburban forest patches and intact forest wildlands. The sites within the binary grouping did not differ significantly enough in landscape context to justify classification beyond the suburban and intact forest categorizations. We selected suburban forest patch sites from sites that were already being utilized for an ongoing avian ecology study (P.S. Warren, unpubl. data). These sites averaged 25 ha in size, had trails frequented by people, and were surrounded by a matrix of roads and residential suburban neighborhoods. Intact forest wildlands tended to be larger than 25 ha, were surrounded by >80% forest cover with no residential areas, few roads, and considerably less human activity than suburban sites (i.e., decreased trail density and visitation). Some intact forest sites had evidence of past logging disturbance. Study design We conducted this research under the University of Massachusetts IACUC permit #2014-0043. At each site, we broadcast playbacks of (1) Chickadee mobbing calls, (2) pishing exemplar recordings, and as a positive control, (3) alarm calls by Cardinals—a resident non-Parid species that typically responds to a different suite of predators than Chickadees. If Chickadees perceive pishing calls as equivalent to mobbing calls, they should respond similarly to both of these stimuli. Cardinal alarm calls, by contrast, are known to result in lower response rates by Chickadees than the chick-a-dee mobbing call variant (Langham et al. 2006). We obtained 3 exemplars of both Chickadee mobbing and Cardinal control calls from the Xeno- Canto database (www.xeno-canto.org/). We obtained pishing exemplar recordings from 3 experienced local birders by recording them pishing in a soundproof recording chamber. We edited the rate and amount of calls within each playback for consistency among playback exemplars. We established playback locations that were a minimum of 150 m from the habitat edge and from other playback locations, and most playbacks were significantly farther than 150 m apart. We selected distances based on the expectation that calls would attenuate enough to minimize detection between playback locations, thereby avoiding re-sampling the same birds, and on the basis of previous studies in a forested setting (Bibby et al. 1992). Each location was first selected on Google Maps (www.google.com/maps) and then located with a Garmin GPS map 62S GPS unit (Garmin Ltd., Olathe, KS). We performed 3 trials of data collection at different sites within each playback location between 0800 and 1400 h in fair weather (i.e., no rain or gusty wind). For each location, we conducted trial 1 from 25 September to 21 October, trial 2 from 22 October to 5 December 2015, and trial 3 from 1 December 2015 to 11 January 2016. We randomly selected which playback types and which exemplar would be used at which locations during a given trial, and tested birds at each location with 1 Northeastern Naturalist 584 M. Kerstens, A.M. Grade, and P.S. Warren 2019 Vol. 26, No. 3 recording from each of the 3 playback types. We played playbacks through a Rock Out portable speaker (model #90401; Goal Zero, Bluffdale, UT) wired with a 3-mm auxiliary cord to a Samsung S4 Android cellular phone (Samsung Group, Seoul, South Korea). We placed the speaker on a branch about 2 m (approximate height of a birder or perching height of a Cardinal or Chickadee) off the ground at each location. We standardized the amplitude of all playbacks to 65 dBA at 5 m from the speaker using a dB-meter (A-weighted, SPL meter model 33-2055; Radio Shack, Fort Worth, TX). We chose this amplitude to maximize the reach of the playback while maintaining signal fidelity (i.e., minimizing “clipping”). Playbacks had an initial 30 s of silence, which allowed the researcher to place the speaker and retreat to 10 m away, followed by 2 min of silence for pre-playback data collection, 2 min of playback, and 4 min of silence for post-playback data collection. In the pre-playback and post-playback data collection phases, we recorded the number of Chickadees seen or heard within a 50-m radius of the speaker (Bibby et al. 1992). We also recorded the total number of “d” notes and “chick” notes produced by all Chickadees in the area before and after the playbacks. To accurately obtain “d” and “chick” note rates in the field, the observer held a clicker counter in each hand and used the clickers to count each “d” note and “chick” note heard, respectively. The density of Chickadees was low enough during each trial (when n > 0, mean = 2.27) to make this possible in situ. We converted “d” note and “chick” note rates into a “d” to “chick” note ratio (d/chick), a metric of Chickadee mobbing call level of