Context‐dependent responses of food‐hoarding to competitors in Apodemus peninsulae: implications for coexistence among asymmetrical species

Abstract Superior species may have distinct advantages over subordinates within asymmetrical interactions among sympatric animals. However, exactly how the subordinate species coexists with superior species is unknown. In the forests west of Beijing City, intense asymmetrical interactions of food competition exist among granivorous rodents (e.g. Apodemus peninsulae, Niviventer confucianus, Sciurotamias davidianus and Tscherskia triton) that have broadly overlapping habitats and diets but have varied body size (range 15–300 g), hoarding habits (scatter vs larder) and/or daily rhythm (diurnal vs nocturnal). The smallest rodent, A. peninsulae, which typically faces high competitive pressure from larger rodents, is an ideal model to explore how subordinate species coexist with superior species. Under semi‐natural enclosure conditions, we tested responses of seed‐hoarding behavior in A. peninsulae to intraspecific and interspecific competitors in the situations of pre‐competition (without competitor), competition (with competitor) and post‐competition (competitor removed). The results showed that for A. peninsulae, the intensity of larder‐hoarding increased and the intensity of scatter‐hoarding declined in the presence of intraspecifics and S. davidianus, whereas A. peninsulae ceased foraging and hoarding in the presence of N. confucianus and T. triton. A. peninsulae reduced intensity of hoarding outside the nest and moved more seeds into the nest for larder‐hoarding under competition from intraspecific individuals and S. davidianus. In most cases, the experimental animals could recover to their original state of pre‐competition when competitors were removed. These results suggest that subordinate species contextually regulate their food‐hoarding strategies according to different competitors, promoting species coexistence among sympatric animals that have asymmetrical food competition.


INTRODUCTION
Food-hoarding is an evolutionary adaption that has evolved in some animals (e.g. rodents and birds) in response to spatial and temporal variation in the availability of resources (Vander Wall 1990). Animals of-ten store food during a food-rich period (e.g. in autumn) and rely on these foods for survival and/or reproduction in the following periods of food scarcity (e.g. in winter and early spring), thereby increasing fitness of survival and reproduction (Vander Wall 1990). The food-hoarding spectrum varies from larder-hoarding (concentrating food items in a few locations) to scatter-hoarding (with multiple small caches) under different conditions (Vander Wall 1990;Dally et al. 2006).
A major requisite for evolution of food-hoarding behavior is that hoarders have to gain an advantage from hoarding events thereby animals strive to protect stores and retrieve more caches during the period of food scarcity than native pilferers (Vander Wall 1990;Gu et al. 2017). As an important consequence of food loss, food pilferage by intraspecific and interspecific competitors threatens the survival and the reproductive capacity of hoarders when food is scarce (Wauters et al. 1995;Vander Wall & Jenkins 2003;Gerhardt 2005). Evidently, food hoarders are sensitive to competitors and have evolved an array of strategies to reduce or prevent the risk of pilferage posed by competitors (reviewed by Vander Wall & Jenkins 2003;Dally et al. 2006;Grodzinski & Clayton 2010).
