Causes and Ecological Context of Dietary Specialization in Neotropical Army Ants

by John Aidan Manubay

B.S. in Biology, May 2013, University of Maryland College Park

A Dissertation submitted to

The Faculty of The Columbian College of Arts and Sciences of The George Washington University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

May 19, 2019

Dissertation directed by

Scott Powell Associate Professor of Biological Sciences

The Columbian College of Arts and Sciences of The George Washington University certifies that John Aidan Manubayhas passed the Final Examination for the degree of

Doctor of Philosophy as of April 12, 2019. This is the final and approved form of the dissertation.

Causes and Ecological Context of Dietary Specialization in Neotropical Army Ants

John Aidan Manubay

Dissertation Research Committee:

Scott Powell, Associate Professor of Biological Sciences, Dissertation Director

John Lill, Professor of Biological Sciences, Committee Member

Adam Smith, Assistant Professor of Biological Sciences, Committee Member

ii

Abstract of Dissertation

Causes and Ecological Context of Dietary Specialization in Neotropical Army Ants

Understanding of the mechanisms that promote coexistence of species-rich predator assemblages in the Neotropics is critical for explaining the structure of diverse tropical communities. New world army ants are voracious predators of ants and other arthropods throughout the Neotropics, and up to twenty species may co-occur in one location. They therefore offer a valuable opportunity to address how predators may partition prey resources and coexist. While we know that each army ant species specializes on different ants and other arthropods, little is known about the mechanisms by which they locate and identify their preferred prey. In studying the closely related army ants in the genus Eciton, we asked the following interconnected questions: 1) Can army ants detect and distinguish between odors of prey and non-prey? 2) Do army ants prefer specific sources of prey-derived odors? 3) Do army ants use vibrational signals from potential prey irrespective of identity? 4) What is the dietary overlap among co- occurring Eciton species in Panamanian tropical forests? 5) Are Eciton burchellii army ants able to detect and attack preferred prey of their congeners?

We filmed and measured army ant behavioral responses to several categories of prey-derived signals to determine the degree to which army ants use odors and vibrations to detect food and direct foraging. We found army ants are specialized on odors derived from preferred ant prey and ignore odors from non-prey ant species. Army ants also had varied responses to different sources of prey odors; prey nest material was most attractive, odors of live and dead prey were next most attractive, while alarm odors were not of any interest. Additionally, we found that army ant species responded differently to

iii vibrations made by struggling arthropods, consistent with their divergent foraging strategies and diet breadths. To confirm this finding, we censused prey intake of the co- occurring Eciton in the Barro Colorado Natural Monument in Panama and found negligible dietary overlap at the prey genus and species level across the feeding guild.

Finally, we found Eciton burchellii to be unable to detect odors of preferred prey of its congeners; however, vibrational stimuli allowed for a modest army ant response to potential prey irrespective of species identity. Taken together, these studies reveal that sensory specialization to prey-derived odor stimuli provides a mechanism by which army ants partition food resources, potentially promoting coexistence.

iv

Table of Contents

Abstract of Dissertation...... iii

List of Figures...... vi

List of Tables...... vii

Chapter 1: Detection of prey odors underpins dietary specialization in a Neotropical top-predator...... 1

Chapter 2: Contrasting sensory modality responsiveness explains differences in diet breadth of co-occurring predators...... 24

Chapter 3: Olfaction mediates food niche breadth in army ants...... 48

References...... 77

v

List of Figures

Figure 1...... 20

Figure 2...... 21

Figure 3...... 22

Figure 4...... 23

Figure 5...... 45

Figure 6...... 46

Figure 7...... 47

Figure 8...... 73

Figure 9...... 74

Figure 10...... 75

Figure 11...... 76

vi

List of Tables

Table 1...... 71

Table 2...... 72

vii

Chapter 1: Detection of prey odors underpins dietary specialization in a Neotropical top- predator

Abstract

Deciphering the mechanisms that underpin dietary specialization and niche partitioning is crucial to understanding the maintenance of biodiversity. New world army ants live in species-rich assemblages throughout the Neotropics and are voracious predators of other arthropods. They are therefore an important and potentially informative group for addressing how diverse predator assemblages partition available prey resources. New World army ants are largely specialist predators of other ants and partition their prey niches by specializing on different ant genera. However, the mechanisms of prey choice are unknown. In this study, we addressed whether the army ant Eciton hamatum: 1) can detect potential prey odors, 2) can distinguish between odors of prey and non-prey, and 3) respond differently to specific types of odors associated with its prey. We tested experimentally the response of army ants to four odor treatments, each with a unique combination of odor sources and other stimuli (alarm odors, dead ant, live ant, and nest material), associated with both prey and non-prey ants. These data were used to determine the degree to which E. hamatum are using specific prey odors and movement cues to detect potential prey and direct their foraging. Army ants responded strongly to odors derived from prey ants, which triggered both increased localized recruitment and slowed advancement of the raid as they targeted the odor source, while odors from non-prey ants were largely ignored. Additionally, the army ants had the strongest response to the nest-material of their preferred prey, with progressively weaker responses across the live ant, dead ant, and alarm odors treatments, respectively. This

1 study reveals that the detection of prey odors, and especially the most persistent odors related to the prey’s nest, provides a mechanism for dietary specialization in army ants. If ubiquitous across the Neotropical army ants, then this olfaction-based ecological specialization may facilitate patterns of resource partitioning and coexistence in these diverse predator communities.

Introduction

Ecological specialization can be a powerful mechanism for reducing the likelihood of competitive exclusion, and thus promoting long-term species coexistence

(Hutchinson, 1959; Connell & Orias, 1964; reviewed in Chase & Leibold, 2003; Sapp,

2016). The competitive exclusion principle states that if two species have complete niche overlap then the species with more efficient use of a shared resource will drive the second species to extinction. However, specialization, which can be conceptualized as any reduction in the breadth of resource usage within an n-dimensional niche space

(Hutchinson, 1957; Devictor et al., 2010), can limit overlap, reduce competitive interactions between species, and promote coexistence (Hardin, 1960; Chase & Leibold,

2003). Despite this long-standing insight, how specialization is mechanistically established remains poorly understood for many taxa. This gap is especially pronounced in hyper-diverse tropical communities, wherein most species must contend with the potential for high intraguild competition and a highly complex and species-rich environment in which to locate resources (MacArthur, 1972; Sapp, 2016).

Specialization on different resource species has been shown to be especially important in reducing niche overlap and competition across a broad diversity of

2 organisms (Futuyma & Moreno, 1988; Chase & Leibold, 2003), including such disparate taxa as Galapagos finches (Grant, 1999), warblers (MacArthur, 1958), and Anolis lizards

(Pacala & Roughgarden, 1985). In particular, ecological specialization on different food resources has long been seen as a primary mechanism promoting coexistence in diverse insect communities, such as insect herbivores (Singer & Stireman, 2005) and pollinators

(Michener, 2007; Armbruster, 2017). In such cases, specialization is often underpinned by sensory specialization (Stevens, 2013), including the combination of multiple senses to maximize foraging efficiency (Siemers et al., 2007; Vincent, Shine, & Brown, 2005).

Nevertheless, this mechanistic knowledge for specialization is lacking in diverse assemblages of predatory insects (but see Rana, Dixon, & Jarošík, 2002; Xue, Wei, &

Huang, 2018 for species specific examples), which despite their trophic footprint are rarely the focus of studies addressing dietary specialization (but see Kaspari, Powell,

Lattke, & O’Donnell, 2011; Hashimoto & Yamane 2014).

New World army ants are well-established as ecologically important predators in

Neotropical systems (Gotwald, 1995), and all existing data indicates that each army ant species is dietarily specialized on one to a few specific genera of other ants (Rettenmeyer,

Chadab-Crepet, Naumann, & Morales, 1983; LaPolla, Mueller, Seid, & Cover, 2002;

Powell & Clark 2004; Powell & Franks 2006; Breton, Dejean, Snelling, & Orivel, 2007;

Powell, 2011). Yet how this specialization is mechanistically achieved is not known. As a group, the New World army ant (a monophyletic group of five genera within the subfamily Dorylinae: Brady, Fisher, Schultz, &Ward, 2014; Borowiec, 2016) therefore represent an ideal taxon for advancing our knowledge of dietary specialization in assemblages of predatory insects. This is especially true given the high regional and local

3 diversity of this group, which only adds to the intraguild competitive context and environmental complexity that these specialist predators operate under. There are 151 described species (Bolton, 2019) of Neotropical army ant, and up to 20 species can co- occur in most tropical forest throughout Central and South America (Rettenmeyer, 1963;

Schneirla, 1971; Rettenmeyer, Chadab-Crepet, Naumann, & Morales, 1983; Kaspari,

Powell, Lattke, & O’Donnell, 2011). Moreover, all species are obligately nomadic group- predators that roam the forests they inhabit in large collective raids, simultaneously seeking out, attacking, and harvesting their preferred ant prey (Rettenmeyer, 1963;

Schneirla, 1933; Schneirla, 1971). Every location within the forest is then intensively raided by all resident army ant species through time (Franks & Bossert, 1983; O'Donnell,

Lattke, Powell, & Kaspari, 2007; Kaspari, Powell, Lattke, & O’Donnell, 2011), with each species harvesting different ant prey from the same area. This biology makes the established dietary specialization even more remarkable and important to explain mechanistically.

Army ants may use a number of non-mutually exclusive mechanisms to locate their specific prey, but none have been tested to date. First, army ants may be detecting specific odor cues, allowing them to localize their prey as they forage through the forest

(Gotwald, 1995). Despite research into the chemical basis of army ant intra-colony communication and chemical signaling used in the prodigious recruitment of army ants

(Chadab & Rettenmeyer, 1975; Chadab, 1979), we have only a cursory understanding of how army ants use chemicals to perceive their environment, and especially their prey.

There are many sources of species-specific odor cues which army ants could be detecting, including cuticular hydrocarbons, volatile pheromones used in alarm and recruitment, and

4 colony odor on nest material (D’Ettorre & Lenoir, 2010). Detection of prey odors may then allow detection of prey across a range of spatial scales and via different odor cues.

Second, prey movement, and associated visual or vibrational cues, represents another plausible mechanism of prey detection and localization. Army ant vision is known to be exceptionally poor, with eyes absent or reduced to a single facet in most species (Bulova,

Purce, Khodak, Sulger, & O’Donnell, 2016). However, visual motion detection may still be possible, and movement may be detected indirectly via vibrations through the substrate (Hill, 2008; Hill, 2009). Nevertheless, it is unclear how visual or vibrational sensory modalities would yield species-specific information to the army ants beyond direct visual identification of the prey. Finally, army ants may eschew directed foraging entirely and simply be engaging in a random walk that lead to direct contact with prey ants at their nests, resulting in subsequent attack.

Our study tests the potential role of different prey-odor cues in detection of preferred prey in army ants, specifically Eciton hamatum in the moist tropical forest of central Panama. This army ant species conducts conspicuous group raids over the forest floor and has a well-documented, quantified dietary specialization on Acromyrmex leaf- cutting ants in the study site (Fig. 1, Powell & Franks, 2006; Powell, 2011). Combined, this foraging ecology and dietary specialization facilitate highly tractable experimental manipulation and control of different prey-odor cues. More specifically, we use this study system to address the following critical questions about specialized detection of prey odors: 1) Does E. hamatum have the ability to detect prey odors in their environment? 2)

Can E. hamatum differentiate between odors from different ant species? 3) Does E. hamatum exhibit differentiated predatory responses to certain types of odor cues

5 produced by its prey? Broadly, this research addresses how an individual species in a diverse guild of predators detects its specific prey in the complex environment of the tropical rainforest, shedding light on how predator assemblages are mechanistically partitioning available prey.

Material & Methods

Study site, study organisms, and colony discovery

Fieldwork was conducted on Barro Colorado Island (BCI), Panama (9°090N,

79°500W) between May and September 2015. BCI is a seasonally dry, 15.6 km2 island rainforest in the central Lake Gatun of the Panama Canal (see Leigh, 1999 and Ziegler &

Leigh, 2011 for detailed information). Colonies of Eciton hamatum army ants were located by walking heavily compacted trails during daylight hours, until raid columns were encountered (following methods of e.g. Franks, 1980; Franks, 1982; Vidal-Riggs &

Chaves-Campos, 2008; O'Donnell, Lattke, Powell, & Kaspari, 2007; Powell 2011). Raid traffic was then tracked in the direction of prey flow, to locate the nest or ‘bivouac’.

Colonies were tracked nightly, by following the emigration traffic to the new bivouac site. This ensured the location of the colony was known for data collection the following day (Powell & Franks, 2006; Powell, 2011).

The ant species Acromyrmex octospinosus was used as the preferred prey of E. hamatum on BCI, as determined by relative frequency of prey items and biomass intake by the colony (Rettenmeyer, Chadab-Crepet, Naumann, & Morales, 1983; Powell &

Franks, 2006; Powell, 2011). The non-prey ant used in this study was Cephalotes atratus, which has never been recorded in the diet records for any army ant species. Easily

6 accessible colonies of both prey A. octospinosus and non-prey C. atratus ants were located and marked while searching for army ant columns.

Odor treatments and preparations

Four distinct odor treatments, each with a unique combination of odor sources and access to the odors for the army ants, were prepared for both prey and non-prey ants and presented to E. hamatum raids. The four odor treatments were as follows: 1) Volatile alarm odors from a living ant that the army ants could not contact directly; 2) A dead ant that was freshly killed, to present the cuticular hydrocarbon odors of the ant to the army ant raid without volatile alarm pheromones or movement cues; 3) A living ant that could be contacted directly by the army ants, to provide the combined presence of cuticular hydrocarbons, alarm pheromones, and movement/vibrational cues; 4) Nest material, to provide high concentrations of gestalt colony odor (Lenoir, d’Ettorre, Errard, & Hefetz,

2001; d’Ettorre & Lenoir, 2010) to the army ants without the presence of any adult ants.

For the alarm odor treatment, live ants were collected and placed in thin wire mesh boxes to generate localized alarm pheromones during each trial without allowing direct contact. Alarm odors were easily detectable to the human nose for both prey and non-prey ant species upon agitating the living ants in the wire mesh box. For the dead ant odor treatment, live ants were rapidly euthanized in a -20ºC freezer and then used in experiments within one hour of removal from the freezer. This procedure provided the cuticular hydrocarbon odors required for the experimental treatment, while ensuring that volatile odors were not present and that the ant had not been dead long enough to be producing “rotting ant” odor cues (Choe & Rust, 2008; Howard & Tschinkel, 1976;

Wilson, Durlach, & Roth, 1958). For the live-ant treatment, the ant was partially

7 immobilized by a non-lethal procedure ablating just above the joint between the tibia and femur of each leg. This preparation provided the army ants with direct access to the ant’s cuticular hydrocarbon and volatile alarm pheromones while the ant also produced movement and vibrational cues without being able to escape. For the final nest odor treatment, fresh nest-material was collected and placed in sealed containers until trials began, with no more than six hours between collection and use.

Data collection

A total of seven E. hamatum army ant colonies were tested between June and

August of 2015. Alarm trials were conducted 140 times across the seven colonies, dead ant trials were conducted 139 times across seven colonies, live ant trials were conducted

107 times across five colonies, and the nest material trials were conducted 156 times across seven colonies. Each treatment trial was recorded with a 4K (3840 x 2160 resolution) GoPro video camera, to allow subsequent data extraction. The camera was setup in the ultra-wide-angle lens mode and held or mounted at an appropriate distance to capture an overhead view of the interaction between the E. hamatum raid front and the odor treatment. More than 300 minutes of video was recorded across the total of 542 treatment trials.

After identifying an E. hamatum colony and associated raiding columns, a randomly selected odor treatment, from either the prey or non-prey ant species, was presented to a raid front at a distance of approximately 20 centimeters. Naïve raid fronts within the overall branching structure of the raid were selected haphazardly for each trial until odor material was depleted. Each trial was replicated a minimum of 10 times per

8 odor treatment, per E. hamatum colony for both prey and non-prey ants, and with as many E. hamatum colonies as could be found.

Data extraction

The response of the army ants to the odor treatments was captured by extracting three metrics from video of each trial; 1) army ant recruitment rates measured in ants per second; 2) army ant raid advancement speed measured in centimeters per second; 3) average running speed of three individual army ants within the raid, measured in centimeters per second. These metrics were chosen because shifts in ant recruitment and speed are reliable proxies for army ant attacks against their prey. The raid front has a steady advancing speed with the ants evenly spread over the forest floor as they collectively search for prey. Then once prey is detected, individuals are rapidly recruited to the site of the prey within the raid front (Chadab & Rettenmeyer, 1975), and the speed of both the overall raid front and individuals within it drops as the ants switch into localized search and attack behavior to capture the prey. Stronger responses to prey are thus represented by higher recruitment rates and slower raid and individual speeds.

To calculate our three metrics of army ant response, a series of still images were captured from each video and analyzed in ImageJ version 1.49 (Schneider, Rasband &

Eliceiri, 2012). The interval between images was 0.5 seconds. Frame dimensions of each individual video were measured using the calibration function of ImageJ. E. hamatum recruitment rate to odor treatments was measured by counting the number of ants that stopped walking and began antennating after the first ant discovered the odor source (see

Chadab & Rettenmeyer, 1975). Raid advancement speed was measured by averaging the time spent in the video frame for ten randomly selected army ant individuals in the

9 presence of a given odor treatment. Only ants entering the frame within the last 20 seconds of each trial were used to account for minor variation in trial length. Raid advancement speed was not assessed for the alarm treatment, because the physical structure of the mesh box altered the collective movement of the raid and therefore the accurate determination of the of this metric. Average individual army ant speed was calculated from instantaneous velocities. The series of captured images was fed into the

Manual Tracker plugin of ImageJ, where three randomly selected army ants had their instantaneous velocities measured between two frames for the duration of each video trial, to allow the calculation of the individual average speeds.

