863 The Condor 115(4):863–873 © The Cooper Ornithological Society 2013 E-mail: nlichti@purdue.edu Manuscript received 5 December 2012; accepted 13 May 2013. ACORN-FORAGING PREFERENCES OF FOUR SPECIES OF FREE-RANGING AVIAN SEED PREDATORS IN EASTERN DECIDUOUS FORESTS K IMBERLY B. R ICHARDSON , N ATHANAEL I. L ICHTI 1 , AND R OBERT K. S WIHART Department of Forestry and Natural Resources, Purdue University, 715 W. State Street, West Lafayette, IN 47907 Abstract Oaks ( Quercus spp.) rely on vertebrates for seed dispersal. However, the extent to which dispersal is facilitated depends on the vertebrate species involved, its preferences for acorns of different species, and the rela- tive availability of acorns. We examined foraging preferences by free-ranging individuals of four species of birds for acorns of white ( Q. alba ), black ( Q. velutina ), and pin ( Q. palustris ) oak. We photographed selection of acorns at feeding platforms and used discrete-choice models to estimate preferences. Blue Jays ( Cyanocitta cristata ) had strong and consistent preferences in the order pin oak > black oak > white oak, whereas Red-bellied Woodpeckers ( Melanerpes carolinus ) exhibited the same but weaker preferences. On the basis of acorn’s chemical and physical attributes, these birds appear to focus on maximizing the net value of cached food. Tufted Titmice ( Baeolophus bicolor ) and White-breasted Nuthatches ( Sitta carolinensis ) preferred acorns of both pin and white oak over those of black oak. These smaller birds may be limited in their ability to crack harder shells. Our analysis also suggests that competition for limited food resources can lead to instances in which the dominant food item selected is not the most preferred. Ecological studies of oak dispersal and recruitment should consider the role of food preferences by avian assemblages in conjunction with spatial and temporal variation in acorn availability. Key words: discrete-choice model, jay, nuthatch, oak, seed size, titmouse, woodpecker Preferencias de Forrajeo de Bellotas de Cuatro Especies de Aves en Libertad Depredadores de Semillas en los Bosques Deciduos del Este Resumen Los robles ( Quercus spp.) dependen de los vertebrados para la dispersión de sus semillas. Sin embargo, el grado de facilitación de la dispersión depende de la especie de vertebrado involucrada, de sus preferencias por las bellotas de diferentes especies y de la disponibilidad relativa de bellotas. Examinamos las preferencia de forrajeo de individuos libres de cuatro especies de aves por las bellotas de Q. alba , Q. velutina y Q. palustris . Fotografiamos la selección de bellotas en plataformas de alimentación y empleamos modelos de selec- ción discreta para estimar las preferencias. Cyanocitta cristata presentó una preferencia marcada y consistente por Q. palustris > Q. velutina > Q. alba , mientras que Melanerpes carolinus exhibió una preferencia similar pero más débil. Sobre la base de los atributos químicos y físicos de las bellotas, estas aves parecen enfocarse en maximizar el valor neto del alimento. Baeolophus bicolor y Sitta carolinensis prefirieron bellotas de Q. palustris y Q. alba por sobre las de Q. velutina . Estas aves más pequeñas pueden estar limitadas por su habilidad en romper cáscaras más duras. Nuestro análisis también sugiere que la competencia por los recursos alimenticios limitados puede llevar a instancias en las cuales el ítem alimenticio dominante seleccionado no sea el más preferido. Los estudios ecológi- cos de dispersión y reclutamiento de robles deberían considerar el rol de las preferencias de los ensambles de aves en conjunto con la variación temporal y espacial en la disponibilidad de bellotas. The Condor, Vol. 115, Number 4, pages 863–873. ISSN 0010-5422, electronic ISSN 1938-5129. © 2013 by The Cooper Ornithological Society. All rights reserved. Please direct all requests for permission to photocopy or reproduce article content through the University of California Press’s Rights and Permissions website, http://www.ucpressjournals.com/ reprintInfo.asp. DOI: 10.1525/cond.2013.120189 INTRODUCTION Most vertebrates consume multiple types of prey, and their for- aging preferences can determine what habitats they use and the distribution and dynamics of their prey (McCann et al. 2005). For example, foraging choices made by granivores can affect the chances of a plant’s reproductive success (Janzen 1970, Howe and Smallwood 1982, Crawley and Long 1995, Hume 1996). In oak ( Quercus spp.) trees, vertebrate predators of acorns destroy many seeds and can limit or preclude establishment of seedlings (Crawley and Long 1995). Yet oaks also depend on granivorous vertebrates for seed dispersal (Bossema 1979, Steele and Smallwood 2002). Birds such as jays (Corvidae), and tit- mice (Paridae) cache acorns and may reduce density-dependent mortality for those seeds that are moved away from the parent tree (Janzen 1971, Steele and Smallwood 2002), protect seeds from invertebrates and other predators by burying them (Barnett 1977), and sometimes transport seeds to locations where seed- lings have a higher probability of establishment if caches are not retrieved (Johnson et al. 1997, Steele and Smallwood 2002). However, only animals that cache intact seeds in places where seeds may be viable can act as dispersers (Vander Wall 1990). Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 864 KIMBERLY B. RICHARDSON ET AL oak (Moore and Swihart 2006), so we expected all species to preferentially select the smaller-seeded pin oak, followed by black oak, and finally by white oak. Our study design enabled us to simultaneously assess acorn availability and selection by individual birds, and therefore to evaluate a true prefer- ence, rather than differences in use. In addition, it allowed us to evaluate competition by quantifying the number and timing of visits by different species. METHODS We studied foraging preferences at six sites in Tippecanoe and Warren counties, west-central Indiana, in 2009. Sites were sep- arated by 2.5–29.5 km. All of the sites were located in a region of small fragments of deciduous forest surrounded by large areas of agriculture, primarily corn and soybeans (see Moore and Swihart 2005 for a detailed description of the study area). At each site, we placed a plywood platform (0.6 m × 0.6 m × 0.2 m) mounted on a wooden fence post (10 cm × 10 cm × 1.5 m tall). Platforms were located at woodland edges, immediately outside the dripline of the nearest tree, and posts were wrapped in loose aluminum flashing to discourage climbing by raccoons and squirrels (Fig. 1a). Data for this study were collected from 19 October to 7 December 2009. However, the platforms had been oper- ated continuously from July 2007 as part of another continu- ing study (Lichti 2012). When not in use for presentation of acorns, they were stocked once per week with sunflower seeds, cracked corn, whole peanuts, and acorns. Acorns were pre- sented on platforms at multiple times from October to March and monitored with remote cameras (Reconyx, Inc., Holmen, WI). A change in camera placement for 2009 enabled collec- tion of the current dataset. Prior to 2009, platforms were monitored by a single camera, mounted on a separate fence post 1.5–2 m from the platform. Beginning in October 2009, we used a system of two cameras to provide better data on acorn selection. Poles 2.4 m in height and 2.5 cm in diameter were bracketed to either side of the platform and supported a pair of cameras, which were positioned 1.2 m above the platform surface, facing directly down. We used two cameras to ensure that all motion on the board was captured while still offering a clear view of the platform. Camera poles were covered in axle grease to prevent mammals from climbing them. On the top of the platform we placed a plywood tray with a 15 × 15 grid (225 cells total) of numbered holes, each 2 cm deep and 2.5 cm in diameter, drilled into its surface (Fig. 1b and c). Each hole was large enough to accommodate an acorn and deep enough to hold the acorns in place but with sufficient room for a bird to easily grasp the acorn and pull it out. Each tray was divided into nine sectors consisting of 25 cells each, and the surface of each sector was painted white, light gray, or dark gray to allow easy discrimination in black-and-white photos (Fig. 1b and c). In the deciduous forests of eastern North America, acorns are eaten by >44 species of wildlife, including squirrels ( Sciurus spp.), mice, eastern chipmunks ( Tamias striatus ), white-tailed deer ( Odocoileus virginianus ), and northern raccoons ( Procyon lotor ), as well as an avian assemblage that includes the Red-bellied Woodpecker ( Melanerpes car- olinus ), Blue Jay ( Cyanocitta cristata ), Tufted Titmouse ( Baeolophus bicolor ), White-breasted Nuthatch ( Sitta caro- linensis ), and Wild Turkey ( Meleagris gallopavo ), among others (Martin et al. 1951). The Blue Jay is considered the pri- mary avian disperser of oaks in eastern North America (John- son et al. 1997, Moore and Swihart 2006) and has received more attention from researchers than have other species. Blue Jays can remove thousands of acorns from oak trees, transport acorns up to several kilometers away from a seed source, and scatter-hoard them in various locations for later consumption (Darley-Hill and Johnson 1981, Johnson and Adkisson 1985). In addition to Blue Jays, Red-bellied Woodpeckers, Tufted Titmice, and White-breasted Nuthatches also scatter- hoard food items in eastern North American forests (Woodrey 1991, Carrascal 1993, Lucas et al. 1993, Pravosudov and Grubb 1997). Although titmice cache acorns in the litter layer, where they may germinate and grow into seedlings (Bent 1946), woodpeckers and nuthatches usually cache in trees (Waite and Grubb 