communicated risk (higher ratio is more risky, lower is less risky; Templeton et al. 2005). We recorded additional site and trial-specific conditions, including temperature (continuous), weather (categorical: e.g., sunny or cloudy), wind (binary: windy or not windy), and a general classification of sitelevel vegetation (binary: dense or sparse, based on visual estimation). Statistical analyses We conducted all analyses in the statistical program R version 1.0.136 (R Core Team 2015) in the R Studio application (RStudio, Inc., Boston, MA). We used Akaike’s information criterion for finite sample sizes (AICC) to select the best-fit models for our statistical hypotheses (Akaike 1973). We included the different playback types, as well as the different exemplars of each playback type as categorical variables in the models. Models with ΔAICC ≤ 2 than the lowest AICC model were considered of equal fit (Burnham and Anderson 2003). Of the models with ΔAICC ≤ 2, we selected the most parsimonious model that also contained the highest number of statistically significant predictor variables (P ≤ 0.05) as our most useful model for describing responses to playbacks. In addition to the 2 primary predictors of interest—playback type (Chickadee, pishing, or Cardinal) and site type (suburban patch or intact forest)—we included the following as other covariates in the set of possible predictors: temperature, wind, and weather, and site-level vegetation. Based on exploratory analyses that indicated that our count data were zero-inflated, we used zero-inflated generalized linear mixed models with negative binomial probability distribution functions (ZINB) and a logistic link function for all models that had the number of Northeastern Naturalist Vol. 26, No. 3 M. Kerstens, A.M. Grade, and P.S. Warren 2019 585 Chickadees seen or heard post-playback as the primary response variable (Zuur et al. 2009). For the models that had “d/chick note ratio after playback” as the primary response variable, we used general linear mixed models (GLMM; Zuur et al. 2009). We also included in all models the starting number of Chickadees present per minute as a random variable to account for variability in starting Chickadee density. We tested 30 ecologically plausible combinations of the above predictor variables as fixed effects. We used the “pscl” package in R (Jackman 2015) to run the ZINB models, the “nlme” package in R (Pinheiro et al. 2017) for the GLMMs, and the “MuMIn” package in R (Barton 2017) to obtain the AICC values. We used the “lsmeans” package in R (Lenth 2016) to obtain post-hoc Tukey LSD statistics for pairwise comparisons on all categorical variables with 3 or more categories. We plotted our results using the “dplyr” package in R (Wickham and Francois 2016) and the “ggplot2” package in R (Wickham 2009). We set statistical significance at a P ≤ 0.05. After selecting the appropriate models, we conducted post-hoc bootstrapping using the “boot” package in R (Canty and Ripley 2015) on the ZINB model to generate confidence intervals to assess if the responses to the 3 levels of playback type were significantly different from each other. With the bootstrapping procedure, we modelled both the count component and the logistic component of the ZINB model and generated exponentiated parameter estimates with percentile confidence intervals (n = 1200, P = 0.05). Results Chickadee approaches as a response to playbacks We completed 135 trials total, in multiple locations spread across 13 sites (n = 7 sites in intact forest, n = 6 sites in suburban patches). The full study dataset is available in the supporting information (S1 Dataset; see Supplemental File S1, available online at http://www.eaglehill.us/NENAonline/suppl-files/n26-3-N1731-Grade-s1, and for BioOne subscribers, at https://dx.doi.org/10.1656/N1731.s1). Out of the 30 models, the model that included playback type (playback typepishing: β = -0.20, SE = 0.38, z124 = -0.53, P = 0.59; playback typeCardinal: β = -0.83, SE = 0.38, z124 = -2.15, P = 0.04) and habitat vegetation (habitat vegetationdense: β = 0.80, SE = 0.38, z124 = -2.10, P = 0.04) was the most parsimonious model of the 3 models with ΔAICC < 2 (Table 1). To assess post-hoc whether the 3 playback types were significantly different from one another in Chickadee approaches, we compared the generated confidence intervals at the P = 0.05 level (n = 1200 iterations). The Table 1. A summary of best-fitting (ΔAICC < 2) models for Chickadee approaches, including pishing trials. Playback type = pishing, Cardinal alarm, or Chickadee mobbing, and habitat vegetation = dense versus sparse vegetation. An asterisk (*) denotes the most parsimonious model with