國立臺灣師範大學理學院

生命科學系 碩士論文

Department of Life Science College of Science National Taiwan Normal University Master’s Thesis

雄性二型雞冠細身赤鍬形蟲之異速生長與打鬥評估策略 The allometry and fighting assessment strategy of male dimorphic Cyclommatus mniszechi

陳震邑 Chen, Zhen-yi

指導教授 Advisor：林仲平 博士 Ph.D.

中華民國 109 年 2 月

February 2020

致謝

感謝我的指導教授林仲平老師在這三年多來的教導、支持與鼓勵，

讓這篇研究可以成形，並依序架構出實驗架構的輪廓，幫助我在研

究之路上可以理出下一步的方向。同時也感謝許鈺鸚老師與打鬥行

為實驗室的各位成員，在每次的實驗室會議中對於實驗設計與統計

分析方法的建議與在實驗空間的協助，提供理論與實務上的建議，

讓打鬥行為實驗可以順利的完成並分析出結果。感謝親愛的實驗室

夥伴們與各位朋友們，陪我進行野外採集、蒐集樣本、實驗室飼

養、操作實驗，也陪伴我在這段時間學習以外的時候可以盡情的休

閒，排解壓力，抒發心情，最後感謝我摯愛的家人，在求學的這段

時間給予物質與精神上的支持，讓我可以專心地完成這份論文。

I

摘要

雄性鍬形蟲具有誇張的大顎作為打鬥時獲取食物與繁殖資源的武

器，大顎武器的形狀與大小可能會影響打鬥的行為與勝負。本研究

檢測雞冠細身赤鍬形蟲(Cyclommatus mniszechi)的異速生長關係、打

鬥行為序列和打鬥評估策略。首先，大顎與體型之間的異速生長關

係的判定可以用來作為決定不同型雄蟲的基礎，再者，藉由行為序

列分析來描述依大顎大小隨機配對與對等配對的雄蟲間打鬥的行為

序列，最後，從打鬥時間(和激烈程度)與資源佔有潛力(RHP, resource holding potential)的相關性來檢驗不同評估策略(自我、累加

和相互評估策略)的預期。異速生長的分析結果顯示大顎長與體長之

間存在非線性正異速生長關係，且在體長(翅鞘長)為 15.03 毫米時

為異速生長係數改變的轉折點，可以此將鍬形蟲個體分為大型與小

型雄蟲。從打鬥行為序列中可以定義出九種行為單元，包含「觸

碰」、「防衛姿勢」、「抬身對峙」、「攻擊」、「推擊」、「纏鬥」、「抓取

1」、「抓取 2」(抓取於對手胸節或腹節)以及「撤退」。大小型雄蟲

具有不同的打鬥行為序列，小型雄蟲傾向持續待在相同打鬥階段(較

多在同一個打鬥階段內的行為轉換)，相對於大型雄蟲則較容易進入

纏鬥階段。大顎長對於打鬥結果有決定性的影響，可以做為代表雞

冠細身赤鍬形蟲的資源佔有潛力相關的可靠特徵。在隨機配對打鬥

II

中，打鬥時間與勝者及敗者的大顎長有顯著正相關，顯示自我評估

可能為此種鍬形蟲所採用的打鬥評估策略。在對等配對打鬥中，打

鬥時間與平均大顎長有顯著正相關，進一步支持鍬形蟲採用自我評

估策略。然而，雄蟲在打鬥中出現對手之間的對等行為，呈現多次

但很少造成傷害的身體接觸，與單方向的行為階段進程(由低度至高

度激烈且鮮少逆向的打鬥階段)。因此，雞冠細身赤鍬形蟲的打鬥可

能不僅使用自我評估策略，累加與相互評估策略可能也在其打鬥決

策中扮演重要角色。

關鍵字：體型、雄性競爭、資源佔有潛力、行為序列分析、性擇、

武器

III

Abstract Male stag beetles (Coleoptera: Lucanidae) possessed exaggerated mandibles as weapons used in fighting contests for access to food and reproduction. The shapes and sizes of these mandibular weapons could influence the fighting behaviours and outcomes of the contests. This study examined the allometry, fighting behaviour and assessment strategy of a stag beetle, Cyclommatus mniszechi. Firstly, the allometric relationships between mandible and body sizes were identified to determine whether the males could be grouped into different morphs based on the allometries. Secondly, the behavioural sequences of male- male fights were characterized using sequential analyses of randomly and size-matched contests. Finally, the correlational predictions between contest duration (and aggressiveness) and RHP (resource holding potential) were examined to test alternative assessment strategies (self-, cumulative and mutual assessment). Allometric analyses show a non- linear positive allometry between mandible and body size in C. mnizechi males, and that they consist of dimorphic males defined quantitatively as the majors and minors by body sizes at the switch point of elytra length of 15.03 mm. Nine behavioural elements were identified from the contests, including ‘touch’, ‘defensive posture’, ‘body raising’ ‘attack’, ‘push’, ‘tussle’, ‘clamp1’(head), ‘clamp2’ (thorax or abdomen) and ‘retreat’. The major and minor males have different fighting behavioural sequences, where the minor males tend to stay within phases (more behavioural transitions within phases) of the contests and more likely to tussle than the major males. Mandible size is the main determinant of the outcomes

IV of the contests and can be used as a reliable proxy for RHP in C. mnizechi. In randomly matched contests, strong positive relations between contest duration and winner’s and loser’s mandible sizes indicate that self-assessment determines strategic decisions in C. mniszechi. In size-matched contests, a positive relation between contest duration and mandible sizes further support the self-assessment strategy. However, males showed behavioural matching in contests, many physical contacts with rare injuries and unidirectional behavioural progressions in phases (from low towards high aggression with rare de-escalation). Therefore, the fighting contests of C. mniszechi may not settle entirely on the basis of pure self-assessment, and that cumulative or mutual assessment may also play an important role in contest decisions.

Key words: Body size, male-male competition, resource holding potential, sequential analysis, sexual selection, weapon

V

Ta ble of Contents Introduction………………………………………………………………1 Materials and methods………………………………….……………...... 9 Study organism………………………………………………………9 Insect rearing……………………………………………………10 Morphological measurements………………………………………11 Allometry analyses………………..………………………………12 Fighting contests………………………………………………… 14 Determination of traits associated with RHP.………………………15 Sequential analyses of fighting behaviours…………………….…16 Statistical analyses of RHP, aggression and contest duration………17 Results……………………………….………………………………...18 Morphological measurement and allometric analyses…………….18 Behavioral sequence of fighting contests…………………………19 Male morph and contest aggression………………………….…..23 Contest duration and RHP………………………………………...24 Discussion…………….…………………………………………………24 Dimorphic males, allometry and fighting behaviours……………….24 Self and mutual assessment strategy…………………………….....27 Probable mechanisms of assessment in C. mniszechi stag beetles…30 References…………………………………………………………….33

VI

List of Tables Table 1. Behavioural elements of the contest interaction………………..39 Table 2. Correlation of the relationships between morphological traits….40 Table 3. Models selection of the allometry analyses……………………..41 Table 4. Parameters estimated for the isometry tests……………….……42 Table 5. Relationships between aggressiveness and morphological traits in randomly matched contests……………………………………………...43 Table 6. the correlation between contest duration and male morphs and average mandible length in the size-matched contests…………………44

VII

List of Figures Figure 1. Mate-guarding, male morphs and morphological measurements of Cyclommatus mniszechi ………………………………………….…45 Figure 2. The allometric relationship between ln-transformed mandible length and elytra length………………………………………………..46 Figure 3. The fighting arena of Cyclommatus mniszechi……………… 47 Figure 4. Sequential analyses of contest behaviours in randomly and size- matched contests……………………………………………………….48 Figure 5. Ethograms and sequential analyses of behabvioural elements of major and minor male of Cyclommatus mniszechi…………………..……..49 Figure 6. Logistic regressions of contest behaviours…………………….50 Figure 7. Simple linear regressions of contest behaviours………………51

VIII

List of Supplementary Information Table S1. The information of Cyclommatus mniszechi specimens………52

IX

Introduction

Male of many insects use weapons to compete for resource essential for their survival and reproduction (Emlen 2008b). Sexually selected traits such as ornaments and weapons often evolve to reach greater extremes of size and elaboration (Andersson 1994; Darwin 1871; Emlen 2008a; Emlen and Nijhout 2000). Males of many insects use enlarged weapons to compete for their survival and reproduction (Emlen 2008a). One group of insects with an amazing diversity of weapons are stag beetles (Lucanidae). Male stag beetles use their exaggerated mandibles as weapons to fight for the access to the sap site, territory and females (Emlen 2008a; Goyens et al. 2015a; Goyens et al. 2015b; Inoue and Hasegawa 2012; Songvorawit et al. 2018). Mandible and body size are reliable indicators for male stag beetle’s resource holding potential (RHP) and positively correlate with their ability to win the contests (Goyens et al. 2015a; Kuan 2011; Songvorawit et al. 2018). Stag beetles exhibit a high diversity in the sizes and shapes of mandibles both within and across species (Emlen 2008a; Kawano 2000; Mizunuma and Nagai 1994; Shiokawa and Iwahashi 2000), which may cause the behaviours they display in fighting to be highly variable (Emlen 2008a). The scaling relationship or allometry of trait size and body size is informative for many important aspects of organism’s biology such as polyphenism (Rowland and Emlen 2009), alternative reproductive tactics (Tomkins et al. 2005) and nature of selection on different body parts (Bonduriansky 2007; Kojima and Lin 2017). The allometry of sexually selected traits can be positive (slope > 1), isometry (slope = 1) or negative

1

(slope <1), and often reflects resource-allocation trade-offs, balance between natural and sexual selection, and genetic constraints of these traits (Bonduriansky 2007; Bonduriansky and Day 2003). The mandible and body size in many stag beetles show a linear and positive allometry, in which larger males have relatively larger mandibles than those of smaller males (Clark 1977; Huxley 1931; Kawano 2000; Knell et al. 2004; Romiti et al. 2015). The positive allometry is explained by greater fitness advantages for larger males to invest proportionally more in the development of weapons versus other body parts (Emlen and Nijhout 2000). Nonlinear (curved, sigmoid or discontinuous) allometric relationships between mandible and body size have also been reported for stag beetles (Emlen and Nijhout 2000; Kawano 2000; Knell et al. 2004). The declined slope of the curved allometry in the largest males is likely due to the depletion of resources available for the development of the weapon, resulting from the competition of imaginal disc within closed pupal chamber (Knell et al. 2004; Kojima and Lin 2017; Nijhout and Wheeler 1996). The sigmoid and discontinuous allometries are characterized with the evolution of size-dependent, alternative reproductive tactics (Andersson 1994). For examples, in dung beetles the larger and horned males use a mate-guarding and fighting strategy against opponents, whereas the smaller and hornless males employ a sneaking tactic for access to females in guarded tunnels (Emlen 1997; Moczek and Emlen 2000).The non-linear allometries between trait size and body size can be characterized following the approach and regression models recommended by Knell 2009 (Knell 2009).The nature of fighting