Data analyses

We conducted all analyses for this study with R version 3.4.4 for Windows (R

Core Team, 2016). Recruitment data were first simplified into binomial yes/no data which was then analyzed to test for equality of proportion between successful and failed recruitment across odor treatments and between prey and non-prey, using Chi-square tests. Differences between mean running speeds (± standard deviation) irrespective of colony identity were assessed with two-sample T-tests assuming unequal variances. Raw recruitment rate, raid speed, and individual running speed data were found to be positively skewed, and were transformed appropriately to conform to the assumptions of normality and equality of variance to conduct parametric statistical analyses. We used mixed design analyses of variance (ANOVA) to test for differences across prey-odor treatments (fixed effect), while accounting for the potential non-independence of the random colony factor. Post hoc analyses for differences among prey-odor treatments were conducted using the Tukey’s Honest Significant Differences test.

10

Results

Responses to prey ant vs. non-prey ant

E. hamatum recruitment was significantly stronger to odor cues derived from prey ants (Acromyrmex) than non-prey ants (Cephalotes) across all four odor treatments

(Alarm, χ2 (1) =54.4, P<0.001; Dead, χ2 (1) =66.3, P<0.001; Live, χ2 (1) =31.1,

P<0.001; Nest, χ2 (1) =62.5, P<0.001). In the Live treatment, E. hamatum had a mean individual running speed of 5.46 ± 1.62 cm/s in response to non-prey odors, which was significantly faster than the mean of 2.47 ± 1.19 cm/s in response to prey odors (t (273,

300) = 18.17, P<0.001). In the Dead treatment, E. hamatum had a mean individual running speed of 6.98 ± 1.57 cm/s in response to non-prey odors, which was significantly faster than the mean of 3.25 ± 1.06 cm/s in response to prey odors (t (368, 420) = 28.38,

P<0.001). Given that all non-prey odors did not cause the army ants to slow down to search the area as prey odors did, and the majority of non-prey odor trials did not elicit successful recruitment, and thus had many zero values, non-prey odors were dropped from further analyses. This allowed for more powerful parametric tests of the differences in army any responses among prey-ant odor treatments in which the army ants showed interest.

Recruitment rates to different prey-odor treatments

Army ant recruitment rates were significantly different across prey-odor treatments (mixed design ANOVA; F (3, 289 = 91.91), and P<0.001; Fig. 2) with no effect of the random colony factor (χ2 (1) = 0.01, and P=0.94). Recruitment was strongest to the nest material, followed by live and dead ant odor treatments respectively, with the

11 alarm-odor treatment eliciting the weakest recruitment (Post hoc Tukey’s HSD, all pairwise comparisons, P<0.001; Fig. 2)

Raid speed in response to different prey-odor treatments

Army ant raid advancement speeds were significantly different across the three odor treatments that allowed the accurate assessment of this metric (mixed design

ANOVA; F (2, 229) = 216.30, and P<0.001; Fig. 3) with no effect of the random colony factor (χ2 (1) = 0.59, and P = 0.44). Army ant raid speeds were slowest for army ants presented with the nest-material treatment, followed by the live odor treatment, while raid speeds were fastest for dead-ant odors (Post hoc Tukey’s HSD, all pairwise comparisons,

P<0.001; Fig. 3).

Individual running speed in response to different prey odors

Individual average running speed was significantly different across all prey odor types (mixed design ANOVA: F (3, 780) = 634.04, and P<0.001; Fig.4), with no effect of the random colony factor (χ2 (1) = 0.63, and P = 0.43). Of the 1561 army ants measured, individuals slowed down the most when presented with the nest material treatment, followed by live and dead ant treatments, and individual army ants slowed down least in the presence of alarm odor treatment (Post hoc Tukey’s HSD all pairwise comparisons,

P<0.001; Fig. 3).

Discussion

Ecological theory posits that tropical organisms are more narrowly specialized, thus allowing a greater number of species to coexist in a given area (Dobzhansky, 1950;

12

MacArthur, 1969; Schemske, Mittelbach, Cornell, Sobel, & Roy, 2009). However, the precise mechanisms of how, and the contexts in which, tropical predators are specialized remain poorly understood for the majority of taxa. In this study we examined the role of prey-derived odor cues in the detection of prey by army ants. We found that army ants can distinguish among potential prey odors in a natural context, and differentially respond to several different categories of odors derived from their preferred prey ants. Of the categories of odors tested, army ants exhibited the strongest response, demonstrated across three metrics, to nest material odors. Responses were weaker to live-prey odors, dead-prey odors, and prey alarm odors, respectively. Our results indicate that patterns of dietary specialization in Eciton hamatum can largely be explained by their reliance on olfaction to detect and discriminate among potential prey. This selectivity in olfactory attention may be a major factor contributing to the continued coexistence of many tropical arthropod predators.

Army ants distinguish between potential prey odor

We observed dramatic army ant recruitment to prey (Acromyrmex) odors relative to non-prey (Cephalotes) odors across all treatments. This provides compelling support for the hypothesis that army ant dietary specialization is underpinned by their ability to detect and discriminate among ant-prey odors. From one of the few quantitative studies on specialized army ant diets, we know an Eciton hamatum raid can reach peak intake rates of 133 Acromyrmex prey items/min, with an average total daily intake of nearly

5000 Acromyrmex prey items (Powell, 2011). Given the apparently patchy distribution and cryptic nature of the nest cavities used by Acromyrmex octospinosus (Fowler,

Pereira-da-Silva, Forti, & Saes, 1986), prey odors from nest material and cuticular

13 hydrocarbon profiles appear to represent reliable cues that E. hamatum can use to detect and attack its preferred prey. Broadly, our results then provide compelling evidence that army ant foraging is directed at local scales by the detection of and recruitment to persistent and reliable odor cues of their preferred prey.

The army ants used in this study exhibited little interest in non-prey ant

(Cephalotes) odors, as evidenced by their general lack of recruitment and fast forager speeds in response to all non-prey odor material. The non-prey odor cues used here do not deter or repel army ants in any significant way, and appear to be ignored almost entirely. Consistent with this observation is the fact that E. hamatum also show absolutely no interest Atta leaf-cutting ants, the sister genus to the preferred and actively tracked

Acromyrmex prey documented here, demonstrating a remarkable ability to differentiated odors of even closely related taxa (Powell & Clark, 2004; Powell, 2011). While chemically based aversion is not published in army ants, years of field observations have yielded insight into specific taxa army ants avoid; for example, most army ants in the

Neotropics have been observed to give wide berth to Azteca nests, and have been observed to even change raid trail direction if presented with Azteca nest carton material

(personal observations by both authors). This is consistent with the observation that many species of social wasps, the preferred non-ant prey item of E. hamatum, often nest in trees with active Azteca colonies, thus receiving indirect defense from the army ants (Chadab-

Crepet & Rettenmeyer, 1982).

Variation in responses to prey odors as a mechanism of dietary specialization

Taken together, the responses measured in this study (Figs. 2-4) indicate that army ants respond more to prey odors that persist in the environment and likely indicate

14 proximity to prey nests, while showing little interest in volatile alarm pheromones of adult ants. Logically, it seems most beneficial for army ants to respond the most to the presence of the nest-material odors of prey, as army ants preferentially harvest brood from prey nests (Fig. 1; Rettenmeyer, 1963; Schneirla, 1971). The relatively high recruitment rates to living and dead ants (Fig. 2) may also be explained by the use of odor cues that indicate proximity to the nest: foragers or refuse piles containing dead ants may also reliably signal proximity to a prey nest. Evidence from the chemical mimicry literature suggests that some myrmecophilous beetles, spiders, etc. rub themselves against nest material inside their respective nests, to gain the specific cuticular hydrocarbon profile, or gestalt odor, of the ants they live with (Lenoir, d’Ettorre, Errard, & Hefetz,

2001). Given that nest material across ant species contains cuticular hydrocarbons

(CHCs) specific to the resident ants (d’Ettorre & Lenoir, 2010), it is therefore fitting that the army ants have the strongest response to this odor source. The experimental results of this study thus provide compelling evidence that army ants are making use of odor cues that signal proximity to ant nests and brood, and thus, the greatest return on foraging investment.

While this study suggests that accumulated CHCs may be the persistent odor source that army ants are responding to so strongly in nest material, we know relatively little about interspecific ant-ant olfactory detection. All ants have a suite of olfactory receptor neurons housed in specialized olfactory sensilla on antenna, representing the primary non-contact detectors of pheromones that convert chemical signals into neuronal activity and convey that information to glomeruli in the antennal lobe (Kelber, Rössler, &

Kleineidam, 2010; Mysore, Shyamala, Rodrigues, 2010; Nakanishi, Nishino, Watanabe,

15

Yokohari, & Nishikawa, 2009). While recent research shows that two sister species of

Camponotus ant antennae have broad spectrum sensitivity to CHCs of nestmates and non-nestmates within their respective species (Sharma et al., 2015), this provides only hints of the capacity for detecting species-specific CHCs in interspecific ant interactions.

Nevertheless, here we have shown that army ants display highly differentiated responses to prey and non-prey ants across a range of contexts where CHCs are likely to be the dominant species-specific olfactory signature that is encountered. Moreover, army ants seem to have very little interest in volatile alarm pheromones emitted by either non-prey or preferred prey ants. Our data therefore suggests that army ant antennae can detect some CHCs of their preferred prey, at least at local spatial scales. Dedicated electrophysiological assays (e.g. D'Ettorre, Heinze, Schulz, Francke, & Ayasse 2004;

Holman, Jørgensen, Nielsen, & D'Ettorre, 2010; and Sharma et al. 2015) will be necessary to confirm the chemical identities of the odor cues that army ant antennae are detecting.

Broadly, this work has demonstrated the importance of olfaction in mechanistically underpinning the dietary specialization of army ants on particular ant prey, but is it unclear how widespread this mechanism of prey detection might be in the ants. The importance of olfaction in food acquisition has been studied in numerous taxa

(reviewed in Stevens 2013), including in specialist arthropod predator-prey interactions, such as in ladybird predation on particular aphids (Al Abassi, Birkett, Pettersson, Pickett,

Wadhams, & Woodcock, 2000), or spiders on specific ant species (Cárdenas, Jiroš, &

Pekár, 2012). The parasitoid Hymenoptera have received particular attention in the chemosensory literature as they are often specialized on one or a few herbivorous insect

16 species, and seek a narrow range of herbivore-induced volatiles emitted by the prey’s host plant (McCormick, Unsicker, & Gershenzon, 2012). In social insects however, mechanisms of olfactory detection have primarily been assessed in the context of intra- specific communication (reviewed in Van Zweden & D’Ettorre, 2010; Leonhardt,

Menzel, Nehring, & Schmitt, 2016). While this is the first demonstration that any army ant uses olfactory cues to detect other ants as prey, previous studies have reported similar olfaction-based prey detection mechanisms in Pachycondyla ants preying on termites

(Yusuf, Crewe, & Pirk, 2014; Yusuf, Gordon, Crewe, & Pirk, 2014) and Crematogaster ants preying on fig wasps (Schatz, Anstett, Out, & Hossaert-McKey, 2003; Schatz &

Hossaert-McKey, 2010). The Pachycondyla-termite interaction offers an especially interesting comparison to the E. hamatum-Acromyrmex interaction for two reasons; both represent instances of food-driven combat between large, eusocial insect colonies, and both predator species appear to be most interested in odors associated with the prey nest- material, which are more stimulating than odors from the prey themselves (Yusuf,

Gordon, Crewe, & Pirk, 2014).

Potential consequences for coexistence:

Mechanisms that allow niche partitioning at fine scales may play an important role in the maintenance of species diversity in the Neotropics (Dobzhansky, 1950;

MacArthur, 1969; Schemske, Mittelbach, Cornell, Sobel, & Roy, 2009; Sapp, 2016). The ability to detect species-specific prey odors, as we have demonstrated here for E. hamatum, provides a new potential mechanism by which army ants could be partitioning ant prey resources within their diverse assemblages. We know that sensory adaptations can increase efficiency in cue reception and transmission, while simultaneously resulting

17 in sensory bias on perception in the environment (Endler, 1992; Fuller, Houle, & Travis,

2005; Stevens 2013). This sensory bias can then play a determinative role in species diet by altering sensory access to food (Barclay & Brigham, 1991; Bernays & Wcislo, 1994;

Raine & Chittka, 2007), resulting in a mechanism to reduce interspecific competition and facilitating coexistence. Sensory-based variation in prey choice among members of a feeding guild is an understudied mechanism of niche partitioning that requires detailed field observations and studies of behavior (but see Siemers & Scnitzler, 2004). This information is sorely lacking for the majority of army ants, which are largely subterranean and nocturnal and typically live in assemblages of up to twenty co-occurring species in Neotropical forests (Rettenmeyer, 1963; Schneirla, 1971; Rettenmeyer,

Chadab-Crepet, Naumann, & Morales, 1983; Kaspari, Powell, Lattke, & O’Donnell,

2011). Nevertheless, it is reasonable to predict that the dietary specialization on specific ant genera seen across the New World army ants is underpinned by the same sensory specialization we have shown here for E. hamatum, providing a sensory-based mechanism of niche partitioning at the level of the whole assemblage.

An ideal focus for further studies on the sensory-based mechanisms of dietary specialization and niche partitioning in army ants would be to contrast our results here with other members of the genus Eciton. Within the New World army ants, members of this genus are among the largest bodied ants and the least subterranean in their foraging activities. Eciton burchellii is perhaps the most tractable candidate species for an initial comparison because it is diurnal and forages entirely above ground, like E. hamatum.

Moreover, E. burchellii is the most closely related co-occurring species to E. hamatum in central American tropical rainforests (Winston, Kronauer, & Moreau, 2017), and it has a

18 distinct mode of swarm foraging that allows it to incorporate non-ant arthropods into its diet, in addition to the typical ant specialization (Rettenmeyer, 1963; Rettenmeyer,

Chadab-Crepet, Naumann, & Morales, 1983; Powell & Franks, 2006). The known contrasts in diets and the potential contrasts in sensory modes of prey detection may therefore be particularly informative with respect to sensory-based mechanisms of species coexistence. Further studies concerning the precise identities of prey ant cuticular hydrocarbons and the electrophysiological basis of odor detection in army ants would also provide a complementary perspective on this issue.

Conclusion

A deeper understanding of the specific mechanisms that maintain diverse predator communities in tropical systems is necessary for identifying the factors facilitating coexistence and promoting biodiversity. Our study demonstrates that olfaction is the primary mechanism by which Eciton hamatum is identifying, localizing, and initiating attacks against its preferred ant prey, mechanistically underpinning a strong dietary specialization. If this olfaction of specific prey-derived odors is used across the

Neotropical army ants, sensory specialization may be an important factor contributing to the observed dietary niche partitioning within ecologically important assemblages of top- predators in the Neotropics. Our results emphasize the need to account for species-level variation in olfaction preferences, to determine whether each army ant in attracted only to a small subset of potential prey odors within the structurally complex and hyper-diverse foraging environments they occupy.

19

Figures

Figure 1. A prey cache of the army ant Eciton hamatum composed primarily of the brood of their preferred prey, Acromyrmex the leaf-cutting ants. After detecting and overpowering a prey nest, E. hamatum harvests the prey’s brood and caches it immediately outside the nest entrance, before it is transported back to the army ants’ own nest. Note the distinctive spherical larvae (e.g. being carried top left) and the already darkening orange pupae of the Acromyrmex brood in the prey cache.

20

Figure 2. Boxplots of Eciton hamatum recruitment rate in ants per second to each odor treatment derived from prey ants (Acromyrmex octospinosus). Each box presents combined data from 7 colonies, except for the Live odor treatment which represents 5 colonies (Alarm trials, N= 70; Dead trials, N= 82; Live trials, N= 55; Nest trials, N= 92). In each boxplot, the box encompasses the interquartile range, a line is drawn at the median, an open circle represents the mean, whiskers extend to the upper and lower quartiles (± 1.5 times the interquartile range), and outliers are shown by the filled circles outside the whiskers. The different letters denote significantly different means among odor treatments, following square-root transformation of the data to meet the assumptions of normality and equal variances (post-hoc Tukey’s HSD, p<0.001).

21

Figure 3. Boxplots of Eciton hamatum raid speeds in centimeters per second to each odor type derived from prey ants (Acromyrmex octospinosus). Each box presents aggregate data from 7 colonies save for the Live odor group which represents 5 colonies. Dead N= 82 trials, Live N= 55 trials, and Nest N= 92 trials. In each boxplot, the box encompasses the interquartile range, a line is drawn at the median, an open circle represents the mean, whiskers extend to the upper and lower quartiles (± 1.5 times the interquartile range), and outliers are shown by the filled circles outside the whiskers. The different letters denote significantly different means among odor treatments, following log10 transformation of the data to meet the assumptions of normality and equal variances (post-hoc Tukey’s HSD, p<0.001).

22

Figure 4. Boxplots of individual average running speed in centimeters per second to each odor type derived from prey ants (Acromyrmex octospinosus). Each box presents aggregate data from 7 colonies save for the Live odor group which represents 5 colonies. Alarm N= 210 ants, Dead N= 210 ants, Live N= 150 ants, and Nest N= 210 ants. In each boxplot, the box encompasses the interquartile range, a line is drawn at the median, an open circle represents the mean, whiskers extend to the upper and lower quartiles (± 1.5 times the interquartile range), and outliers are shown by the filled circles outside the whiskers. The different letters denote significantly different means among odor treatments, following square-root transformation of the data to meet the assumptions of normality and equal variances (post-hoc Tukey’s HSD, p<0.001).

23

Chapter 2: Contrasting sensory modality responsiveness explains differences in diet breadth of co-occurring predators

Abstract

Resource-based niche partitioning contributes to the origination and maintenance of biodiversity by reducing competition and allowing otherwise ecologically similar organisms to coexist. Army ants in the Neotropics are specialist predators, largely of other ants, that partition their prey niches by specializing on different prey ant genera.