1988, Petit et al. 1989, Vander Wall 1990). Nuthatches and woodpeckers therefore act mainly as seed predators rather than as seed dispersers. As a result, differences in the patterns of these four species’ foraging may influence oak dispersal and recruitment substantially. When foraging patterns are considered, it is important to distinguish between usage, which is simply the amount of a resource consumed by a forager, and preference, which is defined as the likelihood of using a particular resource relative to a specific set of alternatives (Manly et al. 2002). Knowledge of foraging preferences can help to predict how consumers might respond to changes in seed availability over space or time as a result of trees’ seasonal phenology or annual vari- ation in their masting (Koenig and Knops 2002, Lusk et al. 2007). Previous studies have demonstrated that Blue Jays preferentially forage on smaller acorns such as those of north- ern pin oak ( Q. palustris ), willow oak ( Q. phellos ), and post oak ( Q. stellata ) (Darley-Hill and Johnson 1981, Scarlett and Smith 1991, Johnson et al. 1997). However, data on foraging preferences in other bird species is limited. To better understand how the guild of avian granivores in eastern forests might affect oak regeneration, we studied the foraging preferences of free-ranging wild birds in Indi- ana for acorns of three species: white oak ( Q. alba ), black oak ( Q. velutina ), and northern pin oak. In general, birds’ prefer- ences for seed size correlate directly with body and bill size (Díaz 1990). Of the three oak species we tested, seed size ranked pin oak < black oak < white oak. The Blue Jay is the largest bird in our study and prefers pin oak to black or white Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 ACORN-FORAGING PREFERENCES OF AVIAN SEED PREDATORS 865 Colors were arranged systematically on the tray so that no two adjoining sectors had the same color. At the beginning of each trial, we assigned white oak, black oak, and northern pin oak acorns randomly to one of the three sector colors and filled the tray with acorns. Acorns of the same species were similar in size, but mass varied significantly by species, with white oak (3.4 ± 0.6 g; mean ± SD, n = 99) > black oak (2.6 ± 0.6 g, n = 102) > pin oak (2.1 ± 0.4 g, n = 106) (ANOVA with Tukey’s HSD, F 2,304 = 134.0, P < 0.001 for all pairwise comparisons). We checked the platforms each morning for the next 4 days or until all acorns had been removed by animals. Acorns that had been removed from a cell but left on top of the platform were put back in the nearest cell of the appropriate color. Animals usually emptied the platforms within 1 or 2 days. We ran 12 trials, 2 at each site. The time between trials at the same site ranged from 19 to 30 days. The two cameras placed above each platform provided a photographic record of animal visits. Cameras were set to their maximum sensitivity level with no delay between exposures. Each time a camera was triggered, it took 10 consecutive photos at intervals of ~0.5 sec. The cameras also were set to take a picture every hour regardless of motion. For each trial, we recorded the time and date of each visit to a plat- form, species identity of visitors, and whether or not they took an acorn. When an animal took an acorn (a selection ), we also recorded the species of acorn selected and attempted to quan- tify the set of acorns that were available on the platform at the time of selection (the choice set ). If a single bird took multiple acorns on a single visit (typical of the Blue Jay), we treated each acorn as a separate selection. To ensure that data collection was as standardized as pos- sible, two technicians interpreted all photos. We also recorded the confidence of our determination of the choice set for each selection on a subjective five-point scale. Confidence took into account factors such as delay between the two cameras, lack of visibility in some areas of the grid because of debris, snow, or poor lighting, or the presence of acorns of an unknown spe- cies on the board (this occurred because animals occasionally moved acorns from cell to cell or brought acorns to the plat- form from elsewhere). Preliminary analyses showed no qual- itative differences between results based on the full dataset and those based on high-confidence assignments only (scores ≥ 4). We therefore present results for the full dataset. STATISTICAL ANALYSES We analyzed relationships among the numbers of visits by birds of each species with log-link negative binomial general- ized linear models, fit with the package M ASS in R (Venables and Ripley 2002, R Development Core Team 2012). We ana- lyzed the choice data with a generalized linear mixed model in WinBUGS 1.4.3 (Spiegelhalter et al. 2003). Because birds were not marked, a single individual may have contributed to multiple selections during a single trial or to selections FIGURE 1. Photographs depicting the manner in which the plat- form and cameras were mounted to posts (a) and the coverage of the two cameras to capture division of the tray into 9 sectors of 25 cells each (b and c). Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 866 KIMBERLY B. RICHARDSON ET AL during multiple trials within a site. In addition, we expect the availability of unobserved ambient mast varied by site and over time; although we cannot know how variation in background availability of seeds influenced choice behavior at our plat- forms, it is reasonable to assume that background conditions were more similar during one trial or between trials within the same site than at different sites. To account for the potential non-independence of observations, mixed models structure the stochastic component in a linear model according to the random effects (they do not assume complete independence). Specifically, our model included two nested levels of random effects, representing trial and site. This allowed for the pos- sibility that selections within a trial (and trials within a site) may be autocorrelated. Such a scenario might arise if individ- ual birds’ preferences differ strongly and a small number of individuals were responsible for the majority of selections at each trial, or if ambient conditions have a strong effect on the choice behavior observed at platforms. For each selection i , the bird had three potential options to select among, corresponding to the three species of acorns. We modeled the probability that it would select acorn species j as a multinomial logistic regression, P y j a X a X N N exp exp , , i ji j ts ki k ts k ts s s 1 3 2 2 ∑ ( ) ( ) ( ) ( ) ( ) = = β β β μ σ μ η σ = β μ where y i is the species of acorn chosen in selection i , a ji is the relative abundance of acorn species j in the choice set for selection i , X j is a dummy-coded vector that indicates acorn species, and β ts is a random effect coefficient that measures the strength of the bird species’ preference for ( β > 0) or against ( β < 0) acorn species j in trial t at site s . The coefficients μ s and η represent the mean preference for the bird species across trials at site s and the species’ overall mean preference, respectively. Similarly, σ 2 β and σ 2 μ estimate the within-site variance and among-site variance in strength of preference, respectively. Bird species with a strong, consistent preference for an acorn species will have relatively high absolute values for η with low variances, indicating that the β values for indi- vidual trials clustered tightly around the species-level mean, whereas high β values coupled with high variance and μ or η near zero indicate strong preferences within individual trials but inconsistent behavior across trials (μ ≈ 0) or among sites ( η ≈ 0). Bayesian statistical approaches require analysts to specify a prior probability distribution for each unknown quantity in a model. These priors describe the researchers’ beliefs regarding the true value of the parameters before the model is fit to any data. Typically, prior distributions with very high variances (“vague priors”) are selected to ensure that estimates are deter- mined only by the data. We used vague, normal(0, 10) priors for β , μ, and η and flat, uniform(0, 10) priors for the standard deviations. We based our inference on 5000 Markov-chain Monte Carlo (MCMC) draws from the joint posterior param- eter estimates. We evaluated convergence by using a Gelman– Rubin statistic cutoff of 1.1 applied to three parallel chains (Gelman et al. 2004). We analyzed each species of bird sepa- rately and used only those instances of selection for which at least two species of acorn were present. We used two approaches to evaluate models’ overall per- formance. First, we calculated a Bayesian posterior predictive p -value, p B (Gelman et al. 2004:162–165), which quantifies a model’s goodness of fit by comparing a test statistic generated with the actual dataset to the distribution of that statistic when calculated from replicate datasets that have been simulated by the model. If >90% of simulated test statistics fall above or below the true value, the model is unlikely to have produced the observed data, and its fit is therefore poor. Ideally, p B = 0.5. We used the sum of the Pearson residuals, ( y ij – p ij ) 2 / p ij (1 – p ij ), where p ij = P ( y i = j ) and y ij = 1 if y i = j and 0 otherwise, as our test statistic. In addition to goodness of fit, we estimated the mean pro- portion of observations correctly predicted by the model, which we refer to as CP. Given a specific set of parameter values, θ , we estimated CP| θ (“CP conditional on θ ”) by generating 500 simulated datasets from the model, determining for each data- set the proportion of choices that concurred with the observed data, and taking the mean over the set of simulations. For