2 behaviour s is essential to understand the evolution and diversification of weapon forms (Emlen 2008a). How male stag beetles interact with and respond to each other in contests, however, remains poorly understood. For the majority of stag beetle species, it was unclear how males use their mandibular weapons in combats and whether males with different weapon sizes and shapes display different contest behaviours. There are approximately 1200 species of stag beetles in the world (Smith 2006). So far, the fighting behaviour has only been reported for five of the species, including Cyclommatus miniszechi(Kuan 2011), Cyclommatus metallifer, (Goyens et al. 2015b), Lucanus maculifemoratus & Prosopocoilus inclinatus, (Okamoto and Hongo 2013) and Aegus chelifer (Songvorawit et al. 2018). These studies indicate that the behavioural elements and sequences are highly variable either between or within species. The commonly used setups and procedures for studying the contest interactions of stag beetles include (1) placing two males some distances apart on a horizontal tree log (Goyens et al. 2015b; Kuan 2011) and (2) allowing the study animals a short time period (15-60mins) to acclimatize (Goyens et al. 2015b; Kuan 2011; Okamoto and Hongo 2013; Songvorawit et al. 2018). These setups and procedures did not appear to be suitable for investigating the contest behaviours of C. mniszechi because 47% of the contest pairs failed to interact (Kuan 2011). Different experimental setups, more resembling the study animal’s natural habitats, are therefore needed for studying the contest behaviours of different species of stag beetles.

3

Selection can often favor individuals engaging in fighting contests for potential fitness advantage. However, the fighting contest can be costly for contestants in terms of time and energy trade-off, risk of predation and injury (Arnott and Elwood 2009; Hsu et al. 2008; Johnson et al. 2006; Lane and Briffa 2017). Therefore, animals have evolved various ways of assessing the benefit and cost of continuing the fights to make decisions about whether to continue or withdraw from the contests (Arnott and Elwood 2009; Hsu et al. 2008). The assessment strategies of aniamal contests can be classified into two categories: self-assessment and mutual assessment strategy. In self-assessment strategy, currently there are two models, war of attrition (WOA) and cumulative assessment model (CAM). There is one model, sequential assessment model (SAM), in mutual assessment strategy. These models differ in how information about the fighting ability (i.e., resource holding potential, RHP, Parker 1974) of the contestants is gathered (Arnott and Elwood 2009). RHP is often related to phenotypical traits that determine the probability of winning a contest, such as body size, weapon size or energy reserve (Arnott and Elwood 2009; Vieira et al. 2013). In WOA or energetic war of attrition (E)WOA, each contestant only has information about its own fighting ability and the contestants persist simply according to their own RHP. Thus, the weaker contestant tends to withdraw first and the contest duration depends on loser’s RHP(Arnott and Elwood 2009). In CAM, the contestants gather information regarding their own fighting ability and the cost (damage, injury) accrued from the rivals. The decision to withdraw is based on the contestant’s own RHP

4 and the rival’s RHP as a proxy of the ability to inflict cost on the opponent. In mutual assessment strategy or SAM, both contestants obtain information considering the rival’s fighting ability relative to their own. The loser with the lower RHP has the benefit of quickly withdrawing from the contest to save time, energy and risk of injury. Thus, mutual assessment appears to be a more efficient process of decision making than (E)WOA and CAM because the difference between rivals’ RHP is constantly compared during the fight. However, (E)WOA and CAM can be more cost-effective than mutual assessment when the time and energy needed for rival assessment are large, or if the signals were unreliable cues of rival’s fighting ability (Elias et al. 2008; Taylor and Elwood 2003). For empirical studies of animal contests, the three alternative assessment strategies can be distinguished by the predicted relations between contest duration and winner/loser’s RHP, and by testing how fighting behaviours progress throughout the contest (Arnott and Elwood 2009; Jennings et al. 2004; Taylor and Elwood 2003). Stag beetles (Lucanidae) use their exaggerated mandibles as weapons in male-male contests for access to the sap sites, territory and females (Emlen 2008a; Goyens et al. 2015b; Inoue and Hasegawa 2012; Songvorawit et al. 2018). Males of stag beetles with larger mandibles than their rivals tend to win more often in the contests (Goyens et al. 2015b; Kuan 2011). However, it was unclear whether stag beetles assess the fighting ability of their own and rivals. And if so, what assessment strategy was applied in male-male contests of stag beetles? In the stag beetle Cyclommatus metallifer, larger males with longer mandibles were

5 found to have larger closer muscles and produce higher bite forces than smaller males, suggesting that the mandibles can be an honest signal of fighting ability (Goyens et al. 2014; Mills et al. 2016). The displaying behaviour of opening mandibles before fights in stag beetles has been suggested to be used to assess RHP of the rivals by visually comparing the widths of mandible openings (Lucanus maculifemoratus and Prosopocoilus inclinatus, (Okamoto and Hongo 2013); C. metallifer,(Mills et al. 2016)). The distribution and densities of bristle mechano-sensors on the mandibles were found to be correlated with material stress of the mandibles in C. metallifer, suggesting that the mechano-sensors are used in a negative feedback system to prevent physical damage, and may function to assess the rivals and optimize fighting strategy in the contests (Goyens et al. 2015a). These findings led to the hypothesis that stag beetles may employ mutual assessment strategy in male-male contests (Goyens et al. 2015c; Goyens et al. 2014; Mills et al. 2016; Okamoto and Hongo 2013) The assessment strategies of stag beetle were tested in three studies (C. miniszechi, (Kuan 2011); C. metallifer, (Goyens et al. 2015b); Aegus chelifer, (Songvorawit et al. 2018)). In C. metallifer, there was a significant positive relation between contest duration and loser RHP (mandible size), but no relation existed between contest duration and winner RHP to conclusively distinguish alternative assessment strategies (Goyens et al. 2015b). In A. chelifer, the study found a significant negative relation of contest duration and RHP (head width) differences, but no relation was found between contest duration and either winner or

6 loser RHP to identify the assessment strategy (Songvorawit et al. 2018). The study of C. miniszechi revealed a significantly positive relation between contest duration and loser RHP (mandible size), a non- significantly positive relation between contest duration and winner RHP (mandible size), and a significantly negative relation of contest duration and RHP differences to suggest self-assessment strategy (Kuan 2011). The commonly used setups and procedures for studying the contest interactions of stag beetles include (1) placing two males some distances apart on a horizontal tree log (Goyens et al. 2015b; Kuan 2011) and (2) allowing the study animals a short time period (≤15-60 mins) to acclimatize (Goyens et al. 2015b; Kuan 2011; Okamoto and Hongo 2013; Songvorawit et al. 2018). These setups and procedures did not appear to be suitable for investigating the contest behaviours of C. miniszechi because 47% of the contest pairs failed to interact (Kuan 2011). Different experimental setups, more resembling the study animal’s natural habitats, are therefore needed for studying the contest behaviours of different species of stag beetles. Considering allometry between body size and RHP traits is important in animal contests because testable predictions of the relation between contest duration and RHP derived from assessment models assume linear scaling relationships (slope = 1) among RHP traits (Palaoro and Briffa 2017). However, many animals like stag beetles show steep allometry (slope >> 1) between body size and weapon size (McCullough et al. 2015; Voje 2016). Instead of a positive linear relation between contest duration and RHP expected by assessment models assuming linear

7 rela tionships, theoretical models of nonlinear scaling (slope ≠ 1) relationships among RHP traits predict contest duration would only increase with body size until a maximum and decrease afterwards (Palaoro and Briffa 2017). If the RHP had a positively non-linear allometric relationship with the body size, the contest duration would be expected to display different correlation coefficient to RHP between different morphs when the stag beetle applied EWOA and CAM. This study investigated the allometry, fighting behaviour and assessment strategy in the stag beetle C. mnizechi. An earlier study classified male C. mnizechi into three morphs based on the shape and size of the mandibles (Figure 1) (Kuan 2011). However, the slope and shape of the allometry between mandible and body size of the beetles, nonetheless, remained unclear. We first examined the relationship between mandible and body size to identify which type of allometry best describe the relationship, and determine whether the males could be grouped into different morphs based on the relationship. The allometric slopes of male morphs provide important information on the level of resource allocation of weapons. Secondly, the behavioural sequences of male-male fights were characterized and tested using sequential analyses of staging contests. Self- and cumulative assessment predict that contest behaviours progress in any direction and without phases; whereas contest behaviours of mutual assessment are predicted to progress uni- directionally in phases of escalating intensity and should not de-escalate to earlier behaviours (Enquist et al. 1990; Green and Patek 2018; Payne 1998; Payne and Pagel 1997). Finally, the correlational predictions of

8 alternative assessment strategies were tested using both randomly and size-matched contests. The relationships between contest duration (and aggressiveness) and RHP were examined. For randomly matched contests, self-assessment predicts a strong positive relation between contest duration and loser RHP, and a weak positive relation between contest duration and winner RHP (Arnott and Elwood 2009; Taylor and Elwood 2003). However, cumulative and mutual assessment both result in a positive relation between contest duration and loser RHP, but an equally strong negative relation between contest duration and winner RHP. Therefore, size-matched contests were used to further discriminate between cumulative and mutual assessment. For RHP-matched rivals, self- and cumulative assessment both predict a positive relation between contest duration and contestant RHP, but mutual assessment will show no relation. To study the contest behaviours of the males, our trial setups enabled the males to encounter each other at the feeding site to compete for food, which resembled the competition occurring in their natural habitats.