all statistically significant predictor variables. Model AICC ΔAICC wi df LogLik 1 + playback type + habitat vegetation* 384.0 0.00 0.183 7 -184.524 1 + playback type + temperature + habitat vegetation 384.6 0.66 0.132 8 -183.720 1 + playback type 385.7 1.71 0.078 6 -186.498 Northeastern Naturalist 586 M. Kerstens, A.M. Grade, and P.S. Warren 2019 Vol. 26, No. 3 Tukey test indicated that there was no meaningful difference between pishing and Chickadee mobbing (t = 1.272, P = 0.422). There was also no difference between pishing and Cardinal alarm (t = 1.008, P = 0.568). There was a significant difference between Chickadee mobbing and Cardinal alarm in approach of Chickadee individuals (t = 2.533, P = 0.043; Fig 1). There was a significant effect of habitat type on Chickadee response regardless of playback type, with significantly higher densities of Chickadees in denser interior habitat than in sparser open habitat (habitat vegetationdense: β = 0.80, SE = 0.38, z124 = -2.10, P = 0.04). Differences in Chickadee response to pishing between pishing exemplars To test whether Chickadees were responding differently to the different pishing exemplars (i.e., if the high variance in response to pishing was due to different Figure 1. Playback (i.e., recording) types versus number of chickadees present (i.e., response of Chickadees) following playbacks. Error bars indicate SE, different letters indicate significant differences between pairwise comparisons via bootstrapped confidence intervals. BCCH represents Black-capped Chickadees and NOCA represents Northern Cardinals. Northeastern Naturalist Vol. 26, No. 3 M. Kerstens, A.M. Grade, and P.S. Warren 2019 587 pishing exemplars), we ran a model identical to those we used above. Pishing exemplar recording was the only fixed variable. The effect of exemplar on Chickadee response was statistically significant (exemplarJP: β = 1.98, SE = 0.64, z80 = 3.11, P = 0.002; exemplarPP: β = 0.47, SE = 0.51, z82 = 0.90, P = 0.37). We were unable to run the bootstrapping protocol for post-hoc comparisons on this subset of trials due to the small sample size. We instead ran pairwise models of subsets of each of the 3 recordings against each other. One of the recordings (JP) was significantly different from the other two (For JP vs. AP, exemplarJP: β = 1.91, SE = 0.81, z25 = 2.35, P = 0.02; for JP vs. PP, exemplarPP: β = -1.50, SE = 0.63, z25 = -2.39, P = 0.02), but the other 2 were not significantly different from each other (for AP vs. PP, exemplarPP: β = 0.50, SE = 0.42, z21 = 1.19, P = 0.23). This result indicates that variance in response among pishing recordings contributed to the wide variance in pishing response. Chickadee communicated perception of risk to playbacks: d/chick note ratios We used a subset of trials in which Chickadees vocally responded to playbacks (n = 36, including trials with pishing) to assess the effects of playback type and site type on Chickadee d/chick note ratios (communicated perception of risk). We included the same 30 models from the other model-selection protocols but utilized a linear mixed-effects model (LME) rather than a zero-inflated model. This selection protocol resulted in the model that included only playback type as the most parsimonious model out of the 2 models with ΔAICC < 2 (Table 2). Playback type was a statistically significant predictor of d/chick note ratios, with higher ratios (indicating more risk) in response to Chickadee mobbing calls than to Cardinal alarm calls (playback typePishing: β = -0.70, SE = 0.55, t30 = -1.27, P = 0.21, playback typeCardinal: β = -1.28, SE = 0.51, t30 = -2.53, P = 0.02; Fig 2). A post-hoc Tukey HSD test indicated that Chickadee and Cardinal calls had significantly different responses, but pishing was not significantly different from either Chickadee mobbing or Cardinal alarm (Chickadee vs. pishing: β = -0.70, SE = 0.55, t30 = 1.27, P = 0.42, Chickadee vs. Cardinal: β = 1.28, SE = 0.51, t30 = 2.53, P = 0.04, pishing vs. Cardinal: β = 0.58, SE = 0.58, t30 = 1.01, P = 0.58). Site type was not statistically significant when tested as a fixed effect (β = -0.21, SE = 0.47, t30 = -0.45, P = 0.65), therefore, it was not included in the model. Discussion Few studies have examined the interaction between urbanization and “active” birding activities, such as the use of pishing, on avian mobbing behavior. Contrary to our predictions, we found that urban context (i.e., suburban forest Table 2. A Summary of best-fitting (ΔAICC < 2) Models for Chickadee d/chick note ratios. An asterisk (*) denotes the most parsimonious model with all statistically si gnificant predictor variables. Model AICC ΔAICC wi df LogLik 1 + playback type* 130.7 