However, the best studied species, Eciton burchellii, is an exception to the trend of dietary specialization exhibited by its congeners, with its highly diverse diet giving many the false impression that army ants are generalist predators. In this study, we hypothesized that two co-occurring and closely related army ants in the genus Eciton have contrasting sensitivities to different stimulus modalities, resulting in differing behavioral responses in the presence of potential prey, thereby reducing niche overlap.

We addressed two predictions in co-occurring army ant species: 1) Eciton burchellii and

E. hamatum share similarly specialized olfaction mechanisms, allowing for detection and identification of their respective ant prey; and 2) E. burchellii show greater sensitivity to vibrational stimuli than E. hamatum, allowing for detection and attack of non-ant prey.

Here, we filmed army ant predation behavior in the field to several types of sensory stimuli from prey and non-prey, and compared the behavioral responses between species.

We found marked similarities in chemical cue sensing by the two army ant species, allowing for a similar mechanism of ant-prey detection and capture. We also observed increased sensitivity to vibrational stimuli by E. burchellii relative to E. hamatum, potentially allowing the former to perceive and attack arthropods it otherwise could not

24 chemically sense. This variation in sensory stimulus usage and responsiveness facilitates resource partitioning, thereby decreasing the intensity of interspecific competition within an ecologically important feeding guild.

Introduction

When related predators co-occur, one might expect behaviors and mechanisms facilitating successful prey-capture to experience divergent selection for increased resource partitioning that promotes coexistence (Brown & Wilson, 1956; Hutchinson,

1959; reviewed in Schluter, 2000). Researchers have an abundance of evidence suggesting that resource specialization leads to niche partitioning (coexistence) in numerous systems, including plankton, finches, anoles, insect herbivores, and pollinators

(Grover, 1997; MacArthur, 1958; Pacala & Roughgarden, 1985; Singer & Stireman,

2005; Michener, 2007; Armbruster, 2016). A common thread in these systems is that coexisting animal species frequently differ in resource use in at least one niche dimension, thus reducing, competition (Hutchinson, 1959; Schoener, 1974). We know that niche partitioning across a broad swath of taxa can occur through many types of mechanisms, including but not limited to behavioral (Hixon, 1980; Pasch et al., 2013), morphological (Losos, 1990; Irschick & Losos, 1999; Colombo et al., 2016), and sensorial mechanisms (Siemers & Schnitzler, 2004; Siemers & Swift, 2006). That said, comparatively little attention has been paid to interspecific differences the in sensory modes used by animals to detect food and their role in promoting resource partitioning within predatory arthropod foraging guilds (Tuttle et al., 1985; Bernays & Wcislo, 1994; but see parasitoid literature reviewed in Godfray, 1994).

25

All animals must find and identify the resources they need to survive in their natural habitat; in the case of predators, this entails extracting useful information from prey derived sensory modalities or cues. Across ecosystem types, predators use information from a wide range of sensory modalities including auditory (Surlykke &

Moss, 2000), chemical (reviewed in Greenstone & Dickens, 2005), electrical (Nelson &

McIver, 1999), and vibratory (Barth, 1982) information to detect and assess both one another and their prey. To successfully initiate a predator-prey interaction, first predators must use sensory information to orient toward prey sources, detect and localize potential prey, and finally discern if the potential prey is edible and able to be subdued (von der

Emde & Bleckmann, 1998). Some of these predators must use several senses simultaneously to locate their preferred prey items (reviewed in Dusenbery, 1992;

Stevens, 2013). For specialist predators these cues must be specific to their preferred prey and offer detailed information to the predator to be of any use. Such predator-prey interactions are vital to the preservation of ecosystem function and community structure, as they are integral in mediating competition, regulating population size, and influencing behavior (Werner et al., 1983; Sih et al., 1985; Schmitz, 2008; Hawlena & Schmitz,

2010). Impacts of these ecological interactions can cascade through trophic levels, which dramatically changes community composition and food web structure (Polis et al., 1989;

Matsuda et al., 1996; Kneital & Chase, 2004; Mooney et al., 2010). For any ecological interaction, a deeper understanding of these cues and how co-occurring organisms make use of them to facilitate specialization is necessary to grasp how interactions influence coexistence.

26

Ants (family: Formicidae) offer researchers a diverse and experimentally tractable taxonomic group in which to investigate sensory specialization and its role in resource partitioning within different foraging guilds. Predatory ants typically feed on other invertebrates, with several prominent lineages specialized on other speciose arthropod groups that are ecologically dominant in their own right, such as the collembolans, myriapods, and termites. Importantly, some carnivorous ant taxa seem to be specialist predators of other ants (Lach et al., 2010), and predation by ants on other ants is known to be both interspecific and intraspecific (Carroll & Janzen, 1973; Czechowski, 2013;

Mabelis, 1979).

One of the most magnificent examples of these specialist ant-predators are the army ants of the Neotropics. New World army ants, a group of five genera within the subfamily Dorylinae (Brady et al., 2014; Borowiec, 2016), are among the most diverse, and understudied groups of Neotropical arthropod predators, and potentially one of the most ecologically important. Past studies have shown conclusively that New World army ant species are predominantly specialist predators of other ants (Rettenmeyer et al., 1983;

LaPolla et al., 2002; Powell & Clark, 2004; Powell & Franks, 2006; Breton et al., 2007;

Powell, 2011), with only a small number of species adding other leaf litter invertebrates and social wasps to their diet (Rettenmeyer, 1963; Chadab, 1979; Rettenmeyer et al.,

1983; O’Donnell et al., 2005; Kaspari et al., 2011). By quantitatively comparing the diets of five army ant species in the genus Eciton, Powell and Franks (2006) showed a high level of prey-ant specificity among these coexisting predators, with each specializing on only one to three genera of other ants. Less attention has been given to the precise mechanisms of prey specialization in army ants, but recent work has shown the

27 conspicuous column-raider Eciton hamatum shows strong preference for and attraction to ant derived odor cues associated with prey items such as nest material, and the prey specific cuticular hydrocarbon profiles (Manubay & Powell 1, In preparation). While it remains unconfirmed that all army ants are using similar olfactory mechanisms to initiate predator behaviors with their respective preferred prey, it is reasonable to hypothesize that the closely related E. burchellii may be as well.

Of the five species of Eciton army ants coexisting on Barro Colorado Island, E. burchellii and E. hamatum stand out in that they are sister species (Winston et al., 2017) which are not strict ant specialists (Powell & Franks, 2006). E. hamatum’s diet principally consists of various other ants, most prominently in the genus Acromyrmex, and it is also known to raid the nests of several social wasp species in the subfamily

Polistinae throughout the Neotropics (Rettenmeyer, 1963; Schneirla, 1971). On the other hand, E. burchellii is among the most famous of army ant species because of its seeming lack of dietary specificity; unlike its congeners, 50% of E. burchellii’s diet consists of an assortment of litter arthropods such as orthopterans, roaches, scorpions, and tarantulas

(Rettenmeyer, 1963; Rettenmeyer et al., 1983; Kaspari et al., 2011). However, the other half of E. burchellii’s diet is almost entirely comprised of one ant genus, Camponotus

(Powell & Franks, 2006). Might variation in sensory modality responsiveness in these two species underpin this clear pattern of dietary divergence, and thus, coexistence?

Ants are known to be receptive to an extremely wide range of chemicals, and use them for communication, navigation, and interspecific interactions (Hölldobler & Wilson,

1990; Lenoir et al., 2001; Lach et al., 2010). Therefore, it is possible that E. burchellii may simply respond to a wider range of prey specific odor cues that E. hamatum, able to

28 identify and home-in-on odors specific to its preferred prey ant, Camponotus, as well as odors from many species of litter arthropods as well. Another possible explanation for the expanded resource breadth of E. burchellii relative to its congeners may not lie in an ability to detect many disparate odors, but rather in the use of differing sensory modalities. It is highly unlikely that E. burchellii are using any visual cues to home in on a particular resource, as all species of army ants are known to be either mostly or entirely blind (Schneirla, 1971; Gotwald, 1995). In fact, the eyes of all Eciton workers comprise of only a single facet, and are unable to form images, though they are responsive to changes in light levels (Bulova et al., 2016). Investigating the sensory cues that army ants are using to detect and choose specific prey will help us understand these common and ecologically impactful interactions, while simultaneously shedding light on how coexisting predators are partitioning available food resources through sensory modalities.

We hypothesize that army ants have varying sensitivities to different stimulus modalities, resulting in differing behavioral responses in the presence of potential prey.

Behavioral variation may limit niche overlap by allowing members of the same feeding guild to detect, identify, and attack only a subset of available prey, and thus, finely partition food resources. To test our central hypothesis, we addressed the following specific predictions in co-occurring army ant species: 1) Eciton burchellii and E. hamatum will share similar olfaction sensory mechanisms, allowing for detection and identification of their respective preferred ant prey; and 2) E. burchellii will show greater sensitivity to vibrational stimuli than E. hamatum, irrespective of prey species identity, allowing for detection and attack of non-ant prey. We tested these predictions by conducting a series of field experiments that compared and contrasted prey detection and

29 attack in E. burchellii and E. hamatum. While both species take only particular hymenopteran prey, E. burchellii has an expanded diet breadth that also includes large litter arthropods, potentially necessitating the use of additional sensory modalities for detection and capture. We assessed the behavioral responses of our focal army ants to prey by measuring two complementary components of army ant collective foraging; 1) localized recruitment rates and 2) individual forager speeds. Experiments that allowed the direct comparison of ant-prey detection and capture between the focal army ant species were complemented by a set of field experiments addressing how E. burchellii may be taking advantage of movement as an additional prey-sourced stimulus modality, allowing for its expanded diet breadth of non-ant, leaf litter arthropods.

Methods

Study site, study organisms, and colony discovery

We conducted all fieldwork on Barro Colorado Island (BCI), Panama (9°090N,

79°500W) between June and August of 2016 and 2017. BCI is a 15.6 km2 island located in Gatun Lake on the northern span of the Panama Canal, covered by semi-deciduous moist tropical forest (see Leigh, 1999; Ziegler & Leigh, 2011 for detailed information).

We located colonies of Eciton burchellii and E. hamatum army ants by walking heavily compacted trails during daylight hours until swarm raids were encountered (following methods of e.g. Franks, 1980; Franks, 1982; Vidal-Riggs & Chaves-Campos, 2008;

O’Donnell et al., 2007; Powell, 2011; Manubay & Powell, 1 In preparation). We then tracked raids in the direction of prey flow, to locate the nest or ‘bivouac’. Colonies were tracked nightly, by following the emigration traffic to the new bivouac site. This ensured

30 the location of the colony was known for data collection the following day (Powell &

Clark, 2004; Powell & Franks, 2006; Powell, 2011).

Prey ant species used in this study were Camponotus sericeiventris and

Acromyrmex octospinosus, among the preferred ant-prey species of E. burchellii and E. hamatum respectively, on Barro Colorado Island (Rettenmeyer et al., 1983; Powell &

Franks, 2006). Here, preferred prey refers to a common member of the prey-ant genera most often collected while sampling army ant prey items with methods developed in

Rettenmeyer et al., 1983. Non-prey ants in this study were Cephalotes atratus ants, which no army ant is known to hunt or consume and is absent from army ant diet records

(Authors’ personal observations). We located and marked easily accessible colonies of both prey species and non-prey ants while walking trails in search of army ant raid columns.

Sensory modality types and preparation

We prepared two treatments of sensory stimuli derived from both prey and non- prey ants, then presented these to army ant raids: 1. Freshly-frozen dead, and thus motionless, ants which offer the olfactory stimuli of the prey ant cuticular hydrocarbon profile to raids of E. burchellii and E. hamatum; and 2. a partially immobilized, living ant treatment combining the presence of ant-specific cuticular hydrocarbons, alarm pheromones, and a movement/ vibrational cue. We collected ants and rapidly euthanized them in a -20 C freezer before use, to present army ants with the freshest possible olfactory stimuli of cuticular hydrocarbon cues. This method provided hydrocarbon profiles in the absence of volatile and quickly dispersed alarm pheromones, and movement cues, and before the ants have been dead long enough to give off “rotting

31 insect” odor (Choe & Rust, 2008; Howard & Tschinkel, 1976; Wilson et al., 1958). We included the treatment of partially immobilized, living ants to test for army ant interest in vibrational stimuli. The living ant treatment was immobilized by a non-lethal procedure ablating just above the joint between the tibia and femur of each leg.

Army ant raid video

We recorded videos using a 4K (3840 x 2160 resolution) GoPro video camera with an ultra-wide-angle lens, held or mounted at an appropriate distance to capture a overhead view of the interaction between army ant raid front and experimental sensory stimuli. After identifying an army ant colony, a haphazardly selected odor cue from either the appropriate prey or non-prey ants was presented to a raid front at a distance of

20 centimeters. We selected naïve sections of the raid front for each trial until all types of sensory stimuli were presented. Each trial was replicated a minimum of 10 times per sensory stimulus type per army ant colony for both prey and non-prey ants. We recorded two metrics of army ant predatory behavior from the video; 1) army ant recruitment rates measured in ants per second, and 2) average running speed of individual army ants within the raid, measured in centimeters per second. These metrics were chosen primarily because ant recruitment is a reliable proxy for detecting a colony’s interest in monopolizing a given resource, and the speeds of individuals should be expected to change relative to normal walking speed based on their interest in the resource. Stronger responses to prey are thus represented by higher recruitment rates and slower forager speeds.

Katydid Collection and Experiment

32

In the summer of 2017, we collected katydids (Family: Tettigoniidae) from lights around the lab clearing on Barro Colorado Island, Panama between the hours of 21:00 and 01:00. Katydids were kept in a well ventilated, plastic holding cage at tropical ambient temperature with food and water until the day of trials. After finding a colony of

Eciton burchellii using the same method used for the ant study, 20 katydids to be used in that day’s trials were randomly selected and chilled at 4 degrees Celsius for ease of handling. We labeled all individuals, ablated the wings and all the legs between the tibia and the femur, and measured and recorded wet weight. All individuals were then allowed sufficient time at ambient temperature for any hemolymph leaking from ablated limbs to coagulate and dry. At this point, we matched the katydids by wet weight, assigned each pair to a moving or motionless treatment, and euthanized katydids in a -20 degrees C freezer to produce the motionless treatment material. Once processed, we transported katydids to army ant raid locations, presented selected individuals from either treatment to the raid front one at a time, and filmed the interaction with the army ants for 30 seconds. This process was repeated until all processed katydids had been offered to the swarm. From the video, we measured army ant recruitment rates, and individual running speeds in response to the katydids.

Video processing and data extraction

We extracted still images from videos and analyzed them in ImageJ version 1.49

(Schneider, Rasband & Eliceiri, 2012). Frame dimensions of each individual video were measured using the calibration function of ImageJ. We measured the rate of army ant recruitment (in ants per second) to sensory stimuli by counting the number of ants that stopped walking and began antennating after the first ant discovered the odor sample (see

33

Chadab & Rettenmeyer, 1975). We also report percentage of trails which elicited successful recruitment, defined as at least one ant being recruited to the stimuli be her sisters. The video was cut into a series of images taken at 0.5 second intervals for each trial. These images were then fed into the Manual Tracker plugin of ImageJ, where three randomly selected army ants had a series of their instantaneous velocities measured between two frames for the duration of each video trial. Instantaneous velocities were then used to calculate average individual army ant forager speed in centimeters per second in response to each stimuli type.

Sample Sizes and Statistical Analyses

We conducted all analyses presented here using R version 3.4.4 for Windows (R

Core Team, 2016). For E. burchellii recruitment trials, a total of 320 video observations were made, spread evenly across all eight colonies and four sensory stimuli treatments.

For E. hamatum recruitment trials, a total of 246 video observations were made across seven colonies (except for the moving prey treatment, where observations were pooled across five colonies). In the E. hamatum trials, we made 52 observations for moving non- prey, 57 for motionless non-prey, 55 for moving prey, and 82 for motionless prey. For E. burchellii mean running speed trials, we measured a total of 960 ants’ average speeds spread evenly across all eight colonies and four sensory stimuli treatments. For E. hamatum mean running speed trials, we measured a total of 738 ant’s average speeds across seven colonies (except for the moving prey treatment, where observations were pooled across five colonies);156 for moving non-prey ants, 171 for motionless non-prey ants, 165 for moving prey ants, and 246 for motionless prey ants. For E. burchellii recruitment trials in the katydid experiment, a total of 160 video observations were made,

34 spread evenly across eight colonies and two sensory stimuli treatments. For E. burchellii mean running speed trials in the katydid experiment, we measured a total of 480 ants’ average speed spread evenly across all eight colonies and two sensory stimuli treatments.

We aggregated all observation data across the seven colonies of E. hamatum (five colonies in the case of moving, prey data), and the eight colonies of E. burchellii after using ANOVA tests to confirm irrelevance of the colony identity factor (p>>0.05 in all cases). For both army ant recruitment rates and running speeds, we tested for differences between E. burchellii and E. hamatum for each sensory stimulus treatment using Welch’s

Two Sample T-tests on untransformed data. Welch’s correction was chosen over traditional Student’s T-tests or non-parametric tests due to increased reliability when comparing samples of unequal variance and unequal sample size, in addition to the tests’ robustness to skewed distributions at moderate to large sample sizes (Fagerland &

Sandvik, 2009; Fagerland, 2012). For the aforementioned reasons and justifications,

Welch’s Two Sample T-tests were again used to test for differences between metrics of

E. burchellii interest in moving and motionless katydids.

Results

Interspecific variation in army ant responses to ant-derived stimuli

Army ant recruitment to potential ant prey was largely similar across treatments between the two predators tested. In the case of both moving and motionless prey ants, both army ant species saw successful recruitment of at least one army ant forager in

100% of trials. However, only 51.25% of motionless, non-prey trials elicited successful recruitment of at least one E. burchellii forager, while only 47.37% of motionless, non-

35 prey trials elicited successful recruitment from E. hamatum. Recruitment rates to moving, non-prey ants were significantly higher in E. burchellii (Mean = 0.32 ± 0.12 ants/second) compared to E. hamatum (Mean = 0.08 ± 0.01 ants/second) (t129.5 = 12.49, P < 0.0001;

Figure 1C), with 100% of E. burchellii trials having non-zero recruitment. This is in strong contrast for E. hamatum, with only 28.85% of trials eliciting any successful recruitment. However, there were no statistically significant differences between army ant species in mean recruitment rates to motionless non-prey ants (t131.1 = 1.32, P = 0.19;

Figure 1A), motionless prey ants (t153.2 = 1.62, P = 0.11; Figure 1B), and moving prey ants (t111.2 = 1.42, P = 0.15; Figure 1D).