the fitted model, we calculated CP| θ separately for each MCMC draw, then took CP as the mean of the conditional values over the posterior. This we compared to the CP value for a null model, in which selection was assumed to be proportional to availability (i.e., β ts = 0 for all trials). We did not use a specific significance threshold to evalu- ate preference parameters. Instead, we considered selection strong if ≥ 95% of the posterior distribution of η fell to one side of zero, moderate if 90–95% fell to one side of zero, and weak or negligible otherwise. RESULTS A total of 26 822 photos were taken at the platforms, which pro- vided data on 2278 individual visits and complete choice-set and selection data for 941 selections. In descending order, the four most common bird species observed at platforms were the Blue Jay, Tufted Titmouse, Red-bellied Woodpecker, and White-breasted Nuthatch (Table 1). Other visitors included the Black-capped Chickadee ( Poecile atricapillus , 66 vis- its), Northern Cardinal ( Cardinalis cardinalis , 41), Dark-eyed Junco ( Junco hyemalis , 21), Downy Woodpecker ( Picoides pubescens , 5), Red-breasted Nuthatch ( Sitta canadensis , 1), and Brown Thrasher ( Toxostoma rufum , 1) as well as two mammals, Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 ACORN-FORAGING PREFERENCES OF AVIAN SEED PREDATORS 867 Fig. 2a). The White-breasted Nuthatch had a similar but non- significant relationship with the Blue Jay (Fig. 2b), while the Red-bellied Woodpecker showed no relationship to Blue Jay visitation (Table 2). For data pooled across sites and trials, the order in which birds removed acorns from platforms varied by species (Fig. 3, χ 2 12 = 72.2, P < 0.001). The Blue Jay was the most likely spe- cies to remove acorns at all times during a trial, but the propor- tion of acorns removed by Blue Jays tended to decline as the set of available acorns was depleted ( χ 2 4 = 9.0, P = 0.06). Tufted Titmice also removed fewer acorns late in trials ( χ 2 4 = 18.7, P < 0.001). In contrast, Red-bellied Woodpeckers removed larger the eastern gray squirrel ( Sciurus carolinensis , 22) and raccoon (15). When mammals visited the platforms, they generally re- moved (squirrels) or ate (both squirrels and raccoons) most of the acorns on the platform. Different assemblages of the four most common birds visited platforms at our six study sites (Table 1). Although Blue Jays and Tufted Titmice were routinely observed at all sites (NIL), each of these two species failed to visit platforms at one site, and Blue Jays only rarely visited the platform at site 1 (Table 1). In addition, we noticed a significant negative relationship between the number of visits by Blue Jays during a trial and the number of visits by Tufted Titmice (Table 2, TABLE 1. Visits to feeding platforms and acorn selection by four species of granivo- rous birds in Indiana. For each bird species and site, columns show the number of visits to platforms (visits), the total number of acorns taken (selections), and the numbers of acorns of the three species of oak tested. Oak Species and site Visits Selections Northern Pin Black White Red-bellied Woodpecker 1 29 53 2 49 2 2 0 0 0 0 0 3 40 9 0 0 9 4 0 0 0 0 0 5 16 5 4 1 0 6 69 46 16 28 2 Total 154 113 22 78 13 Blue Jay 1 6 1 1 0 0 2 75 117 57 46 14 3 126 189 59 72 58 4 30 101 48 35 18 5 213 100 47 36 17 6 0 0 0 0 0 Total 450 508 212 189 107 Tufted Titmouse 1 206 94 40 8 46 2 60 41 8 4 29 3 26 5 1 0 4 4 13 8 2 2 4 5 0 0 0 0 0 6 292 132 65 28 39 Total 597 280 116 42 122 White-breasted Nuthatch 1 17 6 3 3 0 2 0 0 0 0 0 3 14 3 2 0 1 4 6 1 0 0 1 5 0 0 0 0 0 6 33 2 0 0 2 Total 70 12 5 3 4 Grand total 1271 913 355 312 246 Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 868 KIMBERLY B. RICHARDSON ET AL proportions of acorns later in the trials ( χ 2 4 = 35.5, P < 0.001). White-breasted Nuthatches visited the platforms less often than the other three species (Table 1) and removed most acorns at the beginning or end of the trials ( χ 2 4 = 13.0, P = 0.01). By accounting for the relative availability of differ- ent acorn species during each selection, the discrete-choice analysis revealed moderate to strong preferences among the three acorn species tested in all four bird species (Table 3). In general the models fit well and correctly predicted a high proportion of individual choices (Blue Jay: p B = 0.47, CP = 0.65; Red-bellied Woodpecker: p B = 0.24, CP = 0.78; Tufted Titmouse: p B = 0.38, CP = 0.54; White-breasted Nuthatch: p B = 0.27, CP = 0.66). Relative to the null model of selection proportional to availability, the preference model improved CP scores for all four bird species (Blue Jay, 102% improve- ment; Red-bellied Woodpecker, 44%; Tufted Titmouse, 67%; White-breasted Nuthatch, 78%). FIGURE 2. Predictions of generalized linear model (solid line) for the number of visits by Tufted Titmice (a) and White-breasted