0.00 0.286 5 -59.372 1 + playback type + habitat vegetation type 131.1 0.38 0.237 6 -58.113 Northeastern Naturalist 588 M. Kerstens, A.M. Grade, and P.S. Warren 2019 Vol. 26, No. 3 patch vs. intact forest) was not significantly associated with Chickadee approaches or risk communicated in mobbing calls. Also contrary to our predictions, pishing recordings were not significantly different from Cardinal alarm or Chickadee mobbing calls in either response variable. These findings indicate that, although pishing did elicit Chickadee response, the urgency of the call was somewhere between the response to Chickadee mobbing and Cardinal alarm calls, and therefore, was not as salient as Chickadee mobbing in this particular study versus other studies (Langham et al. 2006). The lack of a statistical difference between pishing and the other 2 call types may be due to the wide variance in response to different pishing exemplars recorded by different people. This finding alone is interesting, and shows that individual birder Figure 2. Playback (i.e., recording) types versus d/chick note ratio in Chickadees. Error bars indicate SE, different letters indicate significant differences between pairwise comparisons via post-hoc Tukey HSD test. BCCH represents Black-capped Chickadees and NOCA represents Northern Cardinals. Northeastern Naturalist Vol. 26, No. 3 M. Kerstens, A.M. Grade, and P.S. Warren 2019 589 skill or vocal differences can lead to variable response strength to pishing, reducing its overall effectiveness in eliciting a response comparable to that of Chickadee mobbing. Site vegetation composition was included in the best model for Chickadee approaches. This finding is consistent with previous research that has shown the importance of vegetation structure in birds’ perception of risk, and therefore, their willingness to engage in riskier and stronger mobbing activity (Sieving et al. 2004). More Chickadees responded in areas with denser than sparser vegetation, regardless of whether the site was in a suburban or intact forest patch. More densely foliated sites present safer habitats in which to mob a perched predator than a more open edge habitat (Sieving et al. 2004). This current study may be the first to examine the impact of urbanization on mobbing behavior in passerines, and thus raises new questions. Our results also added complexity to the findings from the only previous study on pishing of which we are aware, as we found that bird responses to pishing were mixed when compared to real Parid mobbing calls (Langham et al. 2006). We did find that, as expected, significantly more birds responded to Chickadee mobbing calls than to the control Cardinal alarm calls. This result leads us to suggest that the effectiveness of pishing can be influenced by the skill and vocal attributes of the person giving the pishing call, but that pishing in general elicits changes in Chickadee behavior and perception of risk similar to that of mobbing or alarm calls. If pishing does indeed effectively mimic Parid mobbing calls, there may be behavioral effects on a wide range of species in areas where the practice is common, such as birding hotspots. Although Chickadee mobbing calls are thought to be directed towards conspecifics within the flock, other species form mixed-species flocks that interact with and follow Chickadees as well as other species in the family Paridae. These flocks are most cohesive during the winter, and individuals within the flock eavesdrop on relevant Parid calls to acquire socially derived information of predation risk. This behavior allows each bird to allocate more time for foraging and less time for individual vigilance during times of decreased food availability (Dolby and Grubb 1998, Sullivan 1984, Templeton and Greene 2007). In these mixed-species flocks, mobbing calls of Parids, often elicit mobbing behavior (Magrath et al. 2007). For example, Templeton and Greene (2007) found that Sitta carolinensis Latham (White-breasted Nuthatch) responded 92% of the time to “chick-a-dee” mobbing calls that indicate the presence of a shared predator threat. If pishing is widespread in a given area, it may cause behavioral changes and wasted energy expenditure in entire flocks and across a wide-range of species (Kronenberg 2014, Sullivan 1984). Although pishing at birding hotspots is unlikely to have widespread ecological consequences, it elevates the ethical implications of active birding techniques, even if birders are only using their voices rather than amplified recordings of bird calls. Implications of urbanization on mobbing behavior Interestingly, we found no difference in response to playbacks between intact forests and