Just as with recruitment, forager running speeds were significantly different between army ant species in the moving non-prey stimuli treatment, with E. burchellii

(Mean = 2.91 ± 1.52 cm/second) being significantly slower than E. hamatum (Mean =

5.46 ± 1.63 cm/s) (t299.9 = 15.46, P < 0.0001; Figure 2C). Additionally, we found significantly different running speeds when comparing army ant responses to motionless prey ants, with E. hamatum (Mean = 3.25 ± 1.07 cm/s) being slightly faster than E. burchellii (Mean = 2.82 ± 1.83 cm/s) (t437.8 = 3.69, P < 0.001; Figure 2B). There were no statistically significant differences between army ants in mean running speeds to motionless non-prey ants (t435.1 = 1.12, P = 0.26; Figure 2A), and moving prey ants (t362.4

= 0.62, P = 0.53; Figure 2D). Mean E. burchellii speeds ranged from 7.14 ± 1.51 cm/s for motionless non-prey trials, 2.82 ± 1.43cm/s for motionless prey trials, 2.91 ± 1.52 cm/s for moving non-prey trials, and 2.56 ± 1.47cm/s for moving prey trials. Average E. hamatum speeds ranged from 6.98 ± 1.57 cm/s for motionless non-prey trials, 3.25 ± 1.07

36 cm/s for motionless prey trials, 5.46 ± 1.63 cm/s for moving non-prey trials, and 2.47 ±

1.19 cm/s for moving prey trials.

Eciton burchellii responses to sensory stimuli of non-ant prey

E. burchellii recruitment rates to moving katydids (Mean=1.05 ± 0.48 ants/second) were significantly higher relative to recruitment to motionless katydids

(Mean= 0.06 ± 0.06 ants/second) (t81.93 = -18.1, P < 0.0001; Figure 3A). Moving katydids were always highly attractive to E. burchellii, with 100% of such trials eliciting successful recruitment. For each paired moving and motionless katydid, recruitment to the moving katydid was always higher in magnitude regardless of wet weight (Figure

3A). In fact, the highest observed recruitment to a motionless katydid was only 10 ants in the 30 second trial, with 30% of motionless katydid trials seeing no recruitment at all

(Figure 3A).

We observed a complimentary pattern when measuring mean running speeds of individual army ants to each katydid stimulus type. E. burchellii mean forager running speeds were significantly slower in the presence of moving katydids (Mean = 2.38 ± 0.87 cm/s) compared to motionless katydids (Mean = 6.92 ± 0.95 cm/s) (t156.91 = 31.54, P <

0.0001; Figure 3B). For each katydid pair, mean running speeds were always slower in the presence of moving katydids relative to the motionless one regardless of wet weight

(Figure 3B).

Discussion

Here we have addressed how differing sensitivity to prey-derived stimulus modalities generates behavioral variation among co-occurring predators, thus facilitating

37 species coexistence. Consistent with the notion that closely related and co-occurring species in the same feeding guild should diverge in at least one trait vital to fitness

(Brown & Wilson, 1956), we found that Eciton burchellii and E. hamatum exhibited differing behavioral responses in the presence of a range of stimuli types. That said, the data suggest some clear similarities between E. burchellii and E. hamatum predatory behaviors elicited in response to ant-prey derived stimuli, suggesting a conserved mechanism for the chemical detection of hymenopteran prey. Additionally, both species of army ants displayed an unwillingness or inability to react to odor stimuli from non- prey ants, specifically when odor cues were decoupled from vibrational modalities.

However, E. burchellii differed significantly from E. hamatum in their interest in moving non-prey ants, as evidenced by E. burchellii’s significantly excited responses relative to its congener. This observation precipitated the related investigation into E. burchellii’s perceptions of and responses to vibrational stimuli from non-ant prey, specifically katydids. We found E. burchellii to be far more interested in moving katydid stimuli while largely ignoring motionless katydids. The apparent inability of E. burchellii to smell odors associated with katydid corpses stands in stark contrast to the great interest both species of army ant in this study showed to corpses of their preferred ant-prey. All told, the differentiated behavioral responses of Eciton species to sensory cues derived from preferred prey and non-prey ants suggests a mechanism by which army ants are successfully partitioning food resources and coexisting.

Army ants share a similarly specialized sensitivity to prey-ant odors

Under selection for an efficient sensory system, one would expect to observe organisms with the specialized ability to detect only a subset of the total sensory stimuli

38 in their environment, ideally those which lead to a potential benefit like the odors of mates and food resources (Bernays & Wcislo, 1994). Clearly, both species of army ants in this study show a similar ability to detect and respond to the odors of their respective prey-ants, regardless of whether the odor was from a moving or motionless source

(Figures 1 and 2). In addition, both species show virtually no interest in odors derived from the non-prey ant used as a control (Figures 1A and 2A). Neither army ant species showed any aversion to the non-prey ants (Cephalotes atratus) selected to be a control in this study, in that they maintained speeds that were generally fast, or close to normal running speeds in other words, when army ants encountered material from this non-prey ant (Hurlbert et al., 2008; Manubay & Powell 1, In Preparation). In terms of ant predation, both species of army ants appear to primarily be using odor based sensory modalities to discriminate between which potential food items to attack and which to ignore, as evidenced by both species’ strong responses to prey-derived stimuli and lack of interest in non-prey odors. While we did find that E. hamatum was significantly faster than E. burchellii when comparing forager speeds in the presence of motionless ant-prey, the magnitude of this difference was not particularly large and may be explained in part by the generally larger bodies and faster walking speeds of E. hamatum (Hurlbert et al.,

2008). Thus, a haphazardly selected sample of E. hamatum may have simply been larger, and thus faster, than the sample of E. burchellii it was compared to.

Researchers have known for some time that cuticular hydrocarbon profiles of insect corpses retain specific information allowing for differentiation between nestmates and non-nestmates, and even corpses of competitors (Bos et al., 2010; Cournault & de

Biseau, 2009; Diez et al., 2013); however, these results suggest that these predators are

39 also able to identify and sort corpses into prey and non-prey categories. These findings are consistent with the notion that army ants are generally excited by ant-odors which indicate some proximity to prey nests, and the army ants’ most prized spoils, namely the prey’s brood (Manubay & Powell 1, In preparation; Rettenmeyer et al., 1983; Gotwald,

1995). Taken together, these findings suggest that E. burchellii and E. hamatum share a similar sensory specialization for the odors of their preferred ant-prey. Precisely which odors of the prey ants the army ants are reacting to is yet unknown; however, given the results of this study coupled with those of Manubay & Powell 1 (In preparation), chemicals associated with the prey ant cuticular hydrocarbon are a good place to start an electroantennographic investigation (e.g. using the methods of Arn et al., 1975; D’Ettorre et al., 2004).

E. burchellii responds to vibratory modalities when interacting with potential prey

The ability to detect vibrations is widespread in insects and other invertebrates

(Cocroft & Rodríguez, 2005; Hill, 2008; Hill, 2009), so we might expect army ants to make use of such cues in foraging. While we have established that closely related army ants share a similarly specialized olfactory mechanism for detecting ant-prey, where E. burchellii differed from E. hamatum was in its relative interest in moving non-prey food resources. Specifically, E. burchellii recruited far more workers to moving non-prey ants than it did to motionless ones; when compared to the results from E. hamatum, E. burchellii’s interest in moving Cephalotes seems particularly stark (Figure 1C). That same interest is seen when looking at the individual army ant speeds as well; clearly, E. burchellii is slowing down far more in the presence of moving non-prey ants than E.

40 hamatum, and while both army ants still largely maintain standard running speeds in the presence of motionless non-prey items (Figure 2C).

Given that there were no differences in preparations of moving and motionless treatments of prey and non-prey ant items between experiments on E. burchellii and E. hamatum, these findings suggest that E. burchellii has a far greater sensitivity to vibrational sensory modalities than does E. hamatum. In fact, the anecdotal observation of this phenomenon prompted the testing of E. burchellii’s predatory behavioral responses to non-ant prey items in the last experiment presented in this study. We used katydids to conduct this experiment because they are known to comprise a sizeable percentage of the litter arthropods E. burchellii feeds on (Authors’ personal observations;

Rettenmeyer et al., 1983), and because of ease of collection.

These results suggest that E. burchellii are not chemically sensing and identifying dead katydids in the same way they sensed and identified dead prey-ants. Initiating predatory behaviors in response to vibrational signals provides a simple mechanism by which E. burchellii can expand its diet beyond social hymenopterans, unlike all other

Eciton army ants. If there is selection for sensory systems to be energetically efficient, this has the added benefit of reducing any potential energetic costs associated with being able to detect and identify the wide range of chemical odor cues of E. burchellii’s diverse arthropod prey, as cuticular hydrocarbon diversity is known to increase with increasing phylogenetic distance across lineages of insects in general and ants specifically

(Drijifhout et al., 2010; Blomquist & Bagneres, 2010). Combining lines of evidence from both ant-prey and katydid experiments, we posit that E. burchellii’s expanded non-ant

41 diet breadth relative to congener E. hamatum may be due to increased responsiveness to vibrational sensory cues caused by moving arthropod prey.

Consequences of divergent modality sensitivity

Complete overlap of resources among co-occurring organisms will often result in competitive exclusion of all but the most efficient exploiters (Hutchinson, 1959; Chesson,

2000; Chase & Leibold, 2003). Avoiding competitive exclusion simply by reducing competition is one obvious benefit of specialization; however, the mechanisms by which ecological specialization is achieved is often unknown. In this study, we have investigated how divergence in sensory modality responsiveness allows for behavioral variation during foraging and resource acquisition, contributing to resource partitioning and specialization among closely related, co-occurring arthropod predators. Foraging is one of the costliest activities that all eusocial insects must perform (Carrol & Janzen,

1973); not only costly in terms of energy expended, but also in terms of the time spent by foragers in dangerous conditions which increase worker mortality. Following this logic, a given ant colony should maximize its benefits by optimizing energetic efficiency, by decreasing time spent searching for desirable resources, by limiting potential risk to workers by only attacking desired prey items or any combination of these (Nonacs &

Dill, 1990; Dornhaus & Powell, 2010).

Army ants foraging strategies can be sub-categorized as either column raiders, in which several tendril-like raiding columns are sent out from the bivouac at the start of each foraging bought, or swarm raiders, in which a single large raid front exits the bivouac and travels in a single cardinal direction for the remainder of the foraging bout

(Schneirla, 1933; Schneirla, 1934; Rettenmeyer, 1963; Schneirla, 1971). Across the army

42 ants, column raiders are considered social insect specialists, whereas the swarm raiders have a more general diet, taking a broader range of litter invertebrates (Schneirla, 1971;

Gotwald, 1995). Within the two closely related Eciton (Winston et al., 2017) in this study, column and swarm raiding are two extremes of a spectrum of foraging behavior, as evidenced by intermediate raiding strategies present in colonies of different sizes

(Schneirla, 1934; Gotwald, 1995; Franks, 2001). Our finding of differing sensitivity to vibrational stimuli adds a new facet to this observation; if swarm raiding is a strategy to cause the majority of litter arthropods to panic, then the vibrations caused by their flight allow E. burchellii to home in on their location and attack indiscriminately. In contrast, the column raiding army ants such as E. hamatum, have a foraging mode more conducive to the rapid detection highly concentrated and defended resource, namely prey-ant nests

(Rettenmeyer, 1963; Rettenmeyer, 1983; Powell & Clark, 2004). Unlike the panic inducing swarm raids of E. burchellii, E. hamatum’s raid columns may give their hymenopteran prey colonies little advanced warning and allow for somewhat of a surprise attack on a heavily defended resource. E. burchellii’s excitability to vibratory stimuli is also consistent with the observation that many nonsocial arthropod species will either lie still and hide or otherwise remain immobile until directly discovered by the army ants, or hang motionless from a line of silk attached to leaf or stem, which the ants cannot traverse (Otis et al., 1986; Authors’ personal observations). Fittingly, increased sensitivity to vibration resulting in aggressive predatory behavior could explain the early and oft repeated observation that E. burchellii are the most excitable and belligerent of army ants (Bates, 1863; Belt, 1874; Wheeler, 1921; Schneirla, 1934).

43

Conclusion

In summary, the data presented in this study show conclusively that two closely related arthropod predators exhibit variation in sensitivity across two stimulus modalities relevant to foraging and prey-capture, namely chemical and vibrational sensing. While the similarities in chemical stimulus sensing by two different army ant species allows for a similar mechanism of ant-prey detection and capture, variation in sensitivity to vibrational stimuli allows E. burchellii to sense and attack novel prey-arthropod types.

This increased sensitivity to vibrational stimuli may be a key innovation allowing E. burchellii to be the only member of its genus to eat a significant amount of non- hymenopteran prey, thereby decreasing the likelihood of serious competition with its congeners. A particularly important avenue of investigation going forward will be testing the extent to which Eciton can detect odors specific to a range of potential prey in the field, whether from ants or other arthropods, and the precise chemical identities of said cues using electroantennographic methods.

44

Figures

Figure 5. Eciton burchellii and E. hamatum recruitment rates in ants per second to A) motionless, non-prey ants; B) motionless, prey ants; C) moving, non-prey ants; and D) moving, prey ants. N=80 observations aggregated from eight colonies of E. burchellii in all panels. For E. hamatum data, A) N= 57 means from seven colonies, B) N= 82 means from seven colonies, C) N= 52 means from seven colonies, D) N= 55 means from five colonies. In each boxplot, the box encompasses the interquartile range, a line is drawn at the median, an open circle represents the mean, whiskers extend to the upper and lower quartiles (± 1.5 times the interquartile range), and outliers are shown by the filled circles outside the whiskers. An asterisk denotes significantly different means in recruitment rates between army ant species (Welch’s T-test, p<0.05).

45

Figure 6. Eciton burchellii and E. hamatum mean individual running speed in presence of A) motionless, non-prey ants; B) motionless, prey ants; C) moving, non-prey ants; and D) moving, prey ants. N=80 means aggregated from 8 colonies of E. burchellii in all panels. For E. hamatum data, A) N= 57 means from seven colonies, B) N= 82 means from seven colonies, C) N= 52 means from seven colonies, D) N= 55 means from five colonies. In each boxplot, the box encompasses the interquartile range, a line is drawn at the median, an open circle represents the mean, whiskers extend to the upper and lower quartiles (± 1.5 times the interquartile range), and outliers are shown by the filled circles outside the whiskers. An asterisk denotes significantly different means in forager running speeds between army ant species (Welch’s T-test, p<0.05).

46

Figure 7. E. burchellii’s A) recruitment rate in ants per second and B) mean individual running speeds (cm/s) to motionless and moving katydids. Each line connects two katydids (circles) paired by wet weight between motionless and moving treatments. N= 80 observations for each treatment, spread across eight E. burchellii colonies. An asterisk denotes significantly different means in E. burchellii responses between katydid statuses (Welch’s T-test, p<0.05).

47

Chapter 3: Olfaction mediates food niche breadth in army ants

Abstract

Investigating mechanisms which support patterns of dietary specialization within ecological communities is crucial to our understanding of the factors maintaining biodiversity. New World army ants are all voracious predators of other ants which live in species-rich assemblages throughout the Neotropics. Therefore, they are an important group in which to investigate how species may be partitioning food resources and limiting competition despite their apparent ecological similarity. However, our basic knowledge of exactly which ant species army ants eat, and how army ants are detecting and discriminating among potential prey in their complex environment is lacking. On

Barro Colorado Island in Panama, we collected prey from four army ant species in the genus Eciton to characterize their diet and identify their prey to the genus and species level. We then conducted a field experiment to determine if a single army ant in the assemblage, E. burchellii, was able to sense and initiate attack of potential ant-prey chosen from a range of preferred prey species of Eciton identified in our diet census. We found very little diet overlap among the four Eciton, with no species or genus found in all predators and only two pairs of the predators sharing one prey species each. Our experiment revealed that E. burchellii strongly responds to odor and movement cues from multiple species of its preferred prey genus Camponotus, which manifests as increased recruitment and decreased localized forager speeds as they target the stimuli. Conversely,

E. burchellii does not respond to odor cues of any of its congeners’ preferred prey; however, E. burchellii showed a modest response to moving prey ants irrespective of species identity. The results of this and previous studies reveal that the detection of only

48 odors from a narrow selection of the total ant tropical ant community underpins observed patterns of dietary specialization in Eciton army ants. If ubiquitous across the Neotropical army ants, then this olfaction-based ecological specialization may facilitate patterns of resource partitioning and coexistence in these diverse predator communities.

Introduction

Specialization is a complex phenomenon, and a full understanding of it in the context of any given organism requires a large amount of detailed information on behavior, ecology, and phylogeny (Devictor et al., 2010; Futuyma & Moreno, 1988;

Irschick, Dyer, & Sherry, 2005). These complexities make empirical studies of specialization difficult to conduct when attempting to address the multifaceted nature of specialization (Holt, 2009; Irschick et al., 2005; Poisot, Bever, Nemri, Thrall, &

Hochberg, 2011). Despite the inherent difficulties, both theoretical and empirical attempts to understand specialization have contributed to broad knowledge of factors facilitating and maintaining biodiversity (Forister, Dyer, Singer, Stireman III, & Lill,

2012; Futuyma & Moreno, 1988; Ravigné, Dieckmann, & Olivieri, 2009). Importantly for this study, researchers have long known specialization can be a powerful mechanism for reducing the likelihood of competitive exclusion, and thus promoting long-term species coexistence (Chase & Leibold, 2003; Connell & Orias, 1964; Hutchinson, 1959,

Sapp, 2016).