Food hoarders can adopt one or more of the following strategies to compete for food and/or minimize and compensate for food loss. First, food hoarders can limit or invalidate the information that a potential pilferer has gathered (MacDonald 1976;Dally et al. 2006;Grodzinski & Clayton 2010). Some hoarders hoard foods in secret locations far away from the sources (Galvez et al. 2009;Zhang et al. 2014a;Steele et al. 2014Steele et al. , 2015, recache items to invalidate pilferers' sensory information (Dally et al. 2005(Dally et al. , 2006, use misinformation (e.g. empty caches) to confuse potential pilferers (Bugnyar & Kotrschal 2004;Steele et al. 2008) and/or conceal auditory information from potential pilferers (Stulp et al. 2009). Second, animals can tolerate a partial loss of their total caches because they hoard more food items than they need or steal caches from other hoarders to compensate for the loss (Vander Wall & Jenkins 2003). Third, animals compensate for food loss by increasing eating and/or hoarding intensity (Huang et al. 2011;Zhang et al. 2011;Luo et al. 2014). Finally, animals aggressively prevent competitors from accessing cached sites (Clarke & Kramer 1994), or shift from scatter-hoarding to larder-hoarding to facilitate cache protection or vice versa to avoid a complete loss (Jenkins et al. 1995;Preston & Jacobs 2001;Huang et al. 2011;Zhang et al. 2011;Luo et al. 2014). These strategies are not mutually exclusive and are context-dependent for a given species. While comprehensive, most of the above responses and adaptations have been described from patterns of intraspecific competition/pilferage, and more research is needed to study food competition between interspecific hoarders (but see Leaver & Daly 2001;Thayer & Vander Wall 2005;Vander Wall et al. 2009;Penner & Devenport 2011;Zhang et al. 2013b;Dittel et al. 2017).
Interactions in food-hoarding between sympatric rodents are associated with species coexistence (Leaver & Daly 2001;Price & Mittler 2003Vander Wall et al. 2009;Penner & Devenport 2011;Luo et al. 2014;Dittel et al. 2017;Gu et al. 2017;Wang et al. 2018). Due to differential capabilities of food competition (e.g. body size), sympatric species usually exhibit asymmetrical interactions in that some species (superior species hereafter) may have distinct advantages over others (subordinate species hereafter) Penner & Devenport 2011;Zhang et al. 2014b;Dittel et al. 2017). Asymmetrical competition in food-hoarding and pilferage has been observed to occur between intraspecific individuals that have different social statuses (e.g. Zhang et al. 2011;Huang et al. 2019) and sympatric species that have different body sizes, hoarding habits, daily rhythms and/or diets (e.g. Vander Wall et al. 2009;Penner & Devenport 2011;Zhang et al. 2014b;Dittel et al. 2017;Gu et al. 2017). In the case of asymmetrical interactions, context-dependent regulation of hoarding strategy is critically important for survival of the subordinate species that endures heavy cache losses, and coexistence of sympatric species that share similar habitats and diets. However, this behavioral plasticity of food-hoarding has received little attention.
At our study site in the Donglingshan mountains west of Beijing City, there exists a rich community of granivorous rodents [e.g. Apodemus peninsulae (Thomas, 1907), Niviventer confucianus (Milne-Edwards, 1871), Sciurotamias davidianus (Milne-Edwards, 1867), Tscherskia triton (de Winton, 1899), Apodemus agrarius (Pallas, 1771) and Eutamias sibiricus (Laxmann, 1769)] that have broadly overlapping habitats (e.g. secondary forests, shrublands and abandoned farmlands) and diets (e.g. seeds of Quercus wutaishanica, Armeniaca sibirica, Juglans mandshurica and Amygdalus davidiana) but have varied body size (range 15-300 g), hoarding habits (scatter vs larder) and/or daily rhythms (diurnal vs nocturnal) (Table 1) (Li et al. 2004;Zhang & Zhang 2008;Zhang et al. 2015). These rodent species not only compete for temporally limited plant seeds of the par-ent trees during the seedfall periods (summer and autumn), but also compete for caches over longer periods after seedfall (winter and early spring), which is essential for their survival during hard winters and reproduction in the following spring . These rodent species have adapted an array of strategies to compete for seed resources before hoarding and protect/pilfer caches after hoarding (Huang et al. 2011;Zhang et al. 2011Zhang et al. , 2014aLuo et al. 2014). For example, A. peninsulae, N. confucianus and S. davidianus increase hoarding intensity even at times when their stores are repeatedly and completely lost (Huang et al. 2011;Luo et al. 2014). S. davidianus first hoards seeds around seed stations to rapidly sequestrate resources and subsequently transfers these seeds to more secure places (e.g. home range) to facilitate protection (Zhang et al. 2014a). E. sibiricus can use auditory and visual cues of intraspecific hoarders to pilfer caches of others (Niu et al. 2019). Besides increasing hoarding intensity, A. peninsulae shifts from larder-hoarding to scatter-hoarding in the presence of intraspecific individuals and vice versa when interspecific competitors (N. confucianus) are present . These interactions of hoarding and pilferage between sympatric species are associated with species coexistence, but interaction networks have not been established at the community level.