Materials and Methods

Study organism The stag beetle C. mnizechi is distributed in SE China, Vietnam and northern Taiwan and inhabits lowland forests below approximately 750m (Li 2004). During the breeding seasons from May to August in Taiwan, males perform mate-guarding behaviours (Figure 1a) and engage in territorial fights (Figure 1b) at the sap sites on the branches or trunks of 9 the broadleaf trees, including Fraxinus griffithii (Euphorbiaceae), Broussonetia papyrifera (Moraceae), Citrus spp. (Rutaceae) and Koelreuteria elegans (Sapindaceae). Male C. mniszechi are highly variable in body sizes but can be classified into three major morphs (alpha, beta and gamma) by the size and shape of the mandibles (Kuan 2011) (Figure 1c). Alpha males have the largest mandibles with tusk-like projections (denticles) at the distal half of the mandibles. Beta males have smaller mandibles with denticles at their proximal halves and close to the base of the mandibles. Gamma males have a pair of the smallest saw-like mandibles with no apparent denticles. In this study, field-collected male C. mniszechi from 2016 to 2018 had 37.5% alpha, 47.2% beta and 15.3% gamma morphs (Table S1).

Insect rearing Adult C. mnizechi were collected from the lowland forests of northern Taiwan by light traps, chopping decayed wood and manually searching at sap-sites on tree trunks between 2016 and 2018 (Table S1). Field- collected males were used in the fighting contests of the same year, except the adults collected in 2016 for breeding and morphological measuring only. Field-collected females were bred in the laboratory then their male offspring were used in fighting contests of the following year. Some larvae in 2016 and 2017 were obtained from local insect breeders. Adult stag beetles were reared individually in plastic containers (15cm×10cm×11cm) at 25°C under 12h: 12h (light: dark) cycle. The larvae of 2016 were reared in 250ml cylindrical plastic containers (7.5cm

10 in diameter; 4.5cm in height) filled with fermented sawdust (good-quality microparticle fermented oak sawdust, Max Piggyfat Insect Feeding Facilities, Taiwan) under 12h: 12h (light: dark) cycle at 25°C. Because synchronous metamorphosis of beetle pupae can be regulated by chemical signal, social cue, and the fluctuation of temperature (Kojima 2014; Kojima et al. 2014). To synchronize the emergence time of adult males in 2018 for setting up fighting contests, the temperature (15.8–25.3°C) used for larval rearing from September of 2017 to May of 2018 was adjusted every two weeks according to monthly mean temperature from 2012 to 2016 in Cyuchih (70m) and Shanjia (48m) weather stations of the Central Weather Bureau of Taiwan, which are the closest stations to the collection sites (Table S1).

Morphological measurements Mandible length (ML), head width (HW) and elytra length (EL) of the beetles were measured to the nearest of 0.01mm using a digital caliper (99MAD027M1, Mitutoyo, Kanagawa, Japan). ML was the average linear distance between the distal tip of the mandible and the axis of mandibular joint of both mandibles (Goyens et al. 2016); HW was the linear distance between the tips of the protrusions anterior to the eyes; EL was the linear distance between the posterior ends of the scutellum and the elytra (Figure 1c). All measurements were obtained on the same day when the beetles were collected in the field or the 21th days after the eclosion of the beetles reared in the laboratory. The body weight (BW) of the beetles was measured to the nearest of 0.001g using a digital scale

11

(C T-50, HIRODA, Shenzhen, China) one day before the fighting contest.

Allometry analyses The morphological measurement of a total of 232 males (34 field- collected and 198 laboratory-reared males) were used for allometry analyses. Among 198 laboratory-reared individuals, 145 larvae were bred from field-collected females and 53 larvae were obtained from local insect breeders. The plot of natural log of body size (elytra length, EL; X) against the natural log of weapon size (mandible length, ML; Y) indicated that the relationship between them was continuous and nonlinear (Figure 2). We thus followed Knell (2009) to characterize possible nonlinear allometries. We first tested for linearity using the quadratic model (Eberhard and Gutierrez 1991; Songvorawit et al. 2017): Quadratic model

2 lnY = β0 + β1 lnX+ β2 (lnX) + ε where β0 is the intercept, β1 and β2 are the regression coefficients and ε is the error. A coefficient β2 which significantly differs from zero would suggest a nonlinear relationship. We then tested for the following statistical models(Eberhard and Gutierrez 1991; Kotiaho 2001) using Akaike (AIC) and Bayesian information criterion (BIC) for model selection: Linear model lnY = β0 + β1 lnX + ε Eberhard and Gutierrez continuous piecewise lnY = β0 + β1 lnX + β2 (lnX − lnX0) D + ε Eberhard and Gutierrez discontinuous piecewise

lnY = β0 + β1 lnX + β2 (lnX − lnX0) D + β3 D + ε Kotiaho and Tomkins linear

lnX = β0 + β1 lnY + ε 12

Kotiaho and Tomkins quadratic 2 lnX = β0 + β1 lnY + β2 (lnY) + ε Kotiaho and Tomkins continuous piecewise

lnX = β0 + β1 lnY + β2 (lnY − lnY0) D + ε Kotiaho and Tomkins discontinuous piecewise

lnX = β0 + β1 lnY + β2 (lnY − lnY0) D + β3 D + ε For Eberhard and Gutierrez continuous and discontinuous piecewise models, where lnX0 is the switch point, D = 0 if lnX  lnX0 (minor males) or D = 1 if lnX  lnX0 (major males). The best estimated value of lnX0 was determined by the maximum adjusted R-square values among 70 of lnX0 (lnX0: from 2.3 to 3.0 in 0.1 interval) in model 2 (Sugiura et al.

2007). If β2 was significantly different from zero, the slope of allometry changed at lnX0. If β3 differed significantly from zero, the allometry was discontinuous at lnX0. The isometry test using the equation lnY = A lnX + lna, a linearized form of Y = aXA (Huxley 1931), as conducted to examine the allometric slope A between mandible size (ML)/ head width (HW) and body size (EL) of different male morphs. The significance of difference between A and 1 was evaluated by two-tailed t-test. An allometric slope of A > 1 (positive allometry) indicates that larger males have disproportionally larger mandibles than that of smaller males. When A = 1 (isometry), the mandible size among males is exactly proportional to their body sizes. A slope of A < 1 (negative allometry) shows that larger males have disproportionally smaller mandibles than that of smaller males. All allometric analyses were conducted in R (version 3.6.0, R Development Core Team, 2019) using the R scripts of Kojima and Lin 2017 (Kojima and Lin 2017). 13

Fighting contests The fighting arena was an acrylic container (32cm×18cm×30cm) filled with 400ml sawdust (1cm height) at the bottom (Figure 3). The arena was divided into one fighting (16cm×18cm) and two resting (8cm×18cm) zones using two acrylic dividers (18cm×30cm). Three to four dry leaves of oak (Quercus glauca, Fagaceae) were placed on the surface of sawdust layer of the resting zones for the beetles to hide under. A piece of half-cut wood (10cm×16cm×5cm) was placed at the centre of fighting zone to serve as a feeding station to facilitate competition between two males. The arena was placed in a room at 25°C and 12h: 12h (light: dark) cycle. All males used in the fighting contests had no prior fighting experience and their bodies were intact with all body parts (without any damages). All contests were conducted in the evening hours (from 6 p.m. to 7 p.m.) because C. mnizechi is nocturnal and shows a distinct day-night cycle of movements (Kuan 2011). The day before the contest day at 6 p.m., the two male contestants were placed in the two resting zones (one male per resting zone) of the fighting arena to acclimatize. The male stag beetles usually hid under the dry leaves in the resting zones without obvious body movements during daytime hours. At 6 p.m. of the contest day, a 2ml tube filled with insect jelly (PPS-801, Champ E Pets Corporation, Taipei, Taiwan) was inserted at the centre of the feeding station (Figure 3). Then the two dividers were removed simultaneously to allow the two males to interact. The two males usually emerged from under the dry leaves, moved around the arena, and walked toward the

14 feeding station within 5 to 10 mins after the dividers were removed. Agnostic interactions occurred when two males encountered each other, usually on the feeding station. This contest setup allowed the males to encounter each other during foraging/feeding and decide how to interact with each other. These resemble the situations that would occur to the males in their nature habitats. The males were allowed to interact and fight until resolved with a clear winner and loser. If the two males did not exhibit sufficient aggression toward each other to resolve with a clear winner and loser, the trials would be terminated after 1 h. After the termination of the trials, the two males were removed from the arena. The contests were recorded using night-vision video monitors (DS-VR7160H, Der Shuenn, Taipei, Taiwan) positioned 65cm above the arena.

Determination of traits associated with RHP Firstly, randomly matched contests were used to examine which morphological traits were associated with the contest outcomes in C. mnizechi. The males reared in 2017 were assigned randomly to the contests. The natural log of mandible length, head width and body weight of winner and loser males differed significantly (Paired t-test: mandible length mean difference=0.082mm, t=2.993, p=0.006, n=30; body weight mean difference=0.106g, t=2.239, p=0.033, n=30), whereas the elytra length of winner and loser males did not differ (elytra length, mean difference=0.02mm, t=1.398, p=0.173, n=30). Multiple paired t-tests were conducted to examined the difference of traits between winners and losers using the Bonferroni method of adjusting the significant p-value to

15

0.0125 (Bland and Altman 1995). Only the difference of mandible length between winners and losers was significant. Therefore, mandible length was used as a proxy for RHP in C. mnizechi. Secondly, size-matched contests were used to remove the effect of unequal RHP to examine the relationship between the contest duration/aggressiveness and the RHP. The males reared in 2018 with similar mandible length paired up for contests. The difference in mandible length between the contestants was less than 5% (0.686mm) of the median mandible length (13.718mm) of all contestants.

Sequential analyses of fighting behaviours We used BORIS v. 7.4 (Behavior Observation Research Interactive Software (Friard et al. 2016) to transcribe the fighting behaviours of the males from the videos based on nine behavioural elements (Table 1). The contests were characterized as less (0: no ‘tussles’) or more intensive (1: with ‘tussles’). The fighting behaviours were coded individually in two adjacent columns, “preceding behaviour” followed by “subsequent behaviour”, for all behavioural transitions. The behavioural sequence data was summarized into adjacency matrices of contest behaviours using igraph network analysis package (Csardi and Nepusz 2006) in R following Green and Patek 2018. The transition rate of each behavioural transition was calculated from the ratio of original behaviour to its subsequent behaviour. To examine whether specific behavioural transition was more frequent than expected by chance, a permutation procedure was conducted by fixing the 1st columns (preceding behaviour), keeping the

16 relative frequency of the behaviours, and randomly sampling the 2nd columns (subsequent behaviour), thus randomizing behavioural transitions between the behaviours (Bakeman et al. 1996; Green and Patek 2018). The permutation process was executed for 10,000 times to generate a null distribution of random transitions. A significant transition was found when the observed transition was more frequent than its respective 95% quantile of the null distribution of random transitions. Significant behavioural transitions of the contests were plotted as network graphs in igraph. The behavioural elements were organized into four phases (Table 1) based on the network graphs. We defined a phase as a behaviour or subset of behaviours, in which these behaviours had relatively similar level of aggressiveness and they were used with relatively equal frequency (Green and Patek 2018).