One of the basic tenets of community ecology is that sympatric species occupying a common trophic level tend to exhibit niche differentiation and resource partitioning, particularly for resources important to fitness (Pianka, 1974; Schoener, 1974; Schoener,

49

1986). Among the most important modes of resource partitioning in ecological communities is the differentiation of food resources, such that cases of extensive dietary overlap between similar species are limited in scope (reviewed in Chase & Leibold

2003). Additionally, longstanding theory also posits that morphologically and behaviorally similar organisms might be expected to consume similar resources, to the extent that related resources types could be used with similar levels of efficiency

(MacArthur, 1969; MacArthur, 1972;). These dual considerations of diet breadth are integral for understanding specialization, with interactions between factors such as the physical environment, resource availability, and competition all affecting the breadth of an organism’s diet over time (Finlay-Doney & Walter, 2012; Steidle & van Loon, 2003).

Our efforts to quantify and compare niche breadth and dietary specialization among species are hampered by the lack of detailed observations of diet in the majority of taxa, a problem especially relevant to arthropods in the tropics where high biodiversity makes collecting such data for every species intractable.

To fully grasp the complexities of dietary specialization, we require more than lists of food consumed by a single organism (Devictor et al., 2010; Irschick et al., 2005;

Poisot et al., 2011; Ravigné et al., 2009); we must ask how, and in which contexts the organism in question is using only a small subset of available resources. Important factors mediating observed patterns of ecological specialization may be sensory perception and behavior (Bernays & Wcislo, 1994; Forister et al., 2012; Sih, Bell, & Johnson, 2004), which can be overlooked in empirical studies of resource specialization which focus on quantifying diet breadth (Bernays, 1998; Bernays, Oppenheim, Chapman, Kwon, &

Gould, 2000; Bernays & Wcislo, 1994). In fact, behavior may be the most ubiquitous

50 mechanism by which patterns of ecological specialization are generated in animals

(Caillaud & Via, 2000; Futuyma & Moreno, 1988; Ravigné et al., 2009). Another facet worth considering when measuring of specialization is phylogenetic context; one might expect co-occurring organisms that are closely related to diverge in resource use in order to avoid competitive exclusion (Brown & Wilson 1956; Hutchinson, 1959). Effort must be made to contextualize the diet breadth of closely related, co-occurring species where a detailed examination of sensory ecology and behavior exhibited therein can offer additional insight into the causes and consequences of specialization in the group.

Questions concerning dietary breadth and overlap among predator species may be especially difficult to address given the inconsistency in the distribution and abundance associated with many of their prey species, and the influence of intraspecific and interspecific competition during food acquisition. Observational studies of predator assemblages during predator-prey interactions in natural systems can reveal the realized trophic niches of study populations, in light of the various extrinsic factors, such as competition, which can influence a species’ realized niche (Sih & Christensen, 2001, Sih et al., 2004). However, a full understanding of specialization requires that we investigate traits that are associated with prey searching and prey handling, which may also shed light on the fundamental diet niche of a given group of predators (Ferry-Graham et al.

2002). In particular, sensory traits are often important determinants of how effective a predator is at detecting and attacking particular types of prey. In the case of specialist predators, the most useful type of sensory stimuli in a predator-prey interaction would be specific to their preferred prey and offer detailed information to the predator (Bernays &

Wcislo, 1994; Steidle & van Loon, 2003; Vet & Dicke, 1992).

51

Well-developed chemical perception in ants (reviewed in d’Ettorre & Lenoir,

2010), coupled with a varied diet across the group, make them an ideal system to investigate the interaction between ecological and behavioral specialization in terms of diet breadth and perceptual ability. Predatory ants typically feed on other invertebrates, with several diverse and abundant lineages specialized on other diverse arthropod groups that are ecologically dominant in their own right, such as the collembolans, myriapods, and termites (reviewed in Hollodobler & Wilson, 1990; Lach et al., 2010). Some carnivorous ant taxa, particularly in the speciose subfamily Dorylinae (Borowiec, 2016), are specialist predators of other ants (reviewed in Gotwald, 1995; Hollodobler & Wilson,

1990; Lach et al. 2010), and predation by ants on other ants is known to be both interspecific and intraspecific (Carroll & Janzen, 1973; Mabelis 1979; Markó et al.,

2014). Moreover, ants may be dangerous to many potential predators, and we might expect strong selection for the evolution of behavioral and physiological traits (Meyers &

Herrel 2005) among myrmecophages that increase capture efficiency and reduce foraging risk on only a narrow selection of ants in a community.

A deeper understanding of the mechanisms underpinning sensorial and behavioral specialization within predator-prey interactions, and resultant diet of army ants is necessary because they are among the most ecologically important predators of other ants throughout the Neotropics ( Breton, Dejean, Snelling, & Orivel, 2007; Gotwald, 1995;

LaPolla, Mueller, Seid, & Cover, 2002; Powell, 2011; Powell & Clark, 2004; Powell &

Franks, 2006; Rettenmeyer, Chadab-Crepet, Naumann, & Morales, 1983). There are 151 described species (Bolton, 2019) of Neotropical army ant, and more than a dozen species can co-occur in most tropical forest throughout Central and South America (Kaspari et

52 al., 2011; Rettenmeyer, 1963; Rettenmeyer et al., 1983; Schneirla, 1971). All species are largely blind (Bulova et al. 2016), obligately nomadic group-predators that roam the forests they inhabit in large collective raids (Rettenmeyer, 1963; Schneirla, 1971). Few locations within tropical forest are spared during intensive raids by all resident army ant species through time (Franks & Bossert, 1983; Kaspari et al., 2011; O’Donnell, Lattke,

Powell, & Kaspari, 2007), with each species harvesting an assortment of different ant prey from the same area.

We know that army ants are specialist predators of other ants (Rettenmeyer,

Chadab-Crepet, Naumann, & Morales, 1983; LaPolla, Mueller, Seid, & Cover, 2002;

Powell & Clark 2004; Powell & Franks 2006; Breton, Dejean, Snelling, & Orivel, 2007;

Powell, 2011), that they can use odor cues to differentiate prey from a non-prey ant

(Manubay & Powell 1; Manubay & Powell 2, In preparation), and that some Eciton species are attracted to odors of their preferred ant-prey (Manubay & Powell 2, In preparation). However, species level diet record information is lacking for most species in the genus Eciton. Despite detailed knowledge of E. burchellii and E. hamatum’s realized diet niche (Powell & Franks, 2006; Rettenmeyer et al., 1983; Schneirla, 1971), the factors limiting the fundamental diet niche of these species is only poorly understood, and details about how all these predators might be specialized to feed on only specific ants are lacking (but see Manubay & Powell 1; Manubay & Powell 2, In preparation).

In this study we collected extensive diet record data across the genus Eciton present on Barro Colorado Island, in order to quantify dietary niche overlap, and to inform the potential ant-prey choices for an experiment aimed at assessing Eciton burchellii behavioral responses to a suite of preferred prey from across the genus.

53

Specifically, the E. burchellii field experiment sought to address the following questions related to the diet breadth of an ecologically dominant guild of Neotropical predators: 1.

Given an array of preferred prey items from across the Eciton assemblage, are the sensory capabilities of Eciton burchellii limited to stimuli derived from their own preferred prey? 2. Does Eciton burchellii exhibit variation in predatory responses to different species within its preferred prey-ant genus? We predict the repertoire of army ant predatory behaviors (measurable responses such as high army ant recruitment rate, and foragers speeds) are only initiated upon detection of certain specific, prey derived stimuli (cuticular hydrocarbons, and/or prey movement). Concomitantly, we expect E. burchellii to exhibit little variation in behavioral responses to members of its known preferred prey genus.

Methods

Study Site, Colony Discovery, and Field Census

Fieldwork was conducted on Barro Colorado Island (BCI), Panama (9°090N,

79°500W) between March and August 2016-2018. BCI is a semi-deciduous moist tropical forested island in the central Lake Gatun of the Panama Canal, with an area of

15.6 km2 (see Leigh, 1999 and Ziegler & Leigh, 2011 for detailed information). Colonies of Eciton army ants were located by hiking heavily compacted trails during daylight hours, until raid columns were encountered ( Franks, 1980; Franks, 1982; O’Donnell et al., 2007; Powell, 2011 Vidal-Riggs & Chaves-Campos, 2008). Raid traffic was then tracked in the direction of prey flow, to locate the nest or ‘bivouac’. Colonies were tracked each night by following the emigration traffic to the new bivouac site, thus

54 ensuring the location of the colony was known for data collection the following day

(Powell, 2011; Powell & Franks, 2006).

Whenever a raid or swarm of Eciton was encountered, we backtracked in the direction of prey flow when possible until multiple foraging columns coalesced into one.

We then used fine forceps to hand-collect prey laden army ant forgers by taking a sample of 10 army ants carrying larval ant-prey and 10 ants carrying adult ant-prey. Samples were preserved in vials of 90% ethanol and stored in a climate-controlled lab environment for later identification. We identified larval ant-prey to the genus level using

George and Jeanette Wheeler’s Ant larvae: Review and synthesis (Wheeler & Wheeler,

1976), and adult ants were identified to species level using keys provided by Jack

Longino’s Ants of Costa Rica database (2010).

Common prey-ant species of each Ection species were identified among these samples and used to inform the selection of prey species presented to Eciton burchellii in the field experiment. The list of common prey species used in the experimental trials was as follows: Camponotus sericeiventris, Camponotus atriceps, Camponotus senex,

Acromyrmex octospinosus, Ectatomma tuberculatum, and Odontomachus bauri.

Cephalotes atratus is an ant that no army ant is known to hunt or consume and was included as an outgroup comparison (following methods of Manubay & Powell 1;

Manubay & Powell 2, In preparation).

Sensory modality types and preparation

We prepared two treatments of sensory stimuli derived from all seven potential prey ants, then presented these to army ant raids: 1. Freshly-frozen dead, and thus motionless, ants which offer the olfactory stimuli of the prey ant cuticular hydrocarbon

55 profile to raids of E. burchellii; and 2. a partially immobilized, living ant treatment combining the presence of ant-specific cuticular hydrocarbons, alarm pheromones, and a movement/ vibrational cue. We collected ants and rapidly euthanized them in a -20 C freezer before use, to present army ants with the freshest possible olfactory stimuli of cuticular hydrocarbon cues. This method provided cuticular hydrocarbon profiles without volatile alarm pheromones and movement cues, and before the ants have been dead long enough to give off “rotten insect” odor (Choe & Rust, 2008; Howard & Tschinkel, 1976;

Wilson, Durlach, & Roth, 1958). We included the treatment of partially immobilized, living ants to assess army ant interest in vibrational stimuli. The moving ant treatment was partially immobilized by a non-lethal procedure ablating just above the joint between the tibia and femur of each leg, thus allowing the prey to cause vibrations through moving but without the capacity to avoid the advancing swarm.

Behavioral Experiment Data Collection

A total of nine army ant colonies were tested from May to July of both 2017 and

2018, resulting in 1260 independent trials of E. burchellii swarms assayed with odors from seven potential prey species, both moving and motionless. Each trial was recorded with a 4K (3840 x 2160 resolution) GoPro video camera, to allow subsequent data extraction. The camera was setup in the ultra-wide-angle lens mode and held or mounted at an appropriate distance to capture an overhead view of the interaction between the E. burchellii swarm front and the stimuli treatment.

After finding an E. burchellii colony and locating the swarm front, a randomly selected stimuli treatment, from one of the seven potential prey-ant species, was presented to a naïve portion of the swarm front at approximately 20 centimeters. Naïve

56 raid fronts within the overall branching structure of the swarm were selected haphazardly for each trial until treatment material was depleted. Each trial was replicated a minimum of 10 times per treatment, per E. burchelliii colony for all seven potential prey-ant species.

Data Extraction

The response of the army ants to the odor treatments was captured by extracting two metrics from footage of each trial; 1) army ant recruitment rates measured in ants per second; and 2) average running speed of three individual army ants within the raid, measured in centimeters per second. These metrics were chosen because shifts in ant recruitment and speed are reliable proxies for army ant attacks against their prey

(Manubay & Powell 1, In preparation). In brief, army ants in the swarm front have a steady walking speed with the ants evenly spread over the forest floor as they collectively search for prey. Once prey is detected, individuals are rapidly recruited to the site of the prey within the raid front (Chadab & Rettenmeyer, 1975), and the speed of individuals within it drops as the ants switch into localized search and attack behavior to capture the prey. Stronger responses to prey are thus represented by higher recruitment rates and slower speeds of individual foragers. We also monitored whether E. burchellii picked up and began stereotypical prey transport behavior (described in Rettenmeyer, 1963;

Schneirla, 1971) of encountered potential prey during the 30 second trials, in order to assess whether army ant foragers identified them as suitable prey.

To calculate the metrics of army ant response, a series of still images were captured from each video and analyzed in ImageJ version 1.49 (Schneider, Rasband &

Eliceiri, 2012). The interval between images was 0.5 seconds. Frame dimensions of each

57 individual video were measured using the calibration function of ImageJ. E. burchellii recruitment rate to odor treatments was measured by counting the number of ants that stopped walking and began antennating after the first ant discovered the odor source (see

Chadab & Rettenmeyer, 1975). Mean individual army ant speed was calculated from instantaneous velocities. The series of captured images was fed into the Manual Tracker plugin of ImageJ, where three randomly selected army ants had their instantaneous velocities measured between two frames for the duration of each video trial, to allow the calculation of the individual average speeds for a total of 3780 ants over 1260 independent trials.

Data Analyses

We conducted all analyses for this study with R version 3.4.4 (R Core Team

2018). To detect differences in taxonomic composition of prey ants, we calculated

Pianka’s overlap index (Pianka, 1974) between the four Eciton species based on relative abundance of each genus of larval and adult prey ants in our collections using the Spaa package (Zhang, 2016). We also calculated Pianka’s overlap index for species level data of adult prey ants. This index ranges between 0 (no diet overlap) and 1 (complete overlap). We directly tested for significance of dietary overlap among Ection at both taxonomic levels using null model randomization procedures in the EcoSimR package

(Gotelli, Hart, & Ellison, 2015), with the RA3 randomization algorithm (designed to detect nonrandom niche overlap patterns e.g. Winemiller & Pianka, 1990).

We asked whether E. burchellii picked up and began stereotypical prey transport behavior of potential prey during the 30 second trials using binomial yes/no data which was then analyzed to test for equality of proportion between successful and failed

58 attempts across prey types, using Chi-square tests. We used mixed design analyses of variance (ANOVA) to test for differences in recruitment rates and running speeds across stimuli and prey treatments (fixed effects), while accounting for the potential non- independence of the random colony factor. Post hoc analyses for differences among prey species and stimuli treatments were conducted using the Tukey’s Honest Significant

Differences test.

Results

Prey Collections

We identified 32 species of prey ant across ten genera among the diet samples collected from four common species of Eciton on BCI (Table 1). Taxonomic composition of the prey ants showed limited to no overlap across the four Eciton species (Fig.1 and

Fig.2). In E. burchellii, the genus Camponotus constituted 100% of the ant prey across all samples. In E. dulcium, 100% of prey came from the subfamily Ponerinae, with the most common prey genus being Odontomachus (66.88%). E. hamatum had the most varied diet at the level of prey subfamily, but by far the most preferred prey was Acromyrmex

(54.83%). Finally, E. mexicanum also had a varied diet, with four prey genera taken, but

Ectatomma was by far the most preferred prey genus (65.19%).

Pianka pairwise comparisons indicated a very low degree of dietary overlap at the levels of prey genus and species across the four Eciton species based on measures of relative abundance of prey ants (Table 2). Null model analysis demonstrated that the observed mean of dietary overlap at the genus level for all Eciton species combined

(0.039) was not significantly less than expected by chance, after 5000 randomizations

59

(simulated mean = 0.177, P = 0.068). However, the observed mean of dietary overlap at the species level for all Eciton species combined (0.011) was significantly less than expected by chance, again after 5000 randomizations (simulated mean = 0.173, P

<0.001).

Behavioral Responses to Potential Prey

Army ant recruitment rates were significantly different across prey species (mixed design ANOVA; F (6, 1260 = 339.05), and P<0.001; Fig. 3) and between stimuli type

(mixed design ANOVA; F (1, 1260 = 568.77), and P<0.001; Fig. 3) with no effect of the random colony identity factor (χ2 (1) = 0.01, and P=0.99). Post hoc pairwise comparisons revealed significantly higher army ant recruitment to moving versus motionless prey within each prey species (Tukey’s HSD, all pairwise comparisons, P<0.001).

Additionally, recruitment was always significantly higher to Camponotus prey species than all other potential prey species, irrespective of stimuli treatment (Tukey’s HSD, all pairwise comparisons, P<0.001). We found no significant differences in army ant recruitment between or among non-Camponotus potential prey, irrespective of stimuli treatment (Tukey’s HSD, all pairwise comparisons, P>0.05).

Army ant mean running speeds were significantly different across prey species

(mixed design ANOVA; F (6, 3750 = 375.67), and P<0.001; Fig. 4) and between stimuli type (mixed design ANOVA; F (1, 3750 = 1877.75), and P<0.001; Fig. 4) with no effect of the random colony factor (χ2 (1) = 1.274, and P=0.259). Post hoc analyses revealed significantly slower mean forager speeds in response to moving versus motionless prey within each prey species (Tukey’s HSD, all pairwise comparisons, P<0.05). Additionally, mean forager speeds were always significantly slower in the presence of Camponotus

60 prey species relative to all other potential prey species assayed, irrespective of stimuli treatment (Tukey’s HSD, all pairwise comparisons, P<0.05). We found no significant differences in mean army ant forager speed between or among non-Camponotus potential prey, irrespective of stimuli treatment (Tukey’s HSD, all pairwise comparisons, P>0.05).