Nuthatches (b) during a trial, as a function of the number of visits by Blue Jays. Dashed lines show the 95% confidence envelope for the predicted relationship, as determined by 50 000 bootstrap samples from the variance–covariance matrix for the fitted model. Points show observed values. TABLE 2. Relationships between the number of visits by Blue Jays during a trial and visits by other bird species. Relationships were estimated by fitting separate negative binomial generalized linear models for each pair of species. The parameter θ accounts for extra-Poisson variation in the number of visits during a trial (Ven- ables and Ripley 2002). Species and parameter Mean SE Wald’s z P Red-bellied Woodpecker Intercept 0.403 0.546 0.739 0.46 Blue Jay 0.010 0.060 0.167 0.87 θ 0.595 0.447 Tufted Titmouse Intercept 4.169 0.468 8.904 <0.001 Blue Jay –0.125 0.057 –2.196 0.03 θ 0.604 0.233 White-breasted Nuthatch Intercept 1.561 0.270 5.791 <0.001 Blue Jay –0.083 0.048 –1.720 0.09 θ 4.19 4.05 FIGURE 3. Changes in the proportion of acorns taken from plat- forms by four granivorous bird species over time. Acorns were clas- sified into five groups according to the order in which they were removed from platforms (rank order, horizontal axis). As platforms became depleted, acorn removal by Blue Jays (squares) and Tufted Titmice (circles) declined while the proportion of acorns removed by Red-bellied Woodpeckers (triangles) increased. The proportion re- moved by White-breasted Nuthatches (plus signs) remained stable. Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 ACORN-FORAGING PREFERENCES OF AVIAN SEED PREDATORS 869 Blue Jays strongly preferred pin oak over black oak, and black oak over white oak; these preferences were rel- atively consistent both within and across study sites (Ta- ble 3). The estimated odds ratios based on the preferences in Table 2 indicated that, on average, a Blue Jay presented with a choice of two acorns was 13.4 times more likely to select a pin oak over a black oak, and 132.0 times more likely to select a pin oak over a white oak. By comparison, Blue Jays were 9.8 times more likely to select a black oak than a white oak acorn. The Red-bellied Woodpecker’s preferences were on average similar to that of the Blue Jay, but more variable, both within and among sites (larger values for σ β and σ μ in Table 3). Despite tak- ing >3 times as many black oak as pin oak acorns from platforms (Table 1), Red-bellied Woodpeckers presented with the choice between pin and black oak acorns were 7.9 times more likely, TABLE 3. Parameter estimates derived from posterior distributions for a mixed- effects discrete-choice model of acorn preferences of four granivorous bird species in Indiana. Species and preference a Parameter b Mean SE 95% C. I. P (>0) c Red-bellied Woodpecker NPO:BLO η 2.07 1.81 (–1.43, 5.82) 0.88 σ β 5.22 2.78 (0.48, 9.74) 1 σ μ 5.91 2.65 (0.65, 9.83) 1 WHO:BLO η –1.18 1.72 (–4.58, 2.24) 0.24 σ β 4.63 2.75 (0.43, 9.70) 1 σ μ 5.06 2.87 (0.31, 9.76) 1 Blue Jay NPO:BLO η 2.59 0.82 (0.99, 4.44) 0.997 σ β 1.23 1.43 (0.09, 5.67) 1 σ μ 2.11 2.35 (0.03, 8.65) 1 WHO:BLO η –2.29 1.17 (–4.71, 0.06) 0.027 σ β 2.65 2.32 (0.16, 8.73) 1 σ μ 3.86 2.72 (0.19, 9.51) 1 Tufted Titmouse NPO:BLO η 2.95 1.11 (0.60, 5.27) 0.992 σ β 3.51 2.44 (0.46, 9.22) 1 σ μ 2.84 2.61 (0.06, 9.05) 1 WHO:BLO η 2.11 1.18 (–0.26, 4.45) 0.96 σ β 3.36 2.46 (0.31, 9.17) 1 σ μ 3.79 2.70 (0.15, 9.43) 1 White-breasted Nuthatch NPO:BLO η 2.56 2.33 (–2.09, 7.21) 0.87 σ β 5.58 2.71 (0.64, 9.83) 1 σ μ 4.71 2.89 (0.19, 9.69) 1 WHO:BLO η 1.83 2.50 (–3.08, 6.88) 0.77 σ β 5.17 2.86 (0.38, 9.77) 1 σ μ 5.51 2.82 (0.36, 9.79) 1 a Acorn species are northern pin oak (NPO), black oak (BLO), or white oak (WHO), and all preferences are reported relative to black oak. b η , overall mean preference; σ β , within-site standard deviation in preference; σ μ , between-site standard deviation in preference. c P (>0) refers to the fraction of the posterior distribution greater than zero. Note that P (<0) = 1 – P (>0). Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 870 KIMBERLY B. RICHARDSON ET AL on average, to select a pin oak. However, a considerable fraction (0.13) of the posterior density of η (pin:black) for the Red-bellied Woodpecker fell below zero, so we cannot confidently reject the null hypothesis that Red-bellied Woodpeckers have no prefer- ence for one of these species over the other. In addition, we found only weak support for the hypothesis that Red-bellied Wood- peckers prefer black oak over white oak (Table 3). However, the model predicted that Red-bellied Woodpeckers presented with pin and white oak acorns were 25.7 times more likely to select the pin oak. Of the posterior MCMC draws, 91% produced odds ratios >1 for this comparison, indicating moderate support for the hypothesis that Red-bellied Woodpeckers prefer pin oak to white oak acorns. Unlike the Blue Jay and