Statistical analyses of RHP, aggression and contest duration Statistical analyses of the regression models were conducted in JMP (ver. 10.0.0, SAS institute Inc., Cary, NC, U.S.A.). A logistic regression model was used to examine the effect of either male morphs or mandible length on the level of aggression (probability of tussle) for all contests. Mandible length was used as the single explanatory factor in regression analyses of the contests of different male morphs. A simple linear regression model was used to investigate the relationships between morphological traits and contest duration. Before conducting regression analyses, the assumption of normality of all response variables were examined using Shapiro-Wilk W tests (Shapiro and Wilk 1965). When non-normality of

17 the variables was found (i.e., contest duration: randomly matched contests, W=0.419, p<0.001; size-matched contests, W=0.569, p<0.002), the variables were ln-transformed in the regression models.

Results

Morphological measurement and allometric analyses The morphological traits of male stag beetles were highly correlated to each other (Table 2). For the allometric relationship between the mandible and elytra length, the β2 coefficient in the quadratic model was significantly different from 0, suggesting that the relationship was significantly non-linear (Table 3). Kotiaho and Tomkins continuous and discontinuous piecewise model had the lowest AIC and BIC score, respectively. Kotiaho and Tomkins discontinuous piecewise model improved the model fit only slightly (ΔAIC=0.17) and the β3 coefficient of the model was not significantly different from 0, thus the simpler Kotiaho and Tomkins continuous piecewise model was considered to be the best fit model for the allometric relationship. The value of the switch

2 point (lnY0) which showed the highest adjusted R in this model was 2.64

(14.01mm). The β2 coefficient of this model was significantly different from 0, indicating that the linear slope between the mandible and elytra length is continuous but changed markedly at either side of the switch point (ML=14.01mm) (Figure 2). Therefore, male C. mnizechi are dimorphic and can be divided allometrically into majors and minors above and below mandible length of 14.01mm, respectively. The allometric slope of the mandibles of minor males (A=2.89±0.10) was 18 steeper than that of major males (A=1.43±0.10). The mandibles and heads of both major and minor males showed significantly positive allometries (Table 4).

Behavioural sequences of fighting contests A total of 31 randomly and 57 size-matched contests were staged in 2017 and 2018, respectively. Males of 10 (32.3%) randomly and 15 (26.3%) size-matched contests escalated to more aggressive ‘tussle’ behaviour. Four (12.9%) randomly and three (5.6%) size-matched contests resolved without aggressive behaviours (‘push’, ‘attack’, ‘tussle’ or ‘clamp’). Only one (3.2%) size-matched contest did not resolve and was excluded from the sequential analyses. The average duration of the contests was 114±271.8 (mean±sd) sec and 37±57.8 sec., respectively, for randomly and size-matched males, respectively. The longest contest continued for about 20 mins (1147 sec) while the shortest resolved in less than 1 sec (both were randomly matched contests). The contest started when one male approached and touched the opponent’s body with its mandibles or fore legs in phase 1 (‘touch’, 100%, Figure 4). After ‘touch’, in both randomly and sized-matched contests males often entered ‘defensive posture’ (≥97%) of phase 2 by raising their heads and keeping their mandibles open towards the opponent. ‘Defensive posture’ then progressed to ‘body raising’ (9-27%) with rapid movement of the antenna and ‘push’ by raising mandibles and knocking on the opponents (30-41%) in both types of contests, or ‘attack’ by biting the opponents with mandibles (28%) and rarely ‘clamp2’ (3%)

19 in randomly matched contests. From ‘body raising’ of phase 2, the males entered either ‘attack’ (45-50%) of the same phase or ‘tussle’ (27-50%) by interlocking mandibles and pushing each other in more aggressive phase 3. ‘Push’ often led to ‘retreat’ (35%) in randomly matched contest, but could result in either ‘attack’ (27%) and ‘retreat’ (31%) in size- matched contests. ‘Attack’ progressed to ‘tussle’ (11-18%) and ‘retreat’ (21-32%) in both contests, but could also enter ‘clamp2’ (5%) in randomly matched contests. In phase 3, ‘tussle’ often led to ‘retreat’ (70%) or ‘clamp1’ in randomly matched contests, but frequently resulted in repetitive ‘tussle’ or ‘clamp1’ in size-matched contests. ‘Clamp1’ and ‘clamp2’ of phase 3 often led to ‘retreat’ (70-100%, 44-50%, respectively). By comparison, the relative frequencies of each behavioural elements in both contests were similar (Figure 4). However, randomly matched contests had more behavioural transitions between phases than size- matched contests (i.e., ‘defensive posture’ of phase 2 and ‘clamp2’ of phase 3; ‘attack’ of phase 2 and ‘clamp2’ of phase 3; ‘tussle’ of phase 3 and ‘retreat’ of phase 4). In contrast, size-matched contests had more behavioural transitions within phases than randomly matched contests (i.e., between ‘push’ and ‘attack’ of phase 2; repetitive ‘tussle’, ‘clamp1’ and ‘clamp2’) Overall, in both randomly and size-matched contests, behaviours progressed from (1) approach and touch, to (2) defense and attack, to (3) clamp and tussle, and finally to (4) contest resolution (Figure 4) (Video 1). The contests appeared to progress from lower to higher aggressive phases (e.g., from phase 1 to phase 2, or from phase 2

20 to phase 3) without de-escalation in one direction and with behavioural transitions occurring within the same phases. To distinguish the behavioural sequence between major and minor males, the asymmetry of opponents’ morphs was removed by excluding the 2 morph-mismatched contests from 57 size-matched contests. There were 55 size-matched contests (24 major morphs and 31 minor morphs) remained for behavioural sequential analyses between male morphs. Males of 14 (25.5%) contests escalated to more aggressive ‘tussle’ behaviour. Three (5.5%) contests resolved without aggressive behaviours (i.e., ‘push’, ‘attack’, ‘tussle’ or ‘clamp’). The average duration of the size-matched contests was 37.2±58.8 (mean±sd) sec. The fighting contests of C. mnizechi stag beetle in size-matched contests followed a variable course of behavioural sequences (Figure 5) (Video 1). The contest started when one male approached and touched its opponent’s body with its mandibles or forelegs (100%, transitional probability from ‘start’ to ‘touch’, Figure 5). After ‘touch’, both major and minor males often entered a ‘defensive posture’ by raising their heads and keeping their mandibles open towards their opponents (100%). ‘Defensive posture’ then progressed to either ‘body raising’ with rapid movement of the antenna (19–30%) or ‘push’ by raising mandibles and knocking the opponents (20–37%). From ‘push’, the contests in both male morphs were either resolved by loser males entering ‘retreat’ (22– 44%) or continuing into ‘attack’ by biting the opponents with mandibles (34–45%). From ‘body raising’, the behavioural sequences differed between major and minor males. In major males, ‘body raising’ often

21 directly led to ‘tussle’ (50%) of phase 3 by interlocking mandibles and pushing each other, but it frequently resulted in ‘attack’ of the same phase 2 in minor males (64%). Within phase 2, ‘push’ was repeated many times in minor males (31%) but not in major males. ‘Attack’ led to ‘tussle’ (12%) and ‘retreat’ (22%) in minor males, whereas ‘attack’ was mainly repeated in major males (45%). In phase 3, minor males performed repetitive clamping onto the head (‘clamp1’, 25%) and body of their opponents (‘clamp2’, 25%) and frequent behavioural transitions between ‘tussle’ and ‘clamp2’ (12–25%), while major males displayed repetitive ‘tussle’ (50%) and ‘clamp2’ (20%). From ‘tussle’ of phase 3, the contests were often resolved by winner males clamping their rivals (‘clamp1’, 29– 38%) and then flipping them into ‘retreat’ (50–100%). Overall, the fighting behaviours of C. mnizechi progressed in phases from (1) approach and touch, (2) defence and attack, (3) clamp and tussle, and finally to (4) contest resolution (Figure 5). Behavioural matching between contestant males was observed for most behavioural elements. The contests of C. mnizechi males were likely to escalate from lower to higher aggressive phases (e.g., from phase 1 to phase 2, or from phase 2 to phase 3), with a few behavioural transitions occurring within the same phases (Figure 5). Once contests escalated to aggressive physical contact (i.e., ‘push’ and ‘attack’ of phase 2; ‘clamp’ and ‘tussle’ of phase 3), behaviours frequently transitioned within phase or to phase 4 of contest resolution without first de-escalating to non-physical interactions in phase 2 (i.e., ‘body raising’ or ‘defensive posture’). Minor males had more frequent behavioural transitions within phases than major males (i.e.,

22 between ‘body raising’ and ‘attack’, and repetitive ‘push’ of phase 2; between ‘tussle’ and ‘clamp2’, and repetitive ‘clamp1’ of phase 3, Figure 5). These results suggest that minor males tend to stay within these phases more than major males, whereas major males were more likely to escalate directly into more aggressive phases.

Male morph and contest aggression Tussle represented the highest level of aggression involving contestants interlocking their mandibles and pushing each other, and was the most frequent escalation behaviour in the phase 3 in C. mnizechi contests (Figure 4). In randomly matched contests, there was a significantly positive relationship between the probability of tussle and winner’s mandible length (estimate±SE=1.334±0.539, Chi-square=6.12, p=0.0134) (Figure 6a) (Table 5), suggesting that winner males with larger mandibles were more likely to reach the highest aggression in the contests. There was no significant relationship between the probability of tussle and loser’s mandible length (estimate±SE=0.034±0.239, Chi-square=0.02, p=0.888) (Figure 6b) (Table 5). In size-matched contests, the probability of tussle had a weakly positive relation with average mandible length (estimate±SE=0.182±0.105, Chi-square=3.01, p=0.083, n=57) (Figure 6c). After excluding two pairs of mismatching male morphs (major vs minor) in size-matched contests, the regression analyses showed that the probability of tussle was not correlated with either mandible length (estimate±se=0.185±0.106, Chi-square=3.02, p=0.082) or male morphs

23

(estimate±se=0.344±0.621, Chi-square=0.31, p=0.579) for all contests (n=55). In the contests of minor males, the probability of tussle had a significant positive relation with mandible length (estimate±se=0.709±0.361, Chi-square=3.86, p=0.049, n=31), whereas no significant relationship was found in the contests of major males (estimate±se=0.341±0.310, Chi-square=1.21, p=0.271, n=24). These results indicate that mandible length and male morphs have no significant effect on the probability of entering tussle. However, within minor morphs males with larger mandibles behaved more aggressively and were more likely to engage in tussles.