In no instance did any E. burchellii foragers pick up and begin to carry a moving ant of any species back to the bivouac within the 30 second trail. Among motionless potential ant-prey, there are significant differences in successful pickup attempts among species (χ2

=129.12, P<0.001), with each Camponotus species having significantly increased rates of pickup than non-Camponotus species (Chi-Squared Tests, P<0.001 for all comparisons).

We found no significant differences in successful pickups among Camponotus species

(Chi-Squared Tests, P>0.05 for all comparisons).

Discussion

Understanding the complexities of specialization requires a multifaceted approach that incorporates ecology, behavior, and phylogenetic context (Devictor et al., 2010;

Futuyma & Moreno, 1988; Irschick et al., 2005; Poisot et al., 2011). Incorporating these disparate fields is a challenge when posing questions and designing experiments that address specialization, but can allow for robust empirical assessment of the causes and context dependencies of specialization. In this study, we measured diet through robust prey sampling within Eciton, a genus of ecologically similar, co-occurring predators, and quantified diet breadth overlap among members of the feeding guild. Additionally, we assessed the degree of sensory and behavioral specialization in Eciton burchellii with experiments in the field where potential prey choice was informed by our diet collection

61 data from across the genus. We found that army ants in the genus Eciton exhibit surprisingly low dietary overlap at the levels of both the prey genus and species (Figures

1 and 2, Table 2). Additionally, we found that when E. burchellii is presented with an ecologically informed range of odors from the preferred prey of its congeners, it does not initiate behaviors associated with foraging or attack (Figures 3 and 4). However, E. burchellii does exhibit great interest in all stimuli from Camponotus ants, and mild interest in movement stimuli regardless of prey species identity (Figures 3 and 4). Our results indicate that observed patterns of ant-based diet breadth in E. burchellii are explained primarily by initiation of prey capture behaviors in response to prey odors intrinsic to the cuticle, and secondarily, to detection of prey vibrational cues irrespective of species identity.

Diet Breadth in Eciton on Barro Colorado Island

Army ants across the genus Eciton take a variety of ant-prey species; however,

Table 1 shows that each Eciton species specializes on between one and four genera of ant prey. The genera taken by E. burchellii and E. dulcium are from a single subfamily

(Formicinae and Ponerinae respectively), but generic level diversity of prey is much higher for the former than the later. Additionally, there is very little overlap in prey- genera among Eciton species, and even less species level overlap (Table 2). In fact, in the shared prey genus between E. burchellii and E. hamatum, only one adult ant species

(Camponotus brevis) was found to overlap between that army ant pair (Figure 1 and

Figure 2). A similar pattern was observed between E. dulcium and E. mexicanum, with only one adult ant prey species (Odontomachus bauri) accounting for the overlap in diet breadth (Figure 1 and Figure 2). No other pair of Eciton species exhibited any level of

62 dietary overlap in our diet sampling (Table 2). To our knowledge, these records are the first to present species-level identifications of the prey ants comprising the diets of E. dulcium and E. mexicanum in Panama. Our results are largely in agreement with both historical anecdotal prey reports (Powell & Baker, 2008; Rettenmeyer et al., 1983), and the only published quantitative prey assessment of these two army ant species in which diet was identified to the genus level (Powell & Franks, 2006).

We acknowledge several limitations to our prey sampling methods; primarily that prey ant eggs were not identified, that ant larva were not identified to species, and that relative to known prey intake rates of army ants, our sample sizes are relatively low

(Powell, 2011). These design limitations may have caused us to miss rarer prey species, and species where army ants forage on prey ants from different developmental stages at unequal proportions. In other words, while any two Eciton might share only one prey species in the adult prey records, it is entirely possible that larval prey might be from many more species resulting in underestimated dietary overlap. We can see an example of this in the data presented in Figures 1 and 2, where E. hamatum was recorded with

Pheidole larval prey, but not adult Pheidole. In future studies using the classic animal barcode Cytochrome Oxidase I might give researchers a more accurate measure of diversity in larval prey species than morphological identification alone; however, a robust

COI database for Central American ants only exists for those in Costa Rica (BOLD

2018), so resolution may not be clearer for these types of analyses. That said, according to David Donoso’s exhaustive (unpublished) list of ant species present on Barro Colorado

Island, our prey records indicate that each Eciton species sampled collected a non-trivial

63 proportion of the total species for each adult ant genus they ate. Thus, we are confident that our sampling presents us an accurate picture of the diet breadth of Eciton on BCI.

Preferences for certain forest stratums in which to raid may also partially account for the limited dietary overlap between the E. burchellii and E. hamatum versus E. dulcium and E. mexicanum pairs, just as seems to be the case in assemblages of Old

World army ants in the genus Aenictus (Hashimoto & Yamane, 2014). While all species have been recorded to forage at ground level, E. burchellii and E. hamatum forage on the forest floor and several layers of the canopy (Hoenle et al., in review; authors observation). This is in direct contrast to E. dulcium and E. mexicanum which are known to mostly raid on the ground, through the leaf litter, and on low vegetation (Hoenle et al.,

In review; authors observation). For example, we found only ground nesting

Odontomachus prey in our E. dulcium and E. mexicanum records, despite the diversity of

Odontomachus which nest arboreally (Longino, 2010). Given that tropical ant communities differ greatly between the forest floor and canopy due to differing nesting biology, and physiological tolerances (Longino & Nadkarni, 1990; Wheeler 1910; Wilson

1959), it follows that these two groups of Eciton would have limited overlap in diet breadth, especially at the prey species level.

Differences in forest stratum usage does not fully explain niche breadth differences, as E. burchellii and E. hamatum also forage extensively on the forest floor

(Rettenmeyer 1963, Schneirla 1971). For example, E. hamatum’s most common prey were in the genus Acromyrmex which; of the two species, A. octospinosus nests in the litter (Longino, 2010) and A. volcanus nests in the canopy (Wetterer, 1993). Interestingly,

E. dulcium and E. mexicanum were once thought to both consume Pachycondyla prey;

64 however, after recent taxonomic revision and splitting in the genus (Schmidt & Shattuck,

2014), E. dulcium consume Pachycondyla while E. mexicanum takes Neoponera. We postulate that these two Eciton species used olfactory cues, just as E. burchellii did in the current study, to distinguish between distinct taxa that were recognized as one in ant taxonomy until recently. It should also be noted that although this sampling design was not meant to detect non-ant prey items, our sampling supports the literature describing E. dulcium and E. mexicanum as strict ant specialists (Powell & Franks, 2006; Rettenmeyer et al., 1983), as we never observed any non-ant prey being attacked or transported by either. However, E. burchellii and E. hamatum are known to take other prey types such as

Polistine wasps (Rettenmeyer 1963, Schneirla 1971), and in the case of E. burchellii, a vast assortment of litter arthropods ( Franks, 1983; Kaspari et al., 2011; Rettenmeyer

1963; Rettenmeyer et al., 1983).

Responses to ant odor cues mediates prey choice in Eciton burchellii

Predation is a multi-step process, in which the first steps are identifying and localizing potential prey (Dicke & Grostal, 2001; Finlay-Doney & Walter, 2012; Vet &

Dicke, 1992). Specific chemical cues, and to a lesser degree, prey movement cues, mediate the prey localization and identification steps of predation, which results in observed patterns of diet breadth in Eciton burchellii. Our results support the notion that army ants are capable of distinguishing between odors of potential prey-ants (Manubay &

Powell 1; Manubay & Powell 2, In preparation), even in the context of being presented with a full assemblage’s worth of preferred prey species. Remarkably, when E. burchellii was presented with the opportunity to attack and consume preferred prey ants of other

Eciton species, almost no recruitment was initiated (Figure 3) or forager slow down

65 recorded (Figure 4) in the case of in the motionless prey treatment. This is in direct contrast to the robust recruitment measured to all species of motionless Camponotus prey, supporting the observation that persistent prey-odor cues associated with the prey cuticle and colony are what direct foraging and initiate attack of prey (Manubay &

Powell, In preparation).

Despite their somewhat stereotypical behaviors that direct prey detection, recognition, localization, and subjugation, many predators are known to take prey items other than primary prey given the opportunity (Finlay-Doney & Walter, 2012). What is astonishing is that even when presented with defenseless prey of its congeners, E. burchellii made no effort to pick up non-Camponotus ants and bring them back to the bivouac for consumption. Moving Camponotus were not picked up for consumption in the time frame of the trial, but were always subdued and removed from the area shortly thereafter (personal observation). This is consistent with the fact that subduing live prey in army ants takes longer than the 30 second trial time allotted, as army ants will grapple and attempt to sting preferred prey once encountered (Powell & Clark, 2004).

Observations of the experimental area after the army ant raid had moved on often revealed the bodies of non-Camponotus ants from both stimuli treatments were simply left behind by E. burchellii; however, solitary foragers of other ants in the genus

Ectatomma would often collect the leftover corpses, presumably for consumption

(personal observation). Our results show that the ant-portion of the diet does not appear to be particularly plastic in E. burchellii in natural contexts, and provides evidence suggesting that food items of close relatives in the local army ant assemblage are not identified as desirable prey.

66

That E. burchellii recruits modestly to any moving potential prey-ant irrespective of species identity is also consistent with previous research demonstrating the mechanism of non-ant prey attack makes use of vibrations generated by struggling prey arthropods

(Manubay & Powell 2, In preparation). That Eciton burchellii refuses to attack and consume motionless non-Camponotus prey ants demonstrates convincingly that E. burchellii is a sensorial specialist of Camponotus odors, recruiting to and attacking on only those particular ants, resulting in observed patterns of diet breadth in the species. In other words, it appears that E. burchellii is not simply failing to encounter and capture preferred prey of its congeners because of foraging stratum portioning or defensive capabilities of prey ants, but instead is not physiologically tuned to the odors of non-prey ants and will not initiate predatory behaviors when they do cross paths. This is consistent with the published observation that close relative E. hamatum does not attack or even seem to respond to the presence of Atta leaf cutting ants, despite its preferred prey item on which it is sensorially specialized (Manubay & Powel 1, In preparation) being the closely related Acromyrmex (Powell & Clark 2004, Powell 2011).

The precise identity of the chemical cues being used by army ants to detect their preferred ant prey is not definitively known; however, the results of this study and previous research suggests cuticular hydrocarbon profiles are the likely candidate (Aidan ch1+2). We know that the primary function of cuticular hydrocarbons across insects is to prevent desiccation (Blomquist & Bagnères, 2010; Gibbs, 1998), but in social insects this class of chemicals have other purposes, namely intracolonial communication and nestmate recognition (reviewed in Van Zweden & D’Ettorre, 2010; Leonhardt, Menzel,

Nehring, & Schmitt, 2016). Cuticular hydrocarbon profiles of the same species typically

67 have the same qualitative make-up, and qualitative differences in profiles are expected to increase with increasing phylogenetic distance (e.g. Nowbahari et al., 1990; Bagnères &

Wicker-Thomas, 2010). Complimentary lines of evidence from the chemical ecology and antennal physiology literature (di Mauro et al., 2015; Pask et al., 2017; Sharma et al.,

2015; Slone et al., 2017) suggests that the broad range of odorant receptors ants possess allow for discrimination not only different classes of hydrocarbons, but also between minute changes in CHC concentrations (di Mauro et al., 2015). Such broad receptivity, coupled with the behavioral evidence presented in this study (Figures 3 and 4), suggest army ants can use CHC profiles to discriminate among potential prey and guide predator- prey interactions. The results of the present study, and the results of previous studies in the closely related E. hamatum (Manubay & Powell 1; Manubay & Powell 2, In preparation), suggests that many, if not all, army ants are capable of detecting, and discriminating between, prey odors for use in localization and attack.

Consequences of sensory biases and divergent diet breadth in an Eciton assemblage

Intense competition and competitive exclusion have been considered the norm in ant ecology (Parr & Gibb, 2010; Cerda, Arnana, & Retana, 2013). Many ant species effectively displace others from high-quality resources through direct interference, and some species can even displace others from large territories (Hölldobler, 1983). Because ants are among the most abundant animals in tropical rain forests ( Hölldobler & Wilson

1990; Lach, Parr, & Abbott, 2010), we should expect strong competition among ant species. However, army ants appear to be a notable exception to this expectation, with more than 20 species often coexisting in small tracts of tropical forest (Kaspari et al.,

2011; Powell & Baker, 2008; Rettenmeyer et al., 1983). Our prey collections support the

68 previously reported notion that army ants in the genus Eciton are primarily specialist predators of a specific subset of other ant genera (Franks & Bossert, 1983; O’Donnell et al., 2007; Powell & Franks, 2006; Rettenmeyer et al., 1983). Additionally, the results of our experiment have shown that E. burchellii display consistent behavioral responses to odor stimuli of only their preferred prey across a range of contexts where cuticular hydrocarbons are likely to be the dominant species-specific olfactory signature that is encountered. Moreover, E. burchellii largely ignores odors of congener’s preferred prey ants, but consistent with results of Manubay & Powell 2 (In preparation), do initiate a weak predatory response if presented with movement/vibrational stimuli irrespective of prey-ant identity. Previous studies have found that the closely related E. hamatum also shares the same olfactory means of dietary specialization (Manubay & Powell 1, In preparation), suggesting this is the common mechanism of prey detection and attack across the genus Eciton. These sensory perception mechanisms can provide army ants with specific and detailed information on the identity of potential prey ants encountered, and thus, localize and attack only a particular subset of ants in the community.

Ecologists have long postulated that specialization on limiting resources can reduce competition and maintain species diversity in tropical communities (Dobzhansky,

1950; MacArthur, 1969; Schemske, Mittelbach, Cornell, Sobel, & Roy, 2009; Sapp,

2016). The ability to detect species-specific prey odors, as we have demonstrated here for

E. burchellii (and previously for E. hamatum in Aidan Ch1), provides the mechanism by which army ants could be partitioning ant prey resources within their diverse assemblages. We know that sensory adaptations can increase efficiency in cue reception and transmission, while simultaneously resulting in sensory bias on perception in the

69 environment (Endler, 1992; Fuller, Houle, & Travis, 2005; Stevens 2013). These sensory biases which can initiate or alter specific behaviors related to prey detection and capture, resulting in observed diet breadth by altering sensory access to food (Barclay & Brigham,

1991; Bernays & Wcislo, 1994; Raine & Chittka, 2007). The result is a fine-grained, chemosensory-based mechanism underpinning dietary specialization (Elizabeth A

Bernays & Wcislo, 1994; Steidle & van Loon, 2003; Vet & Dicke, 1992) which can reduce interspecific competition and facilitate coexistence.

Conclusion

Understanding of the specific mechanisms that maintain diverse predator communities in tropical systems, coupled with detailed accounts of the feeding ecology of species in an assemblage, are necessary for identifying the factors allowing niche partitioning, facilitating coexistence, and promoting biodiversity. Army ants in the genus

Eciton are largely specialist predators of other ants, with diet breadth mediated by chemical perception of their ant-prey. Closely related prey, such as those in the genus

Camponotus, likely share similar CHC profiles which elicit predatory behaviors in E. burchellii; however, non-Camponotus odors appear to either not be detected or not be of interest to E. burchellii. Our diet collection data indicates that the understudied army ants

E. dulcium and E. hamatum are specialists of only a handful of other ant species, and that dietary niche overlap across the genus on Barro Colorado Island is negligible at the prey genus and species level. Our results highlight the need for quantitative examinations of diet in tropical predators, in order to parse how mechanisms of sensorial and behavioral specialization might be influencing observed patterns of food resource partitioning.

70

Tables

Table 1. Summary table of Eciton diet observations made on Barro Colorado Island. The number of unique prey genera and species in the samples appear in parentheses in last two columns.

Eciton Species Colonies Number Number of Number of Number of Sampled of Subfamilies Prey Genera Prey Species Samples (Adult and (Adult) Larva)

E. burchellii 30 600 1 1 (0) 9 (8)

E. dulcium 24 480 1 3 (2) 7 (6)

E. hamatum 30 600 3 4 (2) 10 (9)

E. mexicanum 26 520 2 4 (3) 8 (7)

71

Table 2. Dietary niche overlap at the prey genera and species level expressed as Pianka’s Index between the four Eciton species Barro Colorado Island, Panama. Values of zero indicate there were no shared prey items within the Eciton species comparison, resulting in no dietary overlap.

Eciton Comparison Genera Dietary Overlap Species Dietary Overlap

E. burchellii x E. dulcium 0 0

E. burchellii x E. hamatum 0.134 0.008

E. burchellii x E. mexicanum 0 0

E. dulcium x E. hamatum 0 0

E. dulcium x E. mexicanum 0.101 0.058

E. hamatum x E. mexicanum 0 0

Average Across Eciton 0.039 0.011

72

Figures

Figure 8. Pooled adult and larval ant-prey samples collected from each species of Eciton on Barro Colorado Island, identified to genus. Number of prey samples is displayed on the y-axis, and each color represents a single genus of ant prey.

73

Figure 9. Adult ant-prey samples collected from each species of Eciton on Barro Colorado Island, identified to species. Number of prey samples is displayed on the x-axis, and each color represents a single genus of ant prey.

74

Figure 10. Boxplots of Eciton burchellii recruitment rate in ants per second to each potential prey species and stimuli treatment. Each box presents combined data from 9 colonies (N = 90). Stars above pairs of boxes indicate significant differences in E. burchellii recruitment between prey stimuli treatment (Tukey’s HSD, P<0.001). In each boxplot, the box encompasses the interquartile range, a line is drawn at the median, an open circle represents the mean, whiskers extend to the upper and lower quartiles (± 1.5 times the interquartile range), and outliers are shown by the filled circles outside the whiskers.

75

Figure 11. Boxplots of Eciton burchellii mean individual running speeds in centimeters per second in response to each potential prey species and stimuli treatment. Each box presents combined data from 9 colonies (N = 270). Stars above pairs of boxes indicate significant differences in E. burchellii recruitment between prey stimuli treatment (Tukey’s HSD, P<0.001). In each boxplot, the box encompasses the interquartile range, a line is drawn at the median, an open circle represents the mean, whiskers extend to the upper and lower quartiles (± 1.5 times the interquartile range), and outliers are shown by the filled circles outside the whiskers.