The difference in body size, caching and pilferage ability between the rodent species at our study site seems to create an asymmetry in the species interactions, whereby some large species (e.g. N. confucianus, T. triton and S. davidianus) with high competitive ability gain more advantages from food competition and pilferage than the small species (e.g. A. peninsulae) (also see Vander ). Asymmetrical food competition has been observed at the level of pairwise species interactions in the rodent community at our study site.
For example, N. confucianus (larger-size, larder-hoarding) can pilfer caches from A. peninsulae (small-size, larder and scatter-hoarding), but the converse does not occur (Zhang et al. 2014b). A. peninsulae space their hoarded seeds at a certain depth that can partially prevent pilferage by N. confucianus (Zhang et al. 2014b). However, asymmetrical food competition among multiple species has not been studied. Under the conditions of asymmetrical food competition, an important but less well known question is how the subordinate species (e.g. A. peninsulae) regulate their behavior to maximize fitness during the hoarding events. In the rodent community at our study site, A. peninsulae gives us an ideal model to explore this question because this rodent species has the smallest body size but it has to face high risks of competition by sympatric species (e.g. N. confucianus, S. davidianus and T. triton) when it hoards plant seeds.
Under enclosure conditions, we tested the food-hoarding behavioral response of A. peninsulae individuals to intraspecific and interspecific competitors when competitors were unavailable (pre-competition), available (competition) and then removed (post-competition). We wanted to know how the experimental animals respond to different competitors and if they can recover to pre-competition status when the competitors are removed. We predicted that food-hoarding behavioral responses of A. peninsulae were context-dependent according to different competitors that have different body sizes, food-hoarding habits and/or daily rhythms. If the behavioral plasticity of the subject is caused by competition, then behavioral responses of the experimental animals will recover to the state of pre-competition when competitive pressure is released. Finally, we discussed the implication of behavioral plasticity on survival fitness of subordinate species and the coexistence of sympatric hoarders under asymmetrical competition.

Study sites and animal handling
The experiments were conducted in the Liyuanling field station, Donglingshan Mountain area, approximately 120 km north-west of Beijing City (40°00′N, 115°30′E, 1100 m a.s.l.). This station has been described in detail in our previous works (Zhang & Zhang 2008;Zhang et al. 2013aZhang et al. , 2016. The rodent species used in this study are common around the station (Table 1) (also see Li et al. 2004). All of the experimental animals were captured using live traps (12 × 12 × 25 cm) in 10 plots (2.5 ha, 50-100 m apart) in the secondary forests and shrublands near the field station during the summers of 2015 and 2016 (Table 1). Traps were made of steel wire with an iron sheet attached to the top to protect animals from rain and direct sunlight. Peanuts, local plant seeds (e.g. Q. wutaishanica and A. sibirica), pieces of cucumber or carrot, and local dry leaves were provided in each trap to ensure the captured animals survived well. Twenty-five traps were placed approximately 5-m apart along each of the 2-3 transects (approximately 100-m long, 20-30-m apart) in each plot. Traps were set in an afternoon (1800-1900 hours) and checked twice per day (at 0600-0700 hours and 1800-1900 hours) during the following 2-4 days. Captured animals were individually covered in the trap using a cloth bag and carefully transferred to the laboratory. Females in pregnancy or lactation, juveniles and other unwanted species were released immediately at the site. Following sex determination, weighing and labeling, animals were individually housed in PVC box (37 × 26 × 17 cm) or wheel cages (100 × 100 × 120 cm, one nest box, 20 × 20 × 20 cm, was attached to one corner, for squirrels) and provided commercial mouse chow (Keao Feed, Beijing, China). Water and nest material (wood chips/cotton) were provided ad libitum. Some local rodent-preferred seeds (e.g. Q. wutaishanica, A. sibirica, A. davidiana and/or Juglans regia) and peanuts were provided every week to maintain the animals' natural diets and as a nutritional supplement. Individuals of each species were raised in different rooms with ambient temperature (18-25 °C) and photoperiod (14-16 h of daylight) during the experimental period (August to September in 2015 and 2016). Each animal was acclimatized to the house condition at least 7 d prior to testing. After testing, the experimental animals were released to the sites where they were captured or kept in the laboratory for other experiments. All animals remained healthy until the end of the experiments. All guidelines for animal handling were followed according to Chinese law and the study was permitted by the local government and our institute (permission: SYX-K(é)2015-0052).