Contest duration and RHP In randomly matched contests, simple linear regression found significantly positive relationships between contest duration and both winner’s and loser’s mandible length (n=30) (winner, estimate=0.513, SE=0.148, F=11.955, p=0.002; loser, estimate=0.279, SE=0.109, F=6.595, p=0.016) (Figure 7a&b). In size-matched contests, the contest duration was significantly positively correlated with the average mandible length (n=57) (estimate=0.174, SE=0.049, F=12.488, p=0.001) (Figure 7c). No significant effect was found for the male morphs on the contest duration(n=49) (estimate=-1.238, F=1.688, p=0.196, n=49, Table 6).

Discussion

Dimorphic males, allometry and fighting behaviours Weapon size (mandible length) is the main determinant of the contest 24 outcome and can be used as a reliable proxy for RHP in C. mnizechi. This finding confirms the results from earlier studies of this (Kuan 2011) and other species of stag beetles (Lucanus cervus, (Lagarde et al. 2005); Prosopocoilus inclinatus, (Inoue and Hasegawa 2012); Cyclommatus metallifer, (Goyens et al. 2015b)), in which the males with larger weapons win the majority of fighting contests. The functional advantages of a longer mandible in stag beetle battles may come from the larger biting force and enhanced ability to reach the rival’s legs to detach them (C. metallifer, (Goyens et al. 2015b)). However, the study of a smaller stag beetle species Aegus chelifer suggest that body size is more important than weapon size in fighting success (Songvorawit et al. 2018). Body size and other body traits can be directly related to the physical strength of the males thus enhance their fighting ability. Therefore, the relative importance of weapon and body size on the contest outcome of stag beetles may be species-specific, depending on their fighting behaviours and ecology. Positive allometries were found for the mandibles and heads of C. mnizechi males. Allometric analyses demonstrate that C. mnizechi consists of dimorphic males defined quantitatively as majors and minors by mandible size at the switch point. The allometric slope of mandibles of the major males is not as steep as that of the minor males, indicating that the major males invest proportionally less in the development of mandibles. The reduced allometric slope in weapons of larger males is also found in other stag beetles (Knell et al. 2004), rhinoceros beetles (Hongo 2007; McCullough et al. 2015) and flower beetles (Kojima and

25

Lin 2017). This pattern can be explained by resource exhaustion for pupal development of weapons (Knell et al. 2004; Nijhout and Wheeler 1996), resource allocation between weapons and other body parts (Emlen 2001; Simmons and Emlen 2006), or constraint of natural selection for costs in locomotion (Goyens et al. 2015c) and predation (Kojima et al. 2014). Alternatively, the difference in allometric slopes of mandibles between male morphs of C. mnizechi can be explained by the biomechanical functions of using mandibles as threat signals versus weapons (Eberhard et al. 2018) The mandible opening behaviour of C. mnizechi could be used as a threat display to amplify the size difference between opponents. Whereas the functional advantages of a longer mandible may come from larger biting force or enhanced ability to reach the rival’s bodies to detach them. The reduced allometric slope in mandibles of major males in C. mnizechi may represent a trade-off between these different functions (Dennenmoser and Christy 2013). Our results suggest that the major and minor males of C. mnizechi employ the same behavioural elements in the fighting contests, but they differ in the tendency to stay within phases in the contests and the likelihood of escalating into more aggressive tussle behaviour. Although the regression analyses found no significant effect of male morphs on the probability of tussle, the sequential analyses indicated that the major males were more likely to tussle and escalate directly into more aggressive phases, whereas the minor males tended to stay within phases. In contrast to the observataion of another stag beetle species, A. chelifer, the minor males tend to escalate more frequently than the major males

26

(Songvorawit et al. 2018). Our findings suggest that the major males of C. mnizechi may have been selected to be more aggressive than minor males to compete in contests. The higher aggression of major males in contests may be associated with using mandibles as weapons over threat signals to apply physical force and flap the rivals (Eberhard et al. 2018). Within the minor morphs of C. mnizechi, the regression analysis indicated that males with larger mandibles were more likely to tussle, which is consistent with the expectation from this weapon hypothesis. The elevated aggression in the major males can also be explained by their higher investment in production of relatively larger weapons. The loss of fitness benefit through withdrawing from contests may have a higher cost for the major males, so they tend to be more aggressive than the minor males in contests. Larger males of stag beetles with longer weapons (more muscle mass) have disproportionally higher resting metabolic rates than that of smaller males with shorter weapons (O'Brien et al. 2019). In C. mniszechi, the slope of positive allometry between resting metabolic rates and body weight was less steep in the minor males (Chen et al., unpublished data), suggesting that a higher metabolic cost for major males in contests.

Self and mutual assessment strategy In randomly matched contests, strong positive relations between contest duration and both winner’s and loser’s mandible sizes are in agreement with the hypothesis that self-assessment determines strategic decisions in fighting male C. mniszechi. Especially the positive relation between the

27 winner ’s mandible sizes and contest duration is informative to exclude the hypothesis of cumulative and mutual assessment strategy, which predicts a negative relation. In size-matched contests, a positive relation between contest duration and mandible sizes further support self- assessment strategy in C. mniszechi, whereas no relation is expected by mutual assessment strategy. There is also a trend that winner males with larger mandible and body sizes are more likely to escalate to the highest aggressive tussle behaviour, which is predicted by EWOA model but not CAM nor SAM. Together, the results of correlational analyses clearly support self-assessment strategy as the best model to explain decision making in the fighting contest of C. mniszechi. However, the ethological details of contest behaviours match more closely the predictions from mutual assessment strategy. The results of sequential analyses showed that the fighting contests in C. mniszechi involve many physical contacts with rare injuries, unidirectional behavioural progressions in phases, and progression with no de-escalation from low towards high aggression (Figure 4). There was behavioural matching between winners and losers of C. mniszechi males, which provides further evidence supporting mutual assessment as a possible explanation. Therefore, it seems likely that the fighting contests of C. mniszechi are not settled entirely on the basis of pure self-assessment, and that CAM or mutual assessment may also play a role in contest decisions. This finding joins a growing body of empirical studies suggesting that self- and mutual assessment are not mutually exclusive alternatives, but are extremes of continuum of possible assessment strategies (i.e., partial

28 mut ual assessment, Prenter et al. 2006; reviewed in Chapin et al 2019) (Chapin et al. 2019; Pinto et al. 2019; Prenter et al. 2006). The finding of both self- and mutual assessment strategies in this study emphasizes the importance of a combined approach that integrates both tests of behavioural progressions and correlations of contest duration and RHP to differentiate assessment strategies (Briffa and Elwood 2009; Green and Patek 2018; Taylor and Elwood 2003). The significance of mutual assessment strategy in C. mniszechi stag beetles could have been overlooked without the insights derived from sequential behavioural analyses. Empirical studies using the tests of correlations between contest duration and RHP (Arnott and Elwood 2009; Briffa and Elwood 2009; Taylor and Elwood 2003) can have mixed or inconclusive results because factors other than contestant’s RHP may also affect contest decisions, such as the value of contested resources and individual-level variation of assessment strategies (Chapin et al. 2019). To achieve more natural fighting sequences, the experimental setups of fights in this study often allow only one of the two C. mniszechi males access the contested resources (insect jelly) prior to the fights. This experimental procedure may result in an increase of motivational state for males of the resource holder (residence effect) to win the contests similar to the other stag beetles (Prosopocoilus inclinatus, (Inoue and Hasegawa 2012)). The asymmetry of information about the resource value between the two fighting C. mniszechi males may have under-estimated the resource holder’s RHP based only on mandible size and biased the following

29 correlational tests. However, in this study the results of correlational tests between contest duration and RHP are all consistent with the predictions of self-assessment strategy, suggesting that the potential effect of asymmetric resource assessment on contest decisions is likely negligible.

Probable mechanisms of assessment in C. mniszechi stag beetles Mutual assessment is often viewed to be more cognitively complex than self-assessment because it needs appraisal of both contestants, whereas self-assessment only requires information from one’s own state (Elwood and Arnott 2012; Reichert and Quinn 2017). Moreover, a recent meta-analysis of animal contests concluded that the majority of species settle their contests based on self- rather than mutual assessment strategy but the physical contact, behavioural phases and RHP asymmetry related escalation pattern did not match the prediction of self-assessment strategy (Pinto et al. 2019). Our findings from the fighting contest of C. mniszechi stag beetles suggest that this species is likely to always have reliable information of its own abilities (self-assessment), and may be able to gather some information about its opponents at specific stages of the contest. The possible mechanisms of assessment between opponents of C. mniszechi males may involve contest behaviours using visual, chemical or mechanical cues. At the defensive posture without physical contact, males of C. mniszechi raise their heads and keep their mandibles open toward the opponents (Video 1). Similar to other stag beetles, C. metallifer, L. maculifemoratus and P. inclinatus, this mandible opening

30 behaviour could be used as a visual display by C. mniszechi to assess information regarding their opponents (Goyens et al. 2015b; Mills et al. 2016; Okamoto and Hongo 2013). The blindfolded fight experiments of stag beetle C. metallifer showed that the males can still fight fiercely and lift opponents without vision (Goyens et al. 2015b). This finding suggests that stag beetles may continue the fights based mainly on self-assessment of their own fighting ability when potential visual cues of the opponents are unavailable. During the body raising, C. mniszechi males continue raise their bodies with opening mandibles and rapid antennal movement to match each other (Video 1). In addition to opening mandibles, this body raising behaviour in C. mniszechi presumably also act as a visual display to assess the body size and condition of the opponents before more aggressive behaviours (push, attack, tussle) take place. The rapid antennal movement in defensive posture and body raising suggest that fighting C. mniszechi males may actively obtain information via chemical cues, similar to many male stag beetles using pheromones to locate females near their natal nesting sites (Lin et al. 2009; Rink and Sinsch 2007). Furthermore, C. metallifer stag beetles can response to the substrate vibrations made manually on the tree log (Goyens et al. 2015b). The substrate vibration can potentially be used as mechanical cues in assessment between opponents throughout various stages of the fighting contests. Besides non-physical contact behaviours of mandible opening, body raising and rapid antennal movement, fighting C. mniszechi males may

31 also gather information of the opponents in aggressive behaviours involving physical contact, such as tussles. The mandibles of C. metallifer stag beetles are honest signals and equipped with numerous mechano- sensors which may function to assess the material strength of opponent’s mandibles (Goyens et al. 2015a). Therefore, males of congeneric C. mniszechi may likely have similar mechano-sensors on the mandibles to gauge their opponent’s physical strength in tussles by interlocking each other’s mandibles while attempt to push and lift the rival (Video 1). Because C. mniszechi is a nocturnal species and most active around midnight (Kuan 2011), chemical and mechanical rather than visual signals are more likely candidates to be used in mechanisms for rival assessment. However, the functional roles of these potential signals in each phase of fighting contests of C. mniszechi stag beetles need to be further tested using manipulative experiments on these cues.