76

References

Al Abassi, S., Birkett, M. A., Pettersson, J., Pickett, J. A., Wadhams, A. L., & Woodcock, C. M. (2000). Response of the seven-spot ladybird to an aphid alarm pheromone and an alarm pheromone inhibitor is mediated by paired olfactory cells. Journal of Chemical Ecology, 26(7), 1765-1771.

Armbruster, W. S. (2016). The specialization continuum in pollination systems: diversity of concepts and implications for ecology, evolution and conservation. Functional Ecology. https://doi.org/10.1111/1365-2435.12783

Arn H, Städler E, Rauscher S (1975). The electroantennographic detector – a selective and sensitive tool in the gas chromatographic analysis of insect pheromones. Z Naturforsch C 30, 722–725.

Barclay, R. M., & Brigham, R. M. (1991). Prey detection, dietary niche breadth, and body size in bats: why are aerial insectivorous bats so small? The American Naturalist, 137(5), 693-703. https://doi.org/10.1086/285188

Barth, F. G. (1982). Spiders and vibratory signals: sensory reception and behavioral significance. Spider communication: mechanisms and ecological significance, 67- 120.

Bates, Henry W. (1863). Naturalist on the River Amazons: A Record of the Adventures, Habits of Animals, Sketches of Brazilian and Indian Life, and Aspects of Nature under the Equator, during Eleven Years of Travel. John Murray, London, UK.

Belt, Thomas. (1874). The Naturalist in Nicaragua. Cosimo, New York, USA.

Bernays. (1998). The Value of Being a Resource Specialist: Behavioral Support for a Neural Hypothesis. The American Naturalist, 151(5), 451–464. doi:10.2307/2463523

Bernays, E. A., Oppenheim, S., Chapman, R. F., Kwon, H., & Gould, F. (2000). Taste Sensitivity of insect herbivores to deterrents is greater in specialists than in generalists: A behavioral test of the hypothesis with two closely related caterpillars. Journal of Chemical Ecology, 26.

Bernays, E. A., & Wcislo, W. T. (1994). Sensory Capabilities, Information Processing, and Resource Specialization. The Quarterly Review of Biology, 69(2), 187. Retrieved from http://dx.doi.org/10.1086/418539

Blomquist, G. J., & Bagnères, A. G. (2010). Insect hydrocarbons: biology, biochemistry, and chemical ecology. (G. J. Blomquist & A. G. Bagnères, Eds.). New York: Cambridge University Press.

77

Bolton, B. (2019). An online catalog of the ants of the world. Available from http://antcat.org. (accessed [02/21/2019])

Borowiec, M. (2016). Generic revision of the ant subfamily Dorylinae (Hymenoptera, Formicidae). ZooKeys, 608, 1. Retrieved from http://dx.doi.org/10.3897/zookeys.608.9427

Bos, N., Guerrieri, F. J., & d’Ettorre, P. (2010). Significance of chemical recognition cues is context dependent in ants. Animal Behaviour, 80(5), 839-844.

Bulova, S., Purce, K., Khodak, P., Sulger, E., & O’Donnell, S. (2016). Into the black and back: the ecology of brain investment in Neotropical army ants (Formicidae: Dorylinae). The Science of Nature, 103(3-4), 31. https://doi.org/10.1007/s00114- 016-1353-4

Brady, S. G., Fisher, B. L., Schultz, T. R., & Ward, P. S. (2014). The rise of army ants and their relatives: diversification of specialized predatory doryline ants. BMC Evolutionary Biology, 14(1), 1. https://doi.org/10.1186/1471-2148-14-93

Breton, J.L., Dejean, A., Snelling, G. & Orivel, J. (2007). Specialized predation on Wasmannia auropunctata by the army ant species Neivamyrmex compressinodis. Journal of Applied Entomology, 131, 740–743. https://doi.org/10.1111/j.1439- 0418.2007.01221.x

Brown, W. L., & Wilson, E. O. (1956). Character displacement. Systematic zoology, 5(2), 49-64.

Caillaud, M. C., & Via, S. (2000). Specialized Feeding Behavior Influences Both Ecological Specialization and Assortative Mating in Sympatric Host Races of Pea Aphids. The American Naturalist, 156(6), 606–621. Retrieved from https://www.jstor.org/stable/pdf/10.1086/316991.pdf?casa

Cárdenas, M., Jiroš, P., & Pekár, S. (2012). Selective olfactory attention of a specialised predator to intraspecific chemical signals of its prey. Naturwissenschaften, 99(8), 597-605. https://doi.org/10.1007/s00114-012-0938-9

Carroll, C. R., & Janzen, D. H. (1973). Ecology of foraging by ants. Annual Review of Ecology and Systematics. Retrieved from http://www.jstor.org/stable/2096812

Cerda, X., Arnan, X., & Retana, J. (2013). Is competition a significant hall- mark of ant (Hymenoptera: Formicidae) ecology? Myrmecol. News 18, 131–147.

Chadab, R. (1979). Army-ant predation on social wasps. PhD dissertation, University of Connecticut, Storrs, Conn.

Chadab, R., & Rettenmeyer, C. W. (1975). Mass recruitment by army

78

ants. Science, 188(4193), 1124-1125. https://doi.org/10.1126/science.1215991

Chadab-Crepet, R., & Rettenmeyer, C. W. (1982). Comparative behavior of social wasps when attacked by army ants or other predators and parasites. In Biology of social insects: proceedings, Ninth Congress, International Union for the Study of Social Insects, Boulder, Colorado, August 1982/edited by MD Breed, CD Michener, HE Evans. Boulder, Colo.: Westview Press.

Chase, J. M., & Leibold, M. A. (2003). Ecological niches: linking classical and contemporary approaches. University of Chicago Press. https://doi.org/10.7208/chicago/9780226101811.001.0001

Chesson, P. (2000). Mechanisms of maintenance of species diversity. Annual review of Ecology and Systematics, 31(1), 343-366.

Choe, D. H., & Rust, M. K. (2008). Horizontal transfer of insecticides in laboratory colonies of the Argentine ant (Hymenoptera: Formicidae). Journal of economic entomology, 101(4), 1397-1405. https://doi.org/10.1093/jee/101.4.1397

Cocroft, R. B., & Rodríguez, R. L. (2005). The behavioral ecology of insect vibrational communication. Bioscience, 55(4), 323-334.

Colombo, M., Indermaur, A., Meyer, B. S., & Salzburger, W. (2016). Habitat use and its implications to functional morphology: niche partitioning and the evolution of locomotory morphology in Lake Tanganyikan cichlids (Perciformes: Cichlidae). Biological Journal of the Linnean Society, 118(3), 536-550.

Connell, J., & Orias, E. (1964). The Ecological Regulation of Species Diversity. The American Naturalist, 98(903), 399–414. Retrieved from https://www.jstor.org/stable/pdf/2459143.pdf?casa_token=lt4TAF7obEYAAAAA:d QpkpaXvSS6N8qF8pLSgXwvesSDwZdZ6HNA2Iy9etgvS5Gpx6MjUGryhs3vprY0 JR6Bpsw2_w35jjoEGz94hxVKddgLdqjizsNgNHPWUn_CfQs3Lkp8

Cournault, L., & de Biseau, J. C. (2009). Hierarchical perception of fertility signals and nestmate recognition cues in two dolichoderine ants. Behavioral ecology and sociobiology, 63(11), 1635-1641.

Czechowski, W., Marko, B., Radchenko, A., & Slipinski, P. (2013). Long-term partitioning of space between two territorial species of ants (Hymenoptera: Formicidae) and their effect on subordinate species. European Journal of Entomology, 110(2), 327.

D'Ettorre, P., Heinze, J., Schulz, C., Francke, W., & Ayasse, M. (2004). Does she smell like a queen? Chemoreception of a cuticular hydrocarbon signal in the ant Pachycondyla inversa. Journal of Experimental Biology, 207(7), 1085-1091. https://doi.org/10.1242/jeb.00865

79

D’Ettore, P., Lenoir, A. (2010). Nestmate Recognition. In L. Lach, C. Parr, K. Abbott (Eds.), Ant Ecology (pp. 194-209). New York, NY: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199544639.001.0001

Devictor, V., Clavel, J., Julliard, R., Lavergne, S., Mouillot, D., Thuiller, W., … Mouquet, N. (2010). Defining and measuring ecological specialization. Journal of Applied Ecology, 47(1), 15. Retrieved from http://dx.doi.org/10.1111/j.1365- 2664.2009.01744.x

Diez, L., Moquet, L., & Detrain, C. (2013). Post-mortem changes in chemical profile and their influence on corpse removal in ants. Journal of chemical ecology, 39(11-12), 1424-1432. di Mauro, G., Perez, M., Lorenzi, M. C., Guerrieri, F. J., Millar, J. G., & d’Ettorre, P. (2015). Ants Discriminate Between Different Hydrocarbon Concentrations. Frontiers in Ecology and Evolution, 3, 133. doi:10.3389/fevo.2015.00133

Dicke, M., & Grostal, P. (2001). Chemical Detection of Natural Enemies by Arthropods: An Ecological Perspective. Annual Review of Ecology and Systematics, 32(1), 1. Retrieved from http://dx.doi.org/10.1146/annurev.ecolsys.32.081501.113951

Dobzhansky, T. (1950). Evolution in the Tropics. American Scientist 38, 209-221.

Dornhaus, A., & Powell, S. (2010). Foraging and defense strategies. Ant ecology, 210- 230.

Drijfhout, F. P., Kather, R., & Martin, S. J. (2010). The role of cuticular hydrocarbons in insects. Behavioral and chemical ecology, 91-114.

Endler JA. (1991). Interactions between predators and prey. In J. R. Krebs & N. B. Davies (Eds.) Behavioural ecology: an evolutionary approach (pp. 169–196), 3rd edn. Oxford, UK: Blackwell Scientific. https://doi.org/10.2307/5530

Fagerland, M. W., & Sandvik, L. (2009). Performance of five two-sample location tests for skewed distributions with unequal variances. Contemporary clinical trials, 30(5), 490-496.

Fagerland, M. W. (2012). t-tests, non-parametric tests, and large studies—a paradox of statistical practice?. BMC Medical Research Methodology, 12(1), 78.

Finlay-Doney, M., & Walter, G. H. (2012). The conceptual and practical implications of interpreting diet breadth mechanistically in generalist predatory insects. Biological Journal of the Linnean Society, 107, 737–763. Retrieved from https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1095-8312.2012.01991.x

Forister, M. L., Dyer, L. A., Singer, M. S., Stireman III, J. O., & Lill, J. T. (2012).

80

Revisiting the evolution of ecological specialization, with emphasis on insect–plant interactions. Ecology, 93(5), 981–991. Retrieved from http://dx.doi.org/10.1890/11- 0650.1

Fowler, H.G., V. Pereira-da-Silva, L.C. Forti and N.B. Saes, (1986). Populations dynamics of leaf-cutting ants: a brief review. In C.S. Lofgren & R.K. Vander Meer (Eds.) Fire Ants and Leaf-Cutting Ants: Biology and Management (123–145 pp). Westview Press. Boulder, USA.

Franks, N. R. (1980). The evolutionary ecology of the army ant Eciton burchelli on Barro Colorado Island, Panama, Ph.D. dissertation, The University of Leeds, Leeds, England.

Franks, N. R. (1982). A new method for censusing animal populations: the number of Eciton burchelli army ant colonies on Barro Colorado Island, Panama. Oecologia, 52(2), 266-268. https://doi.org/10.1007/bf00363847

Franks, N. R. (1983). Ecology and population regulation in the army ant Eciton burchelli. In The ecology of a tropical forest: Seasonal rhythms and long-term changes (ed. E. G. Leigh, Jr., A. S. Rand & D. M. Windsor), pp. 389-395. Oxford: Oxford University Press

Franks, N. R. (2001). Evolution of mass transit systems in ants: a tale of two societies. In SYMPOSIUM-ROYAL ENTOMOLOGICAL SOCIETY OF LONDON, 20, 281- 298).

Franks, N. R., & Bossert, W. H. (1983). The influence of swarm raiding army ants on the patchiness and diversity of a tropical leaf litter ant community. Tropical Rain Forest: Ecology and Management. Special Publication of the British Ecological Society, 2, 151–163.

Fuller, R. C., Houle, D., & Travis, J. (2005). Sensory bias as an explanation for the evolution of mate preferences. The American Naturalist, 166(4), 437-446. https://doi.org/10.1086/444443

Futuyma, D. J., & Moreno, G. (1988). The Evolution of Ecological Specialization. Annual Review of Ecology and Systematics, 19(1), 207–233. Retrieved from http://arjournals.annualreviews.org/doi/abs/10.1146%2Fannurev.es.19.110188.0012 31

Gibbs, A. G. (1998). Water-Proofing Properties of Cuticular Lipids 1. American Zoologist, 38, 471–482. Retrieved from https://academic.oup.com/icb/article- abstract/38/3/471/141617

Godfray, H. C. J. (1994). Parasitoids: behavioral and evolutionary ecology. Princeton University Press.

81

Gotelli, N, J., Edmund M. Hart and Aaron M. Ellison (2015) EcoSimR: Null model analysis forecological data. R package version 0.1.0. http://github.com/gotellilab/EcoSimR DOI: 10.5281/zenodo.16522

Gotwald, W. H., Jr. 1995. Army ants: the biology of social predation. Ithaca, NY: Cornell University Press/Comstock Press.

Grant, P.R. 1999. Ecology and Evolution of Darwin’s Finches. Princeton, NJ: Princeton University Press.

Greenstone, M. H. & Dickens, J. C. (2005). The production and appropriation of chemical signals among plants, herbivores and predators. In: Ecology of Predator– Prey Interactions (Barbosa, P. & Castellanos, I., eds). Oxford Univ. Press, New York, pp. 139-165.

Grover, J. P. (1997). Resource Competition. Chapman and Hall, London.

Hardin, G. (1960). The competitive exclusion principle. Science, 131(3409), 1292-1297. https://doi.org/10.1126/science.131.3409.1292

Hashimoto, Y., & Yamane, S. (2014). Comparison of foraging habits between four sympatric army ant species of the genus Aenictus in Sarawak, Borneo. Asian Myrmecology, 6(1), 95–104.

Hawlena, D., & Schmitz, O. J. (2010). Physiological stress as a fundamental mechanism linking predation to ecosystem functioning. The American Naturalist, 176(5), 537- 556.

Hill, P. S. (2008). Vibrational communication in animals. Cambridge, Massachusetts: Harvard University Press.

Hill, P. S. (2009). How do animals use substrate-borne vibrations as an information source? Naturwissenschaften, 96(12), 1355-1371. https://doi.org/10.1007/s00114- 009-0588-8

Hixon, M. A. (1980). Competitive interactions between California reef fishes of the genus Embiotoca. Ecology, 61(4), 918-931.

Hölldobler, B. (1983). Territorial behavior in the green tree ant (Oecophylla smaragdina). Biotropica 15: 241–250.

Hölldobler, B. & Wilson, E. O. (1990). The ants. Harvard University Press, Cambridge, Mass.

Holman, L., Jørgensen, C. G., Nielsen, J., & d'Ettorre, P. (2010). Identification of an ant queen pheromone regulating worker sterility. Proceedings of the Royal Society of

82

London B: Biological Sciences, 277(1701), 3793-3800.

Holt, R. D. (2009). Bringing the Hutchinsonian niche into the 21st century: Ecological and evolutionary perspectives. Proceedings of the National Academy of Sciences, 106(Supplement_2), 19659. Retrieved from http://dx.doi.org/10.1073/pnas.0905137106

Howard, D. F., & Tschinkel, W. R. (1976). Aspects of necrophoric behavior in the red imported fire ant, Solenopsis invicta. Behaviour, 56(1), 157-178. https://doi.org/10.1163/156853976x00334

Hurlbert, A. H., Ballantyne, F., & Powell, S. (2008). Shaking a leg and hot to trot: the effects of body size and temperature on running speed in ants. Ecological Entomology, 33(1), 144-154.

Hutchinson, G.E. (1957). Concluding remarks. Cold Spring Harbor Symposia on Quantitative Biology, 22, 145–159.

Hutchinson, G. E. (1959). Homage to Santa Rosalia or Why Are There So Many Kinds of Animals? The American Naturalist, 93(870), 145. Retrieved from http://dx.doi.org/10.1086/282070

Irschick, D., Dyer, L., & Sherry, T. W. (2005). Phylogenetic Methodologies for Studying Specialization. Oikos, 110(2), 404–408. Retrieved from file:///Oikos/Phylogenetic Methodologies for Studying Specialization - 2005.pdf

Kaspari, M., Lattke, J., O’Donnell, S., Kaspari, M., Powell, S., Lattke, J., & O’Donnell, S. (2011). Predation and patchiness in the tropical litter: do swarm-raiding army ants skim the cream or drain the bottle? Journal of Animal Ecology, 80(4), 818–823. Retrieved from http://doi.wiley.com/10.1111/j.1365-2656.2011.01826.x

Kelber, C., Rössler, W., & Kleineidam, C. J. (2010). Phenotypic plasticity in number of glomeruli and sensory innervation of the antennal lobe in leaf‐cutting ant workers (A. vollenweideri). Developmental neurobiology, 70(4), 222-234. https://doi.org/10.1002/dneu.20782

Kneitel, J. M., & Chase, J. M. (2004). Disturbance, predator, and resource interactions alter container community composition. Ecology, 85(8), 2088-2093.

Lach, L., Parr, C., & Abbott, K. (Eds.). (2009). Ant Ecology. Oxford University Press. Retrieved from http://dx.doi.org/10.1093/acprof:oso/9780199544639.001.0001

LaPolla, J.S., Mueller, U.G., Seid, M. & Cover, S.P. (2002) Predation by the army ant Neivamyrmex rugulosus on the fungus-growing ant Trachymyrmex arizonensis. Insectes Sociaux, 49, 251–256. https://doi.org/10.1007/s00040-002-8310-2

83

Leigh, E. G. (1999). Tropical forest ecology: a view from Barro Colorado Island. New York, NY: Oxford University Press.