Seed preparation
A. sibirica is a common tree/shrub species across northern China. As a dominant species in the secondary forests and shrublands in the study area, A. sibirica produces numerous seeds each summer (July to August). A. sibirica seeds are ideal for experiments because they are highly preferred by the experimental rodent species for consumption and hoarding (Zhang et al. , 2016. The use of A. sibirica seeds is described in our previous works (Huang et al. 2011;Zhang et al. 2011Zhang et al. , 2016. Experimental seeds were randomly collected from more than 50 A. sibirica trees near the field station during the period of natural seedfall (July in 2015 and 2016). All seeds were then mixed to create a single composite sample and kept in the refrigerator (0-4 °C) to maintain their freshness. To facilitate seed-tracking, each experimental seed was marked by a tin-tag: a unique coded tin-tag (30 × 10 mm, 0.1 g weight) was attached to the basal end of the endocarp of each seed using a 3-cm piece of fine steel wire (Zhang & Wang 2001). This method has been widely used to relocate rodent-dispersed seeds under the conditions of enclosures and in the field, although it may delay seed harvest and be used for cues to retrieve/pilfer caches by animals (Xiao et al. 2006).

Enclosures
Four separated semi-nature enclosures (10 × 10 m) used for experiments were constructed in the abandoned farmland at the field station ( Fig. 1). The walls of the enclosures were made of bricks (30-cm thick), extended 30 cm below the ground surface and 100 cm above ground. Wire mesh (1 × 1 cm grid) was used as cover for each enclosure to prevent animals from entering or escaping the enclosure (also see Lu & Zhang 2010;Zhang et al. 2015). Some grass and branches were scattered on the mesh to simulate canopy cover in the field (approximately 60% coverage). The ground surface of each enclosure was a concrete floor which was covered with 15 cm of sandy soil as a hoarding substrate. Grass (e.g. Artemisia spp., Elymus excelsus and Poa spp.) and shrubs (e.g. young A. sibirica and Ulmus laciniata) were planted in the enclosures; the plants were 30-80 cm tall with <60% coverage, similar to the vege-tation around the enclosures. A nest (a wooden box, 20 × 40 × 20 cm, with cotton as nest material) and a water resource (a plastic plate) for the tested animal were placed in one corner of each enclosure, whereas a competitor was located at the opposite corner (Fig. 1). For tests, an individual competitor was held in a steel-wire cage (30 × 30 × 60 cm) with sufficient food, water and cotton nest material, covered with a plank to shelter the animal from rain and direct sunlight. The tested animal could see, smell and even partially touch the competitor animal. A seed station (0.5 m 2 ) was located at the center of each enclosure. Each enclosure was equally divided into 4 areas according to the level of competitive risk: the quarter closest to competitors indicates high competition, the quarter closest to the subject's nest stands for low competition, while the rest of the area of the enclosure represents medium competition (Fig. 1).