Acknowledgements This work was supported by research grants to CPL (MOST 106-2311-B- 003-004-MY3 and 107-2311-B-003-002-MY3) and YH (MOST 106- 2621-B-003-001-MY3) from the Ministry of Science and Technology of Taiwan.

32

Reference

Andersson MB (1994) Sexual selection. Princeton University Press, Arnott G, Elwood RW (2009) Assessment of fighting ability in animal contests. Anim Behav 77:991-1004 https://doi.org/10.1016/j.anbehav.2009.02.010 Bakeman R, Robinson BF, Quera V (1996) Testing sequential association: Estimating exact p values using sampled permutations. Psychol Methods 1:4-15 https://doi.org/10.1037/1082-989x.1.1.4 Bland JM, Altman DG (1995) Multiple significance tests: the Bonferroni method. BMJ 310:170 https://doi.org/10.1136/bmj.310.6973.170 Bonduriansky R (2007) Sexual selection and allometry: a critical reappraisal of the evidence and ideas. Evolution 61:838-849 https://doi.org/10.1111/j.1558- 5646.2007.00081.x Bonduriansky R, Day T (2003) The evolution of static allometry in sexually selected traits. Evolution 57:2450-2458 https://doi.org/10.1111/j.0014- 3820.2003.tb01490.x Briffa M, Elwood RW (2009) Difficulties remain in distinguishing between mutual and self-assessment in animal contests. Anim Behav 77:759-762 https://doi.org/10.1016/j.anbehav.2008.11.010 Chapin KJ, Peixoto PEC, Briffa M (2019) Further mismeasures of animal contests: a new framework for assessment strategies. Behav Ecol https://doi.org/10.1093/beheco/arz081 Clark JT (1977) Aspects of variation in the stag beetle Lucanus cervus (L.) (Coleoptera: Lucanidae). Syst Entomol 2:9-16 https://doi.org/10.1111/j.1365- 3113.1977.tb00350.x Csardi G, Nepusz T (2006) The igraph software package for complex network research. InterJournal, Complex Systems 1695:1-9 Darwin C (1871) The descent of man, and selection in relation to sex. London: Murray 415 Dennenmoser S, Christy JH (2013) The design of a beautiful weapon: compensation for opposing sexual selection on a trait with two functions. Evolution 67:1181- 1188 https://doi.org/10.1111/evo.12018 Eberhard WG, Gutierrez EE (1991) Male Dimorphisms in Beetles and Earwigs and the Question of Developmental Constraints. Evolution 45:18-28 https://doi.org/10.1111/j.1558-5646.1991.tb05262.x Eberhard WG, Rodríguez RL, Huber BA, Speck B, Miller H, Buzatto BA, Machado G (2018) Sexual Selection and Static Allometry: The Importance of Function. The Quarterly Review of Biology 93:207-250 https://doi.org/10.1086/699410

33

Elias DO, Kasumovic MM, Punzalan D, Andrade MC, Mason AC (2008) Assessment during aggressive contests between male jumping spiders. Anim Behav 76:901-910 https://doi.org/10.1016/j.anbehav.2008.01.032 Elwood RW, Arnott G (2012) Understanding how animals fight with Lloyd Morgan's canon. Anim Behav 84:1095-1102 https://doi.org/10.1016/j.anbehav.2012.08.035 Emlen DJ (1997) Alternative reproductive tactics and male-dimorphism in the horned beetle Onthophagus acuminatus (Coleoptera: Scarabaeidae). Behav Ecol Sociobiol 41:335-341 https://doi.org/10.1007/s002650050393 Emlen DJ (2001) Costs and the diversification of exaggerated animal structures. Science 291:1534-1536 https://doi.org/10.1126/science.1056607 Emlen DJ (2008a) The Evolution of Animal Weapons. Annu Rev Ecol Evol S 39:387- 413 https://doi.org/10.1146/annurev.ecolsys.39.110707.173502 Emlen DJ (2008b) The Evolution of Animal Weapons. Annual Review of Ecology, Evolution, and Systematics 39:387-413 https://doi.org/10.1146/annurev.ecolsys.39.110707.173502 Emlen DJ, Nijhout HF (2000) The development and evolution of exaggerated morphologies in insects. Annu Rev Entomol 45:661-708 https://doi.org/10.1146/annurev.ento.45.1.661 Enquist M, Leimar O, Ljungberg T, Mallner Y, Segerdahl N (1990) A test of the sequential assessment game: fighting in the cichlid fish Nannacara anomala. Anim Behav 40:1-14 https://doi.org/10.1016/s0003-3472(05)80660-8 Friard O, Gamba M, Fitzjohn R (2016) BORIS: a free, versatile open-source event- logging software for video/audio coding and live observations. Methods Ecol Evol 7:1325-1330 https://doi.org/10.1111/2041-210x.12584 Goyens J, Dirckx J, Aerts P (2015a) Mechanoreceptor distribution in stag beetle jaws corresponds to the material stress in fights. Arthropod Struct Dev 44:201-208 https://doi.org/10.1016/j.asd.2015.03.003 Goyens J, Dirckx J, Aerts P (2015b) Stag Beetle Battle Behavior and its Associated Anatomical Adaptations. J Insect Behav 28:227-244 https://doi.org/10.1007/s10905-015-9495-3 Goyens J, Dirckx J, Aerts P (2016) Jaw morphology and fighting forces in stag beetles. J Exp Biol 219:2955-2961 https://doi.org/10.1242/jeb.141614 Goyens J, Dirckx J, Aerts P, Davidowitz G (2015c) Costly sexual dimorphism in Cyclommatus metallifer stag beetles. Func Ecol 29:35-43 https://doi.org/10.1111/1365-2435.12294 Goyens J, Dirckx J, Dierick M, Van Hoorebeke L, Aerts P (2014) Biomechanical determinants of bite force dimorphism in Cyclommatus metallifer stag beetles.

34

J Exp Biol 217:1065-1071 https://doi.org/10.1242/jeb.091744 Green PA, Patek SN (2018) Mutual assessment during ritualized fighting in mantis shrimp (Stomatopoda). Proc Biol Sci 285 https://doi.org/10.1098/rspb.2017.2542 Hongo Y (2007) Evolution of male dimorphic allometry in a population of the Japanese horned beetle Trypoxylus dichotomus septentrionalis. Behav Ecol Sociobiol 62:245-253 https://doi.org/10.1007/s00265-007-0459-2 Hsu Y, Lee S-P, Chen M-H, Yang S-Y, Cheng K-C (2008) Switching assessment strategy during a contest: fighting in killifish Kryptolebias marmoratus. Anim Behav 75:1641-1649 https://doi.org/10.1016/j.anbehav.2007.10.017 Huxley JS 46. The relative size of antlers in deer. In: Proc Zool Soc Lond, 1931. vol 3. Wiley Online Library, pp 819-864. https://doi.org/10.1111/j.1096- 3642.1931.tb01047.x Inoue A, Hasegawa E (2012) Effect of morph types, body size and prior residence on food-site holding by males of the male-dimorphic stag beetle Prosopocoilus inclinatus (Coleoptera: Lucanidae). J Ethol 31:55-60 https://doi.org/10.1007/s10164-012-0350-0 Jennings DJ, Gammell MP, Carlin CoM, Hayden TJ (2004) Effect of body weight, antler length, resource value and experience on fight duration and intensity in fallow deer. Anim Behav 68:213-221 https://doi.org/10.1016/j.anbehav.2003.11.005 Johnson HE, Bleich VC, Krausman PR, Koprowski JL (2006) Effects of antler breakage on mating behavior in male tule elk (Cervus elaphus nannodes). Eur J Wildl Res 53:9-15 https://doi.org/10.1007/s10344-006-0060-4 Kawano K (2000) Genera and Allometry in the Stag Beetle Family Lucanidae, Coleoptera. Ann Entomol Soc Am 93:198-207 https://doi.org/10.1603/0013- 8746(2000)093[0198:Gaaits]2.0.Co;2 Knell RJ (2009) On the analysis of non-linear allometries. Ecol Entomol 34:1-11 https://doi.org/10.1111/j.1365-2311.2008.01022.x Knell RJ, Pomfret JC, Tomkins JL (2004) The limits of elaboration: curved allometries reveal the constraints on mandible size in stag beetles. Proc Biol Sci 271:523-528 https://doi.org/10.1098/rspb.2003.2641 Kojima W (2014) Mechanism of synchronous metamorphosis: larvae of a rhinoceros beetle alter the timing of pupation depending on maturity of their neighbours. Behav Ecol Sociobiol 69:415-424 https://doi.org/10.1007/s00265-014-1854-0 Kojima W, Ishikawa Y, Takanashi T (2014) Chemically mediated group formation in soil-dwelling larvae and pupae of the beetle Trypoxylus dichotomus. Naturwissenschaften 101:687-695 https://doi.org/10.1007/s00114-014-1199-6