Red-bellied Woodpecker, the Tufted Titmouse and White-breasted Nuthatch appeared to prefer both pin and white oak acorns over black oak acorns (Table 3). Support for this conclusion was strong in the tit- mouse, but weaker in the nuthatch (see credible intervals and probabilities for η in Table 3). In pairwise comparisons, Tufted Titmice were estimated to be 19.0 and 8.2 times more likely to select pin and white oak acorns, respectively, than black oak. White-breasted Nuthatches were 13.0 and 6.2 times more likely to select pin and white oak, respectively. In the comparison of pin oak versus white oak, mean odds ratios ( P [odds ratios > 1]) for the titmouse and nuthatch to select pin oak were 2.3 (0.71) and 2.1 (0.60), respectively. Thus there is only weak support for the hypothesis that Tufted Titmice prefer pin oak over white oak acorns and no evidence for a preference in the White-breasted Nuthatch. DISCUSSION Our results demonstrated strong preferences for differ- ent species of acorn in the Blue Jay and Tufted Titmouse, but evidence of preferences in the Red-bellied Woodpecker and White-breasted Nuthatch was weak. In addition, prefer- ences varied by bird species, with rankings of pin oak ≥ black oak ≥ white oak for the Blue Jay and Red-bellied Woodpecker but pin oak ≥ white oak ≥ black oak for the Tufted Titmouse and White-breasted Nuthatch. Thus our predictions regard- ing seed size were supported for the Blue Jay and Red-bellied Woodpecker but not for the smaller Tufted Titmouse and White-breasted Nuthatch. This was particularly surprising given that Red-bellied Woodpeckers and Blue Jays weigh 3–4 times more than the smaller birds. Numerous chemical and physical attributes of acorns may influence preference, including energy, protein and tan- nin content, seed size and hardness, and dormancy status. Typically, the effects of chemical and nutrient composition on birds’ decisions to select a food are thought to be secondary; instead, birds tend to focus on those seed species that they can process most easily (Díaz 1996). Moore and Swihart (2006) examined preferences of captive Blue Jays under conditions of varying availability and found preferences for small acorns as well as thinner-shelled seeds. Unlike us, they also found a slight preference for white oak over black oak acorns. How- ever, the birds in their study consumed most of the acorns that they selected and did not have any opportunity to create long-term caches. As a result, they may have attached differ- ent subjective values to seeds’ traits than did the wild birds we studied. Bossema (1979) found that free-ranging Eurasian Jays ( Garrulus glandarius ) favored medium-sized (3–5 g) English oak ( Q. robur ) acorns and also preferred English and sessile oak ( Q. petraea ) to European beech ( Fagus sylvatica ) or hard-shelled northern red oak ( Q. rubra ). Bossema (1979) and Moore and Swihart (2006) both noted that birds had dif- ficulty handling larger, harder-shelled seeds. In another study of the Eurasian Jay, Pons and Pausas (2007) found a consistent preference for Holm oak ( Q. ilex ) over three other Mediterranean oak species in two separate sets of trials. In the first set, acorn size was not controlled and preferences for acorn species were correlated with seed size. However, a second set of trials that controlled for seed size showed a similar order of seed preferences. Moreover, within any given acorn species, Eurasian Jays removed larger seeds before smaller ones. As a result, Pons and Pausas (2007) attributed the preference for Holm oak to interspecific differ- ences in nutrient and tannin content and concluded that, all else being equal, Eurasian Jays preferred larger seeds within the range of sizes tested. Of the three species of oaks we studied, white oak acorns are largest, pin oak acorns are smallest, and black oak acorns are intermediate in size. However, black oak acorns have a harder shell than either pin or white oak. Bomb calorimetry on seed kernels shows the highest energy densities for black oak (20.0 ± 1.1 kJ g –1 mean ± SE), followed by pin oak (16.2 ± 1.1 kJ g –1 ), and white oak (10.4 ± 0.6 kJ g –1 ) (Sundaram and Swihart, unpubl. data). Protein concentration follows a similar pattern (10% dry mass in black oak, 9% in pin oak, and 6% in white oak; Sundaram and Swihart, unpubl. data), but tannin con- centrations are much higher in black oak, intermediate in pin oak (Lewis 1982, Bonner and Vozzo 1987, Briggs and Smith 1989, Vander Wall 2001). High tannin content reduces the protein available to consumers (Johnson et al. 1993), so available protein may actually be greater in pin or white oak acorns. Furthermore, ingestion of tannins may incur the cost of detoxification, constraining dietary choices by the need to regulate tannin intake (Dearing et al. 2005). In addition, acorns of the white oak do not go dormant, whereas acorns of oaks of the