Experimental procedures
Each experimental animal received 4 consecutive days of testing in each trial: day 1, habituation (without competitor); day 2, pre-competition (pre-control, without competitor); day 3, competition (treatment, with competitor); and day 4, post-competition (post-control, competitor removed). During each trial an experimental animal was introduced into an enclosure between 15:00 and 16:00 hours, and kept in the enclosure for 4 days. Ten untagged seeds were provided on day 1 for environmental habituation, 30 tagged seeds were provided on day 2 for the pre-competition test, 30 tagged seeds and a competitor were provided (between 1500 and 1600 hours) on day 3 for the competition test, and, finally, the competitor was removed (between 1500 and 1600 hours) and 30 tagged seeds were provided on day 4 for the post-competition test. Water was provided ad libitum. Peanuts (5-10 g) were provided daily during each trial to balance the nutritional needs and starving motivation of experimental animals. In the early afternoon on each day (between 1200 and 1400 hours), animals were closed in the nest, competitors were removed (on day 3) and seed fates were recorded (see below). Enclosures were renewed by adding water and displacing seeds. At the completion of a trial, enclosures were refreshed by removing all seeds and their fragments, the nest box and water plate were replaced and the soil was loosened; a break of 12 h was allowed to limit possible interference. Experimental animals, competitors and experimental times are presented in Table 1. All experimental animals were used only once, but some competitors (3 A. peninsulae, 2 T. triton and 2 S. davidianus) were reused twice with at least a 7 day break.
Seed fates were recorded as: intact in situ (IIS), when a seed was intact and remained at the seed station; eaten (E), when a seed was consumed and the tag was left on the ground surface or in the nest; scatter-hoarded (SH), when a seed was intact and buried in the soil or grass; and larder-hoarded (LH), when a seed is intact and in the nest (also see Zhang et al. 2011). Total harvested seeds is the sum of E, SH and LH. The seed location of each scatter-hoarded seed was recorded as a high, medium or low competitive area (Fig. 1).

Statistical analyses
The Shapiro-Wilk test was used to determine normality, and if P < 0.05, then data were log(x + 1)-transformed to achieve normality. The generalized linear mixed multivariate model was used to test the main effects of competitor species, treatment and their interactions (fixed factors) on seed fates (seed number) and each competitive level of seed placement (proportion) of the experimental animals. Repeated measures ANOVA, including pairwise comparisons, was used to test for effects of experimental treatment on each seed fate and seed location area in different competitor species. Sexual effects of the experimental animals and competitors were not considered because of the small
The number of eaten seeds was significantly affected by competitor species (χ 2 = 36.835, df = 3, P < 0.001). A. peninsulae ate more seeds in the presence of S. davidianus (r = −1.750, P = 0.046), and returned to the former state after competitors were removed (Fig. 2b).
Food-hoarding behavior of A. peninsulae varied with body size, hoarding habit (scatter vs larder), and dai- Data are mean ± SE. *P < 0.05; **P < 0.01 and ***P < 0.001 indicate differences between treatments. ly rhythm (nocturnal vs diurnal) of competitor species (Fig. 3). Overall, intensity of larder-hoarding was promoted, and scatter-hoarding was restrained by intraspecifics, and by S. davidianus, which shared similar hoarding habits (scatter and larder) but had different daily rhythms (diurnal) and body size. Hoarding intensity was totally restrained by N. confucianus and T. triton, which had large body size, larder-hoarding habit and similar daily rhythm (nocturnal). In addition, eating intensity was promoted by S. davidianus.

Hoarding place
The proportion of seeds hoarded in the high competition area was significantly affected by competitor species (χ 2 = 8.352, df = 3, P = 0.039), treatment (χ 2 = 74.566, df = 2, P < 0.001) and their interaction (χ 2 = 17.720, df = 6, P = 0.007). Hoarding intensity of the experimental animals in the high competition area was significantly restrained by all competitor species (All P < Figure 3 Effects of intraspecific and interspecific competitors on seed-hoarding behavior of Apodemus peninsulae under semi-natural enclosure conditions. + is positive effects, − is negative effects and 0 is neutral effects, whereby a factor was promoted (+), suppressed (−) or not changed (0)    Effects of intraspecific and interspecific competitors on seed-hoarding place selection of Apodemus peninsulae under semi-natural enclosure conditions. Cache places are high competition area (a), medium competition area (b), and low competition area (c) in the enclosure. Data are mean ± SE. *P < 0.05, **P < 0.01 and ***P < 0.001 indicate differences between treatments. 0.05), even though the competitors were removed in the treatments of N. confucianus (r = 6.576, P = 0.023) and S. davidianus (r = 3.890, P = 0.035) were competitors (Fig. 4a). Compared to other competitor species, A. peninsulae hoarded more seeds in the high competition area when S. davidianus were competitors.