35

Kojima W, Lin C-P (2017) It takes two to tango: functional roles, sexual selection and allometry of multiple male weapons in the flower beetle Dicronocephalus wallichii bourgoini. Biol J Linn Soc 121:514-529 https://doi.org/10.1093/biolinnean/blx018 Kotiaho JS (2001) The discrimination of alternative male morphologies. Behav Ecol 12:553-557 https://doi.org/10.1093/beheco/12.5.553 Kuan CY (2011) To win"big": determining factors and sequential analyses in male- male combats of a stag beetle, Cyclommatus mniszechi (Coleoptera: Lucanidae).49 Lagarde F, Corbin J, Goujon C, Poisbleau M (2005) Polymorphisme et performances au combat chez les mâles de Lucane cerf-volant (Lucanus cervus). Revue d'écologie 60:127-137 Lane SM, Briffa M (2017) The price of attack: rethinking damage costs in animal contests. Anim Behav 126:23-29 https://doi.org/10.1016/j.anbehav.2017.01.015 Li HY (2004) Taiwanese stag beetles., vol 1st. Kissnature Publisher, Taipei, Taiwan Lin C-P, Huang J-P, Lee Y-H, Chen M-Y (2009) Phylogenetic position of a threatened stag beetle, Lucanus datunensis (Coleoptera: Lucanidae) in Taiwan and implications for conservation. Conserv Genet 12:337-341 https://doi.org/10.1007/s10592-009-9996-8 McCullough EL, Ledger KJ, O'Brien DM, Emlen DJ (2015) Variation in the allometry of exaggerated rhinoceros beetle horns. Anim Behav 109:133-140 https://doi.org/10.1016/j.anbehav.2015.08.013 Mills MR, Nemri RS, Carlson EA, Wilde W, Gotoh H, Lavine LC, Swanson BO (2016) Functional mechanics of beetle mandibles: Honest signaling in a sexually selected system. J Exp Zool A Ecol Genet Physiol 325:3-12 https://doi.org/10.1002/jez.1961 Mizunuma T, Nagai S (1994) The Lucanid beetles of the world. 1994. Mushisha Tokyo Moczek AP, Emlen DJ (2000) Male horn dimorphism in the scarab beetle, Onthophagus taurus: do alternative reproductive tactics favour alternative phenotypes? Anim Behav 59:459-466 https://doi.org/10.1006/anbe.1999.1342 Nijhout HF, Wheeler DE (1996) Growth Models of Complex Allometries in Holometabolous Insects. Am Nat 148:40-56 https://doi.org/10.1086/285910 O'Brien DM et al. (2019) Muscle mass drives cost in sexually selected arthropod weapons. Proc Biol Sci 286:20191063 https://doi.org/10.1098/rspb.2019.1063 Okamoto K, Hongo Y (2013) Interspecific contests between males of two Japanese stag beetle species, Lucanus maculifemoratus and Prosopocoilus inclinatus:

36

what overcomes a body size disadvantage? Behaviour 150:39-59 https://doi.org/10.1163/1568539x-00003036 Palaoro AV, Briffa M (2017) Weaponry and defenses in fighting animals: how allometry can alter predictions from contest theory. Behav Ecol 28:328-336 https://doi.org/10.1093/beheco/arw163 Payne RJH (1998) Gradually escalating fights and displays: the cumulative assessment model. Anim Behav 56:651-662 https://doi.org/10.1006/anbe.1998.0835 Payne RJH, Pagel M (1997) Why do animals repeat displays? Anim Behav 54:109- 119 https://doi.org/10.1006/anbe.1996.0391 Pinto NS, Palaoro AV, Peixoto PEC (2019) All by myself? Meta-analysis of animal contests shows stronger support for self than for mutual assessment models. Biol Rev Camb Philos Soc 94:1430-1442 https://doi.org/10.1111/brv.12509 Prenter J, Elwood RW, Taylor PW (2006) Self-assessment by males during energetically costly contests over precopula females in amphipods. Animal Behaviour 72:861-868 https://doi.org/10.1016/j.anbehav.2006.01.023 Reichert MS, Quinn JL (2017) Cognition in Contests: Mechanisms, Ecology, and Evolution. Trends Ecol Evol 32:773-785 https://doi.org/10.1016/j.tree.2017.07.003 Rink M, Sinsch U (2007) Radio-telemetric monitoring of dispersing stag beetles: implications for conservation. J Zoo 272:235-243 https://doi.org/10.1111/j.1469-7998.2006.00282.x Romiti F, Tini M, Redolfi De Zan L, Chiari S, Zauli A, Carpaneto GM (2015) Exaggerated allometric structures in relation to demographic and ecological parameters in Lucanus cervus (Coleoptera: Lucanidae). J Morphol 276:1193- 1204 https://doi.org/10.1002/jmor.20411 Rowland JM, Emlen DJ (2009) Two thresholds, three male forms result in facultative male trimorphism in beetles. Science 323:773-776 https://doi.org/10.1126/science.1167345 Shapiro SS, Wilk MB (1965) An Analysis of Variance Test for Normality (Complete Samples). Biometrika 52:591-611 https://doi.org/10.2307/2333709 Shiokawa T, Iwahashi O (2000) Mandible dimorphism in males of a stag beetle, Prosopocoilus dissimilis okinawanus (Coleoptera: Lucanidae). Applied Entomology and Zoology 35:487-494 https://doi.org/10.1303/aez.2000.487 Simmons LW, Emlen DJ (2006) Evolutionary trade-off between weapons and testes. Proc Natl Acad Sci U S A 103:16346-16351 https://doi.org/10.1073/pnas.0603474103 Smith ABT (2006) A Review of the Family-group Names for the Superfamily

37

Scarabaeoidea (Coleoptera) with Corrections to Nomenclature and a Current Classification. Coleopt Bull 60:144-204 https://doi.org/10.1649/0010- 065x(2006)60[144:Arotfn]2.0.Co;2 Songvorawit N, Butcher BA, Chaisuekul C (2017) Different allometric intercepts in major Aegus chelifer chelifer stag beetle males from urban and forest habitats. J Asia Pac Enotmol 20:835-839 https://doi.org/10.1016/j.aspen.2017.05.011 Songvorawit N, Butcher BA, Chaisuekul C (2018) Resource holding potential and the outcome of aggressive Interactions between paired male Aegus chelifer chelifer (Coleoptera: Lucanidae) stag beetles. J Asia Pac Enotmol 31:347-360 https://doi.org/10.1007/s10905-018-9683-z Sugiura S, Yamaura Y, Makihara H (2007) Sexual and male horn dimorphism in Copris ochus (Coleoptera: Scarabaeidae). Zoolog Sci 24:1082-1085 https://doi.org/10.2108/zsj.24.1082 Taylor PW, Elwood RW (2003) The mismeasure of animal contests. Anim Behav 65:1195-1202 https://doi.org/10.1006/anbe.2003.2169 Tomkins JL, Kotiaho JS, LeBas NR (2005) Matters of scale: positive allometry and the evolution of male dimorphisms. Am Nat 165:389-402 https://doi.org/10.1086/427732 Vieira MC, Peixoto PEC, Patek S (2013) Winners and losers: a meta‐analysis of functional determinants of fighting ability in arthropod contests. Func Ecol 27:305-313 https://doi.org/10.1111/1365-2435.12051 Voje KL (2016) Scaling of Morphological Characters across Trait Type, Sex, and Environment. Am Nat 187:89-98 https://doi.org/10.1086/684159

38

Table 1. Behavioural elements of the contest interaction between males of Cyclommatus mniszechi. Behavioural Description Phase elements Touch One of the contestants touches its opponents’ body 1 with mandibles, antenna or legs. Defensive One of the contestants keeps its mandibles open 2 posture and raises its head. Body raising Two contestants face each other and raise their 2 bodies, thoraxes and mandibles with rapid movement of antennas and forelegs, and move toward the opponent. Attack One of the contestants uses its mandibles to bite 2 the opponent. Push One of the contestants raises its mandibles and 2 knocks on the opponent. Tussle Two contestants interlock each other’s mandibles 3 and push each other. Clamp 1 One of the contestants uses its mandibles to clamp 3 onto the head of the opponent. Clamp 2 One of the contestants uses its mandibles to clamp 3 onto the body (thorax or abdomen) of the opponent. Retreat One of the contestants moves backward and away 4 from the other male.

39

T able 2. The Pearson’s (r) and Spearman’s (ρ) correlation coefficient of the relationships between morphological traits of Cyclommatus mniszechi (p0.0001). ρ ML HW EL BW r ML 0.9176 0.8715 0.7375 HW 0.9512 0.9214 0.8328 EL 0.9174 0.9437 0.7889 BW 0.7934 0.8587 0.8621 ML: mandible length, HW: head width, EL: elytra length, BW: body weight.

40

T able 3. Models selection of the allometry between mandible length (ML) and elytra length (EL) in Cyclommatus mniszechi. Parameter Estimate SE P AIC ΔAIC BIC ΔBIC Linear model

β0 -4.333 0.131 <0.001 -471.06 503.32 -460.73 499.69

β1 2.591 0.049 <0.001 Quadratic model

β0 -20.250 2.297 <0.001 -513 461.38 -499.56 460.86

β1 14.698 1.745 <0.001

β2 -2.298 0.331 <0.001 Eberhard and Gutierrez continuous piecewise

β0 -5.372 0.165 <0.001 -535.14 439.24 -521.35 439.07

β1 2.997 0.063 <0.001

β2 -1.627 0.187 <0.001 Eberhard and Gutierrez discontinuous piecewise

β0 -5.367 0.188 <0.001 -533.14 441.24 -515.91 444.51

β1 2.994 0.073 <0.001

β2 -1.633 -1.633 <0.001

β3 0.001 0.019 0.95 Kotiaho and Tomkins linear

β0 1.747 0.017 <0.001 -931.40 42.98 -921.06 39.36

β1 0.356 0.007 <0.001 Kotiaho and Tomkins quadratic

β0 2.359 0.114 <0.001 -957.69 16.69 -943.91 16.51

β1 -0.156 0.094 0.0987

β2 0.105 0.019 <0.001 Kotiaho and Tomkins continuous piecewise

β0 1.860 0.023 <0.001 -974.21 0.17 -960.42 0

β1 0.305 0.010 <0.001

β2 0.202 0.029 <0.001 Kotiaho and Tomkins discontinuous piecewise

β0 1.877 0.025 <0.001 -974.38 0 -957.15 3.27

β1 0.297 0.011 <0.001

β2 0.179 0.033 <0.001

β3 0.011 0.007 0.144

41

Table 4. Parameters estimated for the isometry tests in Cyclommatus mniszechi (ML=mandible length, HW=head width, A= allometric slope). Morph Traits Coefficients Estimate ± SE 95% Lower Limit 95% Upper Limit P Major ML A 1.43 ± 0.10 1.24 1.62 <0.001* lna -1.13 ± 0.27 -1.66 -0.60 HW A 1.30 ± 0.07 1.17 1.44 0.001* lna -0.94 ± 0.18 -1.30 -0.57 Minor ML A 2.89 ± 0.10 2.69 3.10 <0.001* lna -5.12 ± 0.27 -5.65 -4.58 HW A 1.97 ± 0.06 1.86 2.08 0.001* lna -2.77 ± 0.14 -3.06 -2.49 *: p-value<0.05

42

Table 5. Multiple logistic regression models of the relationships between aggressiveness (the probability of tussle) and morphological traits in randomly matched contests of Cyclommatus mniszechi in 2017. Mandible length (ML), head width (HW), elytra length (EL) and body weight (BW) (n=30).