suburban forest patches. Despite this finding, birds in densely foliated Northeastern Naturalist 590 M. Kerstens, A.M. Grade, and P.S. Warren 2019 Vol. 26, No. 3 habitat responded with greater strength than birds in edge habitat. In the intact forest fragments, the edge habitats were places where the forest patch met a meadow, pond, hiking trail, or dirt logging road, whereas edge habitat in suburban forest fragments were places where the patch met a meadow, pond, hiking trail, residential backyard, or paved road. It seems that in this experiment, within-patch habitat characteristics were more important to the strength of Chickadee responses than landscape context. We caution that, in addition to edge characteristics, many other factors differ between suburban forest patches and intact forest sites. For example, birds are often found in higher densities in suburban forest fragments (Chace and Walsh 2006). It is feasible that there may simply have been more birds within range in these patches to respond to our playbacks. To address this possibility, we included initial densities of Chickadees as a random effect in the models. In addition, there was a large range of temperatures and wind levels across the fall and winter months in the study region, with winter months tending to be colder and windier. Weather factors such as cloud cover, precipitation, temperature, and wind are known to affect bird activity levels, and these varied conditions could have tempered some of the effects of the playbacks at a level that was not selected for in the models (Shedd 2018). Finally, rather than following and targeting mixed-species flocks, we chose random locations to begin the playbacks. This protocol resulted in overall low approaches by Chickadees, and zero-inflated data. Future studies should consider a more active approach of finding mixed-species flocks to increase the power to detect effects and reduce the number of trials with no approaches. Conclusions Birding can offer positive benefits, such as providing people with an opportunity to spend time outside or to increase appreciation of and conservation efforts for wildlife, but it can also disturb natural areas (Sekercioglu 2002, Straka and Turner 2013). Our results further suggest that birding activities can have impacts on birds, and that behaviors such as pishing may modify avian behavior and increase birds’ perception of predation risk. In times of stress (such as winter months in the northeastern US), birding could potentially present a fitness costs to birds. People who are keen on bird-watching may want to follow ethical guides (e.g., Sekercioglu 2002) for viewing wildlife to minimize disturbing the natural areas and species that they have come to enjoy. Acknowledgments We thank J. Finn, B. Kane, J. Podos, J. Smith, and K. Straley for their assistance in the design and analysis of this research. Literature Cited Akaike, H. 1973. Maximum likelihood identification of Gaussian autoregressive moving average models. Biometrika 60:255–265. Baker, M.C., and A.M. Becker. 2018. Mobbing calls of Black-Capped Chickadees: Effects of urgency on call production. Wilson Ornithological Bulletin 114:510–516. Northeastern Naturalist Vol. 26, No. 3 M. Kerstens, A.M. Grade, and P.S. Warren 2019 591 Barton, K. 2017. MuMIn: Multi-Model Inference. Available online at https://cran.r-project. org/package=MuMIn. Accessed 4 April 2019. Bedneko, P.A., and S.L. Lima. 1998. Re-examining safety in numbers: Interactions between risk dilution and collective detection depend upon predator targeting behaviour. Proceedings of the Royal Society of London B 265:2021–2026. Bibby, C.J., N.D. Burgess, D.A. Hill, and S. Mustoe. 1992. Bird Census Techniques, 2nd Edition. Academic Press, London, UK. 302 pp. Bowers, M.A., and B. Breland. 2016. Foraging of Gray Squirrels on an urban–rural gradient: Use of the GUD to assess anthropogenic impact. Ecolological Applications 6:1135–1142. Burnham, K.P., and D.R. Anderson. 2003. Model Selection and Multimodal Inference: A Practical Information-theoretic Approach. Springer, New York, NY. 488 pp. Canty, A., and B. Ripley. 2015. boot: Bootstrap R (S-Plus) functions. Available online at https://cran.r-project.org/web/packages/boot/boot.pdf. Accessed 4 April 2019. Chace, J.F., and J.J. Walsh. 2006. Urban effects on native avifauna: A review. Landscape and Urban Planning 74:46–69. Courter, J.R., and G. Ritchison. 2010. Alarm calls of Tufted Titmice convey information about predator size and threat. Behavioral Ecology 21:936–942. Desrochers, A., M. Belisle, and J. Bourque. 2002. Do mobbing calls affect the perception of predation risk by forest birds? Animal Behaviour 64:709–714. Dolby, A.S., and T.C. Grubb. 