The proportion of seeds hoarded in the medium competition area was significantly affected by competitor species (χ 2 = 7.975, df = 3, P < 0.047) and treatment (χ 2 = 69.171, df = 2, P < 0.001). The experimental animals reduced intensity of hoarding seeds in the medium competition area in the presence of all competitor species, and then returned to the original state of pre-competition after removal of competitors (all P < 0.05), with an exception of the intraspecific competition treatment (Fig.  4b).
The proportion of seeds hoarded in the low competition area was significantly affected by treatment (χ 2 = 69.171, df = 2, P < 0.001) and the interaction between competitor species and treatment (χ 2 = 22.382, df = 6, P = 0.001). A. peninsulae reduced hoarding intensity in the low competition area when each species of interspecific competitor was present, and then returned to the state of pre-competition after competitors were removed (all P < 0.05) (Fig. 4c).
Overall, seed-hoarding site selection of A. peninsulae was affected by competitor species, treatment and/ or their interaction. Together with the results of larder-hoarding in the nests (Fig. 2d), the results showed that A. peninsulae reduced the intensity of scatter-hoarding outside the nest, and moved more seeds into the nest for larder-hoarding under competition from intraspecific individuals and S. davidianus, whereas they larder-hoarded all seeds in the nests when N. confucianus and T. triton were competitors. The subjects avoided highly competitive areas for scatter-hoarding when they faced competitors of intraspecific individuals and S. davidianus. In most cases, the experimental animals could return to the pre-competition state when competitive pressure was released.

Discussion
We found that hoarding behavioral responses of A. peninsulae were context-dependent according to intraspecific and interspecific competitors that have varied body size, hoarding habit and/or daily rhythm but overlapped habitat utilization and diet preferences. Overall, the experimental animals increased larder-hoarding but decreased scatter-hoarding intensity in the presences of competitors that share similar hoarding habit (scatter and larder) (e.g. intraspecific and S. davidianus), whereas they reduced/ceased hoarding when they faced competitors that have similar daily rhythms (nocturnal), but have large body size and larder-hoarding habits (e.g. N. confucianus and T. triton) (Figs 2 and 3). A. peninsulae tried to avoid highly competitive areas for scatter-hoarding, and larder hoarded more seeds in the nests under the mediation of competitors (Fig. 4). Interactions of food competition were asymmetrical between A. peninsulae and the superior sympatric competitors (Fig. 3). Behavioral plasticity was deemed a functional response of the experimental animals to food competition because they returned to the pre-competition state in most cases when competitive pressure was released (Vander Wall & Jenkins 2003;Dally et al. 2006). Our results are meaningful for understanding the coexistence of sympatric rodents from the point of view of food competition in this asymmetrical system.