Lenoir, A., d'Ettorre, P., Errard, C., & Hefetz, A. (2001). Chemical ecology and social parasitism in ants. Annual review of entomology, 46(1), 573-599.

Leonhardt, S. D., Menzel, F., Nehring, V., & Schmitt, T. (2016). Ecology and evolution of communication in social insects. Cell, 164(6), 1277-1287. https://doi.org/10.1016/j.cell.2016.01.035

Longino, J. T., & Nadkarni, N. M. (1990). A Comparison of Ground and Canopy Leaf Litter Ants (Hymenoptera: Formicidae) in a Neotropical Montane Forest. Psyche: A Journal of Entomology, 97(1–2), 81–93.

Longino, J.T. (2010). Ants Of Costa Rica. http://ants.biology.utah.edu/~longino/AntsofCostaRica.html, retrieved 12. January 2019

Losos, J. B. (1990). Concordant evolution of locomotor behaviour, display rate and morphology in Anolis lizards. Animal Behaviour, 39(5), 879-890.

Mabelis, A. A. (1979). Wood Ant Wars the Relationship Between Aggression and Predation in the Red Wood Ant (Formica Polyctena Forst.) by. Netherlands Journal of Zoology, 29(4), 451-620.

MacArthur, R. H. (1958). Population ecology of some warblers of northeastern coniferous forests. Ecology, 39(4), 599-619. https://doi.org/10.2307/1931600

MacArthur, R. H. (1969). Patterns of communities in the tropics. Biological Journal of the Linnean Society, 1(1-2), 19-30.

MacArthur, R. H. (1972). Geographical ecology: patterns in the distribution of species. Princeton, NJ: Princeton University Press.

Markó, B., Erős, K., Czechowski, W., Maák, I., Babik, H., & Ślipiński, P. (2014). Cues or meaningless objects? Differential responses of the ant Formica cinerea to corpses of competitors and enslavers. Animal Behaviour, 91, 53–59. doi:10.1016/j.anbehav.2014.02.014

Matsuda, H., Hori, M., & Abrams, P. A. (1996). Effects of predator-specific defence on biodiversity and community complexity in two-trophic-level communities. Evolutionary ecology, 10(1), 13-28.

McCormick, A. C., Unsicker, S. B., & Gershenzon, J. (2012). The specificity of herbivore-induced plant volatiles in attracting herbivore enemies. Trends in plant science, 17(5), 303-310. https://doi.org/10.1016/j.tplants.2012.03.012

84

Michener C. D. (2007). The bees of the world (2nd ed.). Baltimore, MD: The Johns Hopkins University Press.

Meyers, J. J., & Herrel, A. (2005). Prey capture kinematics of ant-eating lizards. Journal of Experimental Biology, 208(1), 113-127.

Mooney, K. A., Gruner, D. S., Barber, N. A., Van Bael, S. A., Philpott, S. M., & Greenberg, R. (2010). Interactions among predators and the cascading effects of vertebrate insectivores on arthropod communities and plants. Proceedings of the National Academy of Sciences, 107(16), 7335-7340.

Mysore, K., Shyamala, B. V., & Rodrigues, V. (2010). Morphological and developmental analysis of peripheral antennal chemosensory sensilla and central olfactory glomeruli in worker castes of Camponotus compressus (Fabricius, 1787). Arthropod structure & development, 39(5), 310-321. https://doi.org/10.1016/j.asd.2010.04.003

Nakanishi, A., Nishino, H., Watanabe, H., Yokohari, F., & Nishikawa, M. (2009). Sex‐ specific antennal sensory system in the ant Camponotus japonicus: Glomerular organizations of antennal lobes. Journal of Comparative Neurology, 518(12), 2186- 2201. https://doi.org/10.1002/cne.22326

Nelson, M. E., & Maciver, M. A. (1999). Prey capture in the weakly electric fish Apteronotus albifrons: sensory acquisition strategies and electrosensory consequences. Journal of Experimental Biology, 202(10), 1195-1203.

Nonacs, P., & Dill, L. M. (1990). Mortality risk vs. food quality trade‐offs in a common currency: ant patch preferences. Ecology, 71(5), 1886-1892.

Nowbahari, e., Lenoir, a., clément, J. L., Lange, c., bagnères, a. G. and Joulie, c. (1990). Individual, geographical and experimental variation of cuticular hydrocarbons of the ant Cataglyphis cursor (hymenoptera: Formicidae): their use in nest and subspecies recognition. Biochem. Syst. Ecol., 18, 63–73.

O’Donnell, S., Kaspari, M. & Lattke, J. (2005) Extraordinary Predation by the Neotropical Army Ant Cheliomyrmex andicola: Implications for the Evolution of the Army Ant Syndrome. Biotropica, 37, 706–709. https://doi.org/10.1111/j.1744- 7429.2005.00091.x

O’Donnell, S., Lattke, J., Powell, S., & Kaspari, M. (2007). Army ants in four forests: geographic variation in raid rates and species composition. Journal of Animal Ecology, 76(3), 580–589. Retrieved from http://doi.wiley.com/10.1111/j.1365- 2656.2007.01221.x

Otis, G. W., Santana, E., Crawford, D. L., & Higgins, M. L. (1986). The effect of foraging army ants on leaf-litter arthropods. Biotropica, 56-61.

85

Pacala, S. W., & Roughgarden, J. (1985). Population experiments with the Anolis lizards of St. Maarten and St. Eustatius. Ecology, 66(1), 129-141. https://doi.org/10.2307/1941313

Parr, C. L., & GIBB, H. (2010). Dominance hierarchies. In C. L. Parr, L. Lach, and K. Abbot (Eds.). Ant ecology, pp. 33–35. Oxford University Press, Oford.

Pasch, B., Bolker, B. M., & Phelps, S. M. (2013). Interspecific dominance via vocal interactions mediates altitudinal zonation in neotropical singing mice. The American Naturalist, 182(5), E161-E173.

Pask, G. M., Slone, J. D., Millar, J. G., Das, P., Moreira, J. A., Zhou, X., … Ray, A. (2017). Specialized odorant receptors in social insects that detect cuticular hydrocarbon cues and candidate pheromones. Nature Communications, 8(1). Retrieved from http://dx.doi.org/10.1038/s41467-017-00099-1

Pianka, E. R. (1974). Niche Overlap and Diffuse Competition. Proceedings of the National Academy of Sciences, 71(5), 2141. Retrieved from http://10.0.4.49/pnas.71.5.2141

Poisot, T., Bever, J. D., Nemri, A., Thrall, P. H., & Hochberg, M. E. (2011). A conceptual framework for the evolution of ecological specialisation. Ecology Letters, 14(9), 841–851. Retrieved from http://doi.wiley.com/10.1111/j.1461- 0248.2011.01645.x

Polis, G. A., Myers, C. A., & Holt, R. D. (1989). The ecology and evolution of intraguild predation: potential competitors that eat each other. Annual review of ecology and systematics, 20(1), 297-330.

Powell, S. (2005) Polymorphism and Ecology in the New World Army Ant Genus Eciton. Ph.D. dissertation, University of Bristol, Bristol, England.

Powell, S. (2011). How much do army ants eat? On the prey intake of a neotropical top- predator. Insectes Sociaux, 58(3), 317–324. Retrieved from http://link.springer.com/10.1007/s00040-011-0152-3

Powell, S., & Baker, B. (2008). The hidden big predators of the Neotropics : The behaviour , diet , and impact of New World army ants ( Ecitoninae ). Insetos Sociais Da Biologia à Aplicação, 18–37.

Powell, S., & Clark, E. (2004). Combat between large derived societies: A subterranean army ant established as a predator of mature leaf-cutting ant colonies. Insectes Sociaux, 51(4), 342–351. doi:10.1007/s00040-004-0752-2

Powell, S., & Franks, N. R. (2006). Ecology and the evolution of worker morphological diversity: a comparative analysis with Eciton army ants. Functional Ecology, 20(6),

86

1105. Retrieved from http://dx.doi.org/10.1111/j.1365-2435.2006.01184.x

Raine, N. E., & Chittka, L. (2007). The adaptive significance of sensory bias in a foraging context: floral colour preferences in the bumblebee Bombus terrestris. PLoS One, 2(6), e556. https://doi.org/10.1371/journal.pone.0000556

Rana, J. S., Dixon, A. F. G., & Jarošík, V. (2002). Costs and benefits of prey specialization in a generalist insect predator. Journal of Animal Ecology, 71(1), 15- 22. https://doi.org/10.1046/j.0021-8790.2001.00574.x

Ravigné, V., Dieckmann, U., & Olivieri, I. (2009). Live Where You Thrive: Joint Evolution of Habitat Choice and Local Adaptation Facilitates Specialization and Promotes Diversity. The American Naturalist, 174(4), E141–E169. doi:10.1086/605369

Rettenmeyer, C. W. (1963). Behavioral studies of army ants. University of Kansas Science Bulletin 44, 281-465.

Rettenmeyer, C. W., Chadab-Crepet, R., Naumann, M. G., & Morales, L. (1983). Comparative foraging in Neotropical army ants. In P. Jaison (Ed.), Social insects in the tropics: Proceedings of the first international symposium organized by the International Union for the Study of Social Insects and the Sociedad Méxicana de Entomologia Université Paris-Nord (pp. 59–73).

Sapp, J. (2016). Coexistence: The Ecology and Evolution of Tropical Biodiversity. Oxford, UK: Oxford University Press.

Schatz, B., Anstett, M. C., Out, W., & Hossaert-McKey, M. (2003). Olfactive detection of fig wasps as prey by the ant Crematogaster scutellaris (Formicidae; Myrmicinae). Naturwissenschaften, 90(10), 456-459. https://doi.org/10.1007/s00114-003-0457-9

Schatz, B., & Hossaert-McKey, M. (2010). Ants use odour cues to exploit fig–fig wasp interactions. Acta Oecologica, 36(1), 107-113. https://doi.org/10.1016/j.actao.2009.10.008

Schemske, D. W., Mittelbach, G. G., Cornell, H. V., Sobel, J. M., & Roy, K. (2009). Is there a latitudinal gradient in the importance of biotic interactions?. Annu. Rev. Ecol. Evol. Syst., 40, 245-269.

Schluter, D. (2000). Ecological character displacement in adaptive radiation. The american naturalist, 156(S4), S4-S16.

Schmidt, C.A. & Shattuck, S.O. (2014). The higher classification of the ant subfamily Ponerinae (Hymenoptera: Formicidae), with a review of ponerine ecology and behavior. Zootaxa. 3817, 1–242. doi:10.11646/zootaxa.3817.1.1

87

Schneider, C. A.; Rasband, W. S. & Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of image analysis. Nature methods 9 (7), 671-675.

Schneirla, T. C. (1933). Studies on army ants in Panama. Journal of Comparative Psychology, 15(2), 267. https://doi.org/10.1037/h0071685

Schneirla, T. C. (1934). Raiding and other outstanding phenomena in the behavior of army ants. Proceedings of the National Academy of Sciences, 20(5), 316-321.

Schneirla, T. C. 1971. Army ants: a study in social organization. San Francisco, CA: W. H. Freeman.

Schnitzler, H. U., Kalko, E., Miller, L., & Surlykke, A. (1987). The echolocation and hunting behavior of the bat, Pipistrellus kuhli. Journal of Comparative Physiology A, 161(2), 267-274.

Schoener, T. W. (1974). Resource partitioning in ecological communities. Science, 185(4145), 27-39.

Schoener, T. W. (1986). Mechanistic Approaches to Community Ecology: A New Reductionism?’. American Zoologist, 26, 81–106. Retrieved from https://academic.oup.com/icb/article-abstract/26/1/81/114045

Sharma, K. R., Enzmann, B. L., Schmidt, Y., Moore, D., Jones, G. R., Parker, J., … Ray, A. (2015). Cuticular Hydrocarbon Pheromones for Social Behavior and Their Coding in the Ant Antenna. Cell Reports, 12(8), 1261. Retrieved from http://dx.doi.org/10.1016/j.celrep.2015.07.031

Siemers, B. M., & Schnitzler, H. U. (2004). Echolocation signals reflect niche differentiation in five sympatric congeneric bat species. Nature, 429(6992), 657. https://doi.org/10.1038/nature02547

Siemers, B. M., Goerlitz, H. R., Robsomanitrandrasana, E., Piep, M., Ramanamanjato, J. B., Rakotondravony, D., ... & Ganzhorn, J. U. (2007). Sensory basis of food detection in wild Microcebus murinus. International Journal of Primatology, 28(2), 291. https://doi.org/10.1007/s10764-007-9135-7

Siemers, B. M., & Swift, S. M. (2006). Differences in sensory ecology contribute to resource partitioning in the bats Myotis bechsteinii and Myotis nattereri (Chiroptera: Vespertilionidae). Behavioral Ecology and Sociobiology, 59(3), 373-380.

Sih, A., & Christensen, B. (2001). Optimal diet theory: when does it work, and when and why does it fail?. Animal behaviour, 61(2), 379-390.

Sih, A., Bell, A., & Johnson, J. C. (2004). Behavioral syndromes: an ecological and

88

evolutionary overview. Review TRENDS in Ecology and Evolution, 19(7). doi:10.1016/j.tree.2004.04.009

Singer, M. S., & Stireman, J. O. (2005). The tri‐trophic niche concept and adaptive radiation of phytophagous insects. Ecology Letters, 8(12), 1247-1255. https://doi.org/10.1111/j.1461-0248.2005.00835.x

Slone, J., Pask, G., Ferguson, S., Millar, J., Berger, S., Reinberg, D., … Zwiebel, L. (2017). Functional characterization of odorant receptors in the ponerine ant, Harpegnathos saltator. Proceedings of the National Academy of Sciences, 114(32), 8586–8591. doi:10.1073/pnas.1704647114

Steidle, J. L. M., & van Loon, J. J. A. (2003). Dietary specialization and infochemical use in carnivorous arthropods: testing a concept. Entomologia Experimentalis et Applicata, 108(3), 133. Retrieved from http://dx.doi.org/10.1046/j.1570- 7458.2003.00080.x

Stevens M (2013). Sensory ecology, behaviour, and evolution. Oxford, UK: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199601776.001.0001

Surlykke, A., & Moss, C. F. (2000). Echolocation behavior of big brown bats, Eptesicus fuscus, in the field and the laboratory. The Journal of the Acoustical Society of America, 108(5), 2419-2429.

Tuttle, M. D., Ryan, M. J., & Belwood, J. J. (1985). Acoustical resource partitioning by two species of phyllostomid bats (Trachops cirrhosus and Tonatia sylvicola). Animal Behaviour, 33(4), 1369-1371.

Van Zweden JS, D’Ettorre P. (2010). Nestmate recognition in social insects and the role of hydrocarbons. In G. Blomquist & A. Bagneres (Eds.) Insect hydrocarbons: biology, biochemistry, and chemical ecology (pp. 222-243). Cambridge, UK: Cambridge University Press. https://doi.org/10.1017/cbo9780511711909

Vet, L. E. M., & Dicke, M. (1992). Ecology if info chemical use by natual enemies in a tritrphic context. Annual Review of Entomology, 37, 141–72. Retrieved from www.annualreviews.org

Vidal‐Riggs, J. M., & Chaves‐Campos, J. (2008). Method review: estimation of colony densities of the army ant Eciton burchellii in Costa Rica. Biotropica, 40(2), 259-262. https://doi.org/10.1111/j.1744-7429.2007.00347.x

Von der Emde, G., & Bleckmann, H. (1998). Finding food: senses involved in foraging for insect larvae in the electric fish Gnathonemus petersii. Journal of Experimental Biology, 201(7), 969-980.

Wetterer, J. K. (1993). Foraging and nesting ecology of a Costa Rican leaf-cutting

89

ant, Acromyrmex volcanus. Psyche, 100, 65-76.

Wheeler, W. M. (1910). Ants: Their Structure, Development and Behavior. Columbia University Press, New York.

Wheeler, G. C., & Wheeler, J. (1976). Ant larvae: review and synthesis (No. 7). Washington, DC: Entomological Society of Washington.

Wilson, E. O., Durlach, N. I., & Roth, L. M. (1958). Chemical releasers of necrophoric behavior in ants. Psyche, 65(4), 108-114. https://doi.org/10.1155/1958/69391

Wilson, E. O. (1959). Some ecological characteristics of ants in New Guinea rain forest. Ecology 40, 437-447.

Winemiller K O, and Pianka E R. (1990). Organization in natural assemblages of desert lizards and tropical fishes. Ecological Monographs 60, 27 – 55.

Winston, M. E., Kronauer, D. J., & Moreau, C. S. (2017). Early and dynamic colonization of Central America drives speciation in Neotropical army ants. Molecular ecology, 26(3), 859-870.

Xue, H. J., Wei, J., Huang, Z. Z., Li, W. Z., & Yang, X. K. (2018). Your chemical coat tells me you are my delicacy: a predatory stink bug uses cuticular hydrocarbons to identify prey. Chemoecology, 1-5.

Yusuf, A. A., Crewe, R. M., & Pirk, C. W. (2014). Olfactory detection of prey by the termite-raiding ant Pachycondyla analis. Journal of insect science, 14(53), 1-10. https://doi.org/10.1673/031.014.53

Yusuf, A. A., Gordon, I., Crewe, R. M., & Pirk, C. W. W. (2014). Prey choice and raiding behaviour of the Ponerine ant Pachycondyla analis (Hymenoptera: Formicidae). Journal of Natural History, 48(5-6), 345-358. https://doi.org/10.1080/00222933.2013.791931

Zhang, Jinlong. (2016). spaa: SPecies Association Analysis. R package version 0.2.2. https://CRAN.R-project.org/package=spaa

Ziegler, C., & Leigh, E. G. (2011). A Magic Web: The Tropical Forest of Barro Colorado Island. Washington, DC: Smithsonian Institution Scholarly Press.

90