section Lobatae , including both pin and black oak, remain dormant over the winter (Bonner 2008) and thus may be more suitable for long-term storage in caches. Given these characteristics, Blue Jays and Red-bellied Woodpeckers preferred smaller, dormant acorns with Downloaded from https://academic.oup.com/condor/article/115/4/863/5152884 by University of Florida - Chiles Info Ctr user on 23 October 2024 ACORN-FORAGING PREFERENCES OF AVIAN SEED PREDATORS 871 intermediate energy and tannin content, whereas Tufted Titmice and White-breasted Nuthatches preferred acorns with thinner shells and low or moderate energy and tan- nin concentrations. If the harder shell of black oak imposes greater handling costs on the smaller Tufted Titmouse and White-breasted Nuthatch than on the larger birds, then the data for all four species may be consistent with a strategy that maximizes the rate of energy intake. In many of the New World jays, including the Blue Jay, morphological ad- aptations of the mandible and quadrate have evolved that allow hard seeds to be opened by chiseling, rather than by hammering (as in titmice) or crushing (Zusi 1987). Chis- eling allows Cyanocitta and Aphelocoma jays to eat hard- shelled seeds such as northern red oak that the much larger Eurasian Jay and magpies ( Pica spp.) rarely succeed in opening. In woodpeckers, adaptations for digging into wood may serve a similar function. Moore and Swihart (2006) considered a wider range of species in terms of seed hardness, including both northern red oak and shagbark hickory ( Carya ovata ), finding birds preferred smaller nuts with thinner seed coats. We do not know how most of the seeds in our study were used. However, if our subjects were focused primar- ily on caching rather than eating seeds immediately, then they may have aimed to maximize the rate at which they stored recoverable resources (Waite and Ydenberg 1994). Blue Jays load themselves with multiple food items, so they may prefer smaller seeds, more of which they can carry in the esophagus simultaneously (Bossema 1979, Bartlow et al. 2011). In addition, because white oak acorns do not remain dormant over winter, they are more likely to escape from ground caches by germinating (Steele and Smallwood 2002). The risk of germination should be lower for the Red- bellied Woodpecker and White-breasted Nuthatch, both of which consistently cache in trees where acorns are likely to desiccate before germinating (Waite and Grubb 1988, Petit et al. 1989). We cannot test this hypothesis with our current dataset, but it is consistent with the White-breasted Nuthatch’s relative lack of discrimination between white and pin oak. We found that in addition to seed preferences, seed usage was clearly affected by availability. For instance, when both black and pin oak acorns were available, Red- bellied Woodpeckers were 7.9 times more likely to choose pin oaks. However, they transported three times more black oak than pin oak acorns from our platforms. Dur- ing feeding trials Blue Jays typically were the first birds to visit platforms, followed by smaller birds and Red-bel- lied Woodpeckers. By visiting platforms later, Red-bellied Woodpeckers had fewer choices than in the original set of acorns and instead were left to choose from the subset of acorns left by the Blue Jays, mostly black and white oak. We also observed direct interactions between Blue Jays and Red-bellied Woodpeckers on the platforms. However, in- terference competition probably does not explain the heavy use of black oak by the Red-bellied Woodpecker, as it ap- pears dominant over the Blue Jay (Kilham 1963; NIL, pers. obs ). Under natural conditions competition for limited food resources frequently may make the availability of a food item more important than innate preferences and lead to circumstances in which the dominant food item selected is not the most preferred. Direct competition may play a larger role in interactions between Blue Jays and Tufted Titmice (and, to a lesser extent, Blue Jays and White-breasted Nuthatches) than between Blue Jays and Red-bellied Woodpeckers. Our analysis of the num- bers of visits by Blue Jays and Tufted Titmice suggests that the latter species avoids foraging in locations where the former is active. Corvids, including Cyanocitta jays, may pilfer food they observe other animals caching (Heinrich and Pep- per 1998, Bugnyar and Kotrschal 2002, Thayer and Vander Wall 2005). Indeed, we observed an example of this behavior during our study. A Tufted Titmouse cached a peanut at the base of a tree near a platform while being observed by a Blue Jay in another tree nearby. Almost immediately after the tit- mouse left the cache, the jay flew to the location, collected the peanut, and flew away (NIL, pers. obs ). If hoarders know that potential cache pilferers are active in an area, they may stop foraging at that location and move elsewhere to