Our results suggest that, compared to the focal hoarders, competitors using similar means of sharing resources (e.g. between intraspecific individuals, A. peninsulae vs S. davidianus) have higher competitive pressure to the hoarders even though they have different active rhythms (nocturnal vs diurnal) ( Table 1, Figs 2 and 3). It is easy to understand that intense competition exists between animals that use food items in the same manner Dittel et al. 2017;Gu et al. 2017). For example, intense competition exists in the granivorous rodent communities (e.g. Tamias spp., Peromyscus spp., S. lateralis) in eastern Sierra Nevada, USA that hoard plant seeds (e.g. Pinus jeffreyi, Purshia tridentate, Arctostaphylos patula and Castanopsis sempervirens) in fall and rely on these seeds for survival in winter (Briggs et al. 2009;Vander Wall et al. 2009;Dittel et al. 2017). The experimental animals increased larder-hoarding and reduced scatter-hoarding under the mediation of competitors that share similar manners of food utilization, suggesting that larder-hoarding strategy is more suitable for food protection than scatter-hoarding in A. peninsulae (Fig. 2c,d) (also see Zhang et al. 2013b). This observation is consistent with our previous studies in A. peninsulae when it faced competitors of intraspecific individuals (Zhang et al. , 2014b, or when its stores were completely lost (Huang et al. 2011;Luo et al. 2014), but not when it faced competitors of N. confucianus . In a study conducted in a tropical forest in southwestern China where N. confucianus performs both larder-hoarding and scatter-hoarding, N. confucianus increased larder-hoarding and shifted from scatter-hoarding to larder-hoarding when competitors (Rattus flavipectus) were present (Zhang et al. 2013b). Our observations are different from those of other studies, in which animals increased scatter-hoarding intensity, spaced caches further apart or did not change hoarding strategy when exposed to competitors (Preston & Jacobs 2001;Leaver et al. 2007;Male & Smulders 2007). For example, presence of intraspecific individuals promoted cache spacing in S. carolinensis (Leaver et al. 2007), whereas pilfering of intraspecific individuals impelled D. merriami to shift from scatter-hoarding to larder-hoarding (Preston & Jacobs 2001). Larder-hoarding often makes cache defense easier for superior animals (e.g. large-size species) by preventing thieves, while scatter-hoarding generally avoids total cache loss for subordinate hoarders (e.g. small-size species) by minimizing all caches preferred by thieves at one time (Vander Wall 1990;Macdonald 1997;Preston & Jacobs 2001;Vander Wall & Jenkins 2003;Dally et al. 2006). Larder-hoarding is successful only if the hoarders are superior to their competitors because larder caches often suffer from high pilferage (e.g. high value, stronger odors) and need aggressive defenses (Clarkson et al. 1986;Dally et al. 2006). Scatter-hoarding, by contrast, is successful only when the hoarders can retrieve more caches to compensate for costs in terms of travel, memory and predation risks (Stapanian & Smith 1978, 1984Dally et al. 2006). However, our findings are inconsistent with these predictions because the smallest A. peninsulae increased larder-hoarding, rather than scatter-hoarding, in the presences of competitors (also see Preston & Jacobs 2001). Several factors may determine the experimental animals' responses of larder-hoarding enhancement. First, A. peninsulae often hoard seeds in burrows and small caves in the field, which prevents large thieves from gaining access to these seeds due to the small size of the entrance (Table 1) (Zhang et al. 2014b). Second, scatter-hoarding does work in the field, but it does not work under enclosure conditions because risks of competition/pilferage and predation are very high, and/or suitable sites for scatter-hoarding are limited in the narrow and enclosed space. Our other studies conducted in a large and more complex enclosure (40 m × 50 m, which contain enough suitable cache sites and refuges) indirectly support these arguments (Zhang et al. 2014a;Wang et al. 2018;Wang 2018;Huang et al. 2019). In these studies, S. davidianus (large body and primary scatter hoarders, Table 1) rapidly hoarded seeds around seed stations and then transferred these seeds to nest areas for scatter-hoarding, rather than move these seeds into the nest for larder-hoarding (Zhang et al. 2014a), whereas A. peninsulae adopted both scatterand larder-hoarding throughout the period of tests under the mediation of intra-and/or inter-specific competitors (Wang 2018;Huang et al. 2019).