1998. Benefits to satellite members in mixed-species foraging groups: An experimental analysis. Animal Behaviour 56:501–509. Ficken, M.S., and J. Popp. 2018. A comparative analysis of passerine mobbing calls. The Auk 113:370–380. Forsman, J.T., and M. Mo. 2001. Responses by breeding birds to heterospecific song and mobbing call playbacks under varying predation risk. Animal Behaviour 62:1067–1073. Gabrielson, G.W., and E.N. Smith. 1995. Physiological responses of wildlife to disturbance. Pp. 95–107, In R.L. Knight and K.J. Gutzwiller (Eds.). Wildlife and Recreationists: Coexistence Through Management and Research. Island Press, Washington, DC. 389 pp. Grade, A.M., and K.E. Sieving. 2016. When the birds go unheard: Highway noise disrupts information transfer between bird species. Biology Letters 12:20160113. Jackman, S. 2015. pscl: Classes and methods for R developed in the Political Science Computational Laboratory, Stanford University. Department of Political Science, Stanford University, Stanford, CA. Available online at http://pscl.stanford.edu/. Accessed 4 April 2019. Kronenberg, J. 2014. Environmental impacts of the use of ecosystem services: Case study of birdwatching. Environmental Management 54:617–630. Langham, G.M., T.A. Contreras, and K.E. Sieving. 2006. Why pishing works: Titmouse (Paridae ) scolds elicit a generalized response in bird communities. Ecoscience 13:485–496. Leavesley, A.J., and R.D. Magrath. 2005. Communicating about danger: Urgency alarm calling in a bird. Animal Behaviour 70:365–373. Lenth, R.V. 2016. Least-squares means: The R package lsmeans. Available online at https:// cran.r-project.org/web/packages/lsmeans/lsmeans.pdf. Accessed 4 April 2019. Lerman, S.B., P.S. Warren, H. Gan, and E. Shochat. 2012. Linking foraging decisions to residential yard bird composition. PLoS ONE 7:1–8. Magrath, R.D., B.J. Pitcher, and J.L. Gardner. 2007. A mutual understanding? Interspecific responses by birds to each other’s aerial alarm calls. Behavioral Ecology 18:944–951. Northeastern Naturalist 592 M. Kerstens, A.M. Grade, and P.S. Warren 2019 Vol. 26, No. 3 McKinney, M.L. 2002. Urbanization, biodiversity, and conservation. BioScience 52(10)883–890. Pinheiro, J., D. Bates, S. DebRoy, and D. Sarkar. 2017. nlme: Linear and nonlinear mixed effects models. Available online at https://cran.r-project.org/package=nlme. Accessed 4 April 2019. R Core Team. 2015. R: A language and environment for statistical computing. R foundation for Statistical Computing. Vienna, Austria. Available online at https://www.r-project. org/. Accessed 4 April 2019. Rajala, M., O. Ratti, and J. Suhonen. 2003. Age differences in the response of Willow Tits (Parus montanus) to conspecific alarm calls. Ethology 109:501–509. Sekercioglu, C.H. 2002. Impacts of birdwatching on human and avian communities. Environmental Conservation 29:282–289. Shedd, D.H. 2018. Seasonal variation in mobbing intensity in the Black-Capped Chickadee. The Wilson Bulletin 95:343–348. Sieving, K.E., T.A. Contreras, and K.L. Maute. 2004. Heterospecific facilitation of forest-boundary crossing by mobbing understory birds in north-central Florida. Auk 121:738–751. Sieving, K.E., S.A. Hetrick, and M.L. Avery. 2010. The versatility of graded acoustic measures in classification of predation threats by the Tufted Titmouse, Baeolophus bicolor: Exploring a mixed framework for threat communication. Oikos 119:264–276. Sol, D., O. Lapiedra, and C. González-Lagos. 2013. Behavioural adjustments for a life in the city. Animal Behaviour 85:1101–1112. Steven, R., C. Pickering, and J.G. Castley. 2011. A review of the impacts of nature-based recreation on birds. Journal of Environmental Management 92:2287–2294. Straka, J.R., and M.E. Turner. 2013. Sympathy for the twitcher: What those who seek rare birds do for ornithology. Birding 40–46. Sullivan, K.A. 1984. Information exploitation by Downy Woodpeckers in mixed-species flocks. Behaviour 91:294–311. Templeton, C.N., and E. Greene. 2007. Nuthatches eavesdrop on variations in heterospecific Chickadee mobbing alarm calls. PNAS 104:5479–5482. Templeton, C.N., E. Greene, and K. Davis. 2005. Allometry of alarm calls: Black-Capped Chickadees encode information about predator size. Science 308:1934–1938. Wickham, H. 2009. ggplot2: Elegant graphics for data analysis. Springer-Verlag, New York, NY. 213 pp. Wickham, H., and R. Francois. 2016. dplyr: A grammar of data manipulation. Available online at https://cran.r-project.org/package=dplyr. Accessed 4 April 2019. Zuur, R.A.F., E.N. Leno, N.J. Walker, A.S. Saveliev, and J.M. Smith. 2009. Mixed Effects Models and Extentions in Ecology with R. Springer , New York, NY. 574 pp.