Our results also show that food competition of subordinate species (small-size) was totally suppressed by superior competitors (large-size) that have overlapped habitat preference, nocturnal rhythm and diets (e.g. N. confucianus and T. triton) ( Table 1, Figs 2 and 3). Inconsistent with this observation, our previous studies conducted under the similar enclosure conditions show that A. peninsulae individuals harvested, larder-hoarded and consumed food with great intensity when N. confucianus were in audience (caged, cannot compete/pilfer directly, as same as this study), and ate more of their larder-hoarded seeds and tried to transfer their scatter-hoarded caches into the nests for larder-hoarding when N. confucianus were pilferers (not caged, can freely pilfer) (Zhang et al. 2014b). This difference reflects the plasticity of hoarding behavior in A. peninsulae, a largely unknown phenomenon that animals can regulate hoarding strategies according to conditions of time/space/environment and risks of predation/competition/pilferage. In addition, our results suggest that there were definitely asymmetrical interactions in food competition between sympatric species that have similar ecological niches, supported by several previous studies that were conducted under enclosure conditions and in the field (e.g. Penner & Devenport 2011;Zhang et al. 2014b;Dittel et al. 2017;Gu et al. 2017). For example, N. confucianus (large-size) harvested more seeds from the seed station than A. peninsulae (small-size), and showed a unidirectional pilferage of seeds cached by A. peninsulae (Zhang et al. 2014b). In the pine forests (e.g. P. jeffreyi) of southern Reno, Nevada USA, S. lateralis (primary larder hoarders) were more able to forage for nuts on trees or on the ground surface, while T. amoenus (primary scatter hoarders) were more professional in cache-pilfering ). Within the asymmetrical interactions, whether or not one makes a profit is associated with its body-size, hoarding habit and competitive ability for food resources and pilferage of caches Dittel et al. 2017;Wang et al. 2018). It is interesting how a subordinate species coexists with superior competitors that shared similar ecological niches. Here, by considering interactions of seed-hoarding, we argued that context-dependent regulation of hoarding strategies according to competitors promotes subordinate species coexistence with their superior neighbors. However, this argument needs more observations in multiple species and ecosystems.
Several animals reduce foraging intensity or cease caching events to avoid further loss by competitors/pilferers (Bugnyar & Kotrschal 2002;Emery et al. 2004;Dally et al. 2006;Baudoin et al. 2013). Here, we argued that high competitive pressure compelled A. peninsulae to reduce foraging or cease caching under enclosure conditions because the subjects recovered to the original state of pre-competition in most cases when competitors were removed (Fig. 2). The observation that experimental animals avoided the high competitive areas for scatter-hoarding and moved more seeds into the nest for larder-hoarding implies that they could use alternative strategies, rather than cease foraging, to compete for resources with superior species if there were enough spaces (Fig. 4). Competition between sympatric species is less intense in the field than under enclosure conditions because there are light differentiations at micro-levels of habitat selection, rhythm, diets and hoarding behavior. At our study site, A. peninsulae (small-size, larder and scatter hoarder) is different from other species in micro-habitat selection, active peak time, seed preferences and/or hoarding behavior (Zhang & Zhang 2008;Zhang et al. 2011Zhang et al. , 2014bHuang et al. 2019). These differentiations can help A. peninsulae to avoid high competition and pilferage caused by superior species, and therefore promote species coexistence. Ecological niche difference at multiple levels from other species is also a contribution to prevalence of A. peninsulae in the study area. Generally, A. peninsulae reduced/ceased foraging and hoarding, increased larder-hoarding, reduced scatter-hoarding, and/or avoided the high competitive area for hoarding according to the presence of different competitor species. Exactly which strategies were adapted should be associated with the body-size, habitat preference, diet, hoarding habit and/or daily rhythm of competitors. Together with our previous studies (e.g. Zhang et al. 2011Zhang et al. , 2014bLuo et al. 2014;Huang et al. 2019), we argued that context-dependent behavioral regulation according to competitors is beneficial for species coexistence and survival of the subordinate species in asymmetrical system. In order to demonstrate how widespread this phenomenon of behavioral plasticity occurs in this asymmetrical system, further studies are needed in multiple communities and ecosystems .  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51