Tussle probability

(N=30, df=2, χ2=11.940, p=0.003)

Variables df EstimateSE χ2 p-value

Winner ML (mm) 1 1.3340.539 6.12 0.0134*

Loser ML (mm) 1 0.0340.239 0.02 0.888

(N=30, df=2, χ2=4.662, p=0.097)

Winner HW (mm) 1 0.8610.493 3.04 0.081

Loser HW (mm) 1 0.0530.285 0.03 0.852

(N=30, df=2, χ2=7.723, p=0.021*)

Winner EL (mm) 1 1.6620.747 4.96 0.026*

Loser EL (mm) 1 0.0720.427 0.03 0.866

(N=30, df=2, χ2=3.778, p=0.151)

Winner BW (g) 1 4.1222.470 2.78 0.095

Loser BW (g) 1 -0.6291.587 0.17 0.683 *: p-value<0.05

43

Table 6. The multiple regression for the correlation between contest duration and male morphs and average mandible length in the size-matched contests in Cyclommatus mniszechi. (n=55) Ln contest duration (N=55, df=2, F=6.3179, p=0.0035*) Variables df Estimate SE t ratio P-value Male morphs (major=1; 1 -0.4940.584 -0.85 0.4016 minor=0) Average ML (mm) 1 0.2380.092 2.59 0.0123* *: p-value<0.05

44

List of Figures

Figure 1. Mate-guarding of Cyclommatus mniszechi at the feeding sites of Koelreuteria elegans and Citrus spp. in the evening (a & b). Male morphs and morphological measurements of C. mniszechi (c). EL: elytra length; HW: head width; ML: mandible length.

45

Figure 2. The allometric relationship between ln-transformed mandible length (ML) and elytra length (EL) of Cyclommatus mniszechi used in the fighting contests. The linear regression lines of the major (circle) and minor (triangle) males cross at the switch point of ln elytra length at 2.71 (15.029 mm). The 95% confidence intervals are shown in grey.

46

Figure 3. The fighting arena of Cyclommatus mniszechi used in this study. (a) the overhead view, and (b) front view.

47

Figure 4. Sequential analyses of contest behaviours in randomly (left) and size- matched (right) contests of Cyclommatus mniszechi. Circles represent behaviours; circle size is scaled to the percentage of total contest behaviours. Arrows represent significant behavioural transitions; arrow width is scaled to transitional probability of the behaviours. Phases are coloured for visualization.

48

Figure 5. The ethograms and sequential analyses of behabvioural elements of Cyclommatus mniszechi in male-male fighting contests. The winner of (a) major males (b) minor males in size-matched contests. The losers of (c) major males and (d) minor males in the size matched contest. The size of circle represents the behaviour’s percentile of total behavioural events. The shade of red color of the circle from the phase 1 to phase 3 represent the aggressiveness of the behaviour.

49

Figure 6. Logistic regressions of contest behaviours in Cyclommatus mniszechi. (a) The relationship between the probability of tussle and the winner’s and mandible length in randomly matched contests (estimate±SE=1.334±0.539, Chi-square=6.12, p=0.0134). (b) The relationship between the probability of tussle and the loser’s mandible length in randomly matched contests (estimate±SE=0.034±0.239, Chi- square=0.02, p=0.888). (c) The relationship between the probability of tussle and averaged mandible length in size-matched contests (estimate±SE=0.182±0.105, Chi- square=3.01, p=0.083, n=57).

50

Figure 7. Simple linear regressions of contest behaviours in Cyclommatus mniszechi. (a) The relationship between ln-transformed contest duration and the winner’s mandible length in randomly matched contests (df=1, estimate=0.513, SE=0.148, F=11.955, p=0.002, n=30). (b) The relationship between ln-transformed contest duration and the loser’s mandible length in randomly matched contests (df=1, estimate=0.279, SE=0.109, F=6.595, p=0.016, n=30). (c) The relationship between ln- transformed contest duration and average mandible length in size-matched contests (df=1, estimate=0.174, SE=0.049, F=12.488, p=0.001, n=57).

51

Table S1. The information of Cyclommatus mniszechi specimens used in this study. ID No. Location Date Sex Status Method Host plant ML MT Note YGWD1703 21B Donghu, Yingge Dist., New Taipei City 2017/6/10 M A L 17.26 3 YGWD1701 19A Donghu, Yingge Dist., New Taipei City 2017/6/10 M A L 15.24 2 YGWD1706 Donghu, Yingge Dist., New Taipei City 2017/6/13 M A L 18.91 3 YGWD1702 13A Donghu, Yingge Dist., New Taipei City 2017/6/13 M A L 18.02 3 YGWD1710 Donghu, Yingge Dist., New Taipei City 2017/6/25 F A L -- 0 SXWD1701 Tiaogukeng Mt., Shuangxi Dist., New Taipei City 2017/6/27 M A L 17.42 3 YGWD1707 Donghu, Yingge Dist., New Taipei City 2017/7/1 M A L 12.49 2 YGWD1704 20A Donghu, Yingge Dist., New Taipei City 2017/7/1 M A L 8.22 1 SXWD1710 Pinglin, Shuangxi Dist., New Taipei City 2017/7/1 F A L -- 0 SGYWD1701 27B Chajiao, Sanxia Dist., New Taipei City 2017/7/4 M A L 14.41 2 XDWD1701 Tutan, Xindian Dist., New Taipei City 2017/7/5 M A L 9.4 1 MZWD1704 Zhinan, Wenshan Dist., Taipei City 2017/7/8 F A L -- 0 MZWD1701 25B Zhinan, Wenshan Dist., Taipei City 2017/7/10 M A L 14.27 2 YGWD1708 Donghu, Yingge Dist., New Taipei City 2017/7/11 M A L 17.31 3 M.G. for MZWD1702 Zhinan, Wenshan Dist., Taipei City 2017/7/13 M A T Unknown 13.62 2 MZWD1702 MZWD1703 Zhinan, Wenshan Dist., Taipei City 2017/7/13 F A T Unknown -- 0 SGYWD1706 31A Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 16.65 3 SGYWD1705 33B Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 17.26 3 SGYWD1704 32A Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 11.85 2 SGYWD1703 28B Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 17.46 3 SGYWD1702 26B Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 18.02 2 SGYWD13 Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 15.22 2

SGYWD12 Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 16.33 3

52

SGYWD11 Chajiao, Sanxia Dist., New Taipei City 2017/7/15 M A L 18.26 3 YGWD1705 33A Donghu, Yingge Dist., New Taipei City 2017/7/16 M A L 18.70 3 ZHWD1701 Nanshijiao Mt., Zhunghe Dist., New Taipei City 2017/7/17 M A L 17.15 3 SGYWD1707 34A Chajiao, Sanxia Dist., New Taipei City 2017/7/19 M A L 13.61 2 SGYWD15 Chajiao, Sanxia Dist., New Taipei City 2017/7/19 M A L 19.15 3 SGYWD14 Chajiao, Sanxia Dist., New Taipei City 2017/7/19 M A L 12.42 2 XDWD1702 Sishifen, Xindian Dist., New Taipei City 2017/7/26 M A L 13.73 2 SGYWD16 Chajiao, Sanxia Dist., New Taipei City 2017/7/30 M A L 8.97 1 WLWD1701 Fushan, Wulai Dist., New Taipei City 2017/8/15 M A L 14.21 2 XDWDL1810 31L Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 M L W 16.87 3 XDWDL1809 30L Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 M L W 17.20 3 XDWDL1808 32R Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 M L W 16.00 3 XDWDL1807 31R Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 M L W 16.67 2 XDWDL1801 44R Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 M L W 20.17 3 XDWDL1806 Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 F L W -- 0 XDWDL1804 Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 F L W -- 0 XDWDL1802 Xiacukeng, Xindian Dist., New Taipei City 2018/3/10 F L W -- 0 XDWDL1812 Xiacukeng, Xindian Dist., New Taipei City 2018/3/13 M L W 16.80 3 XDWDL1811 32L Xiacukeng, Xindian Dist., New Taipei City 2018/3/13 M L W 15.80 3 ZHWDL1803 30R Nanshijiao Mt., Zhunghe Dist., New Taipei City 2018/3/14 M L W 17.89 3 ZHWDL1801 Nanshijiao Mt., Zhunghe Dist., New Taipei City 2018/3/14 M L W 15.36 2 ZHWD1804 Nanshijiao Mt., Zhunghe Dist., New Taipei City 2018/3/14 M L W -- ZHWD1802 Nanshijiao Mt., Zhunghe Dist., New Taipei City 2018/3/14 M L W -- XDWDL1816 46R Sishifen, Xindian Dist., New Taipei City 2018/3/18 M L W 12.19 2 XDWDL1813 34L Sishifen, Xindian Dist., New Taipei City 2018/3/18 M L W 13.81 2

XDWDL1815 Sishifen, Xindian Dist., New Taipei City 2018/3/18 F L W -- 0 53

XDWD1817 Sishifen, Xindian Dist., New Taipei City 2018/3/18 F L W -- 0 XDWDL1818 42L Xiacukeng, Xindian Dist., New Taipei City 2018/3/19 M L W 15.55 3 ZHWD1805 Nanshijiao Mt., Zhunghe Dist., New Taipei City 2018/5/11 F L W -- 0 XDWD1802 34R Xiacukeng, Xindian Dist., New Taipei City 2018/5/21 M A W 13.94 2 XDWD1801 33R Xiacukeng, Xindian Dist., New Taipei City 2018/5/21 M A W 11.27 2 XDWD1803 33L Xiacukeng, Xindian Dist., New Taipei City 2018/5/23 M A L 10.56 2 WLWD1801 Neidong, Wulai Dist., New Taipei City 2018/5/29 M A L 6.81 1 YGWD1801 36L Donghu, Yingge Dist., New Taipei City 2018/5/30 M A L 13.51 2 XDWD1804 36R Sishifen, Xindian Dist., New Taipei City 2018/6/7 M A L 12.93 2 XDWD1806 38R Guangxing, Xindian Dist., New Taipei City 2018/6/16 M A L 9.53 1 Koelreuteria M.G. for YGWD1802 39L Asikeng, Yingge Dist., New Taipei City 2018/6/17 M A T 7.11 1 elegans YGWD1815 Koelreuteria YGWD1815 Asikeng, Yingge Dist., New Taipei City 2018/6/17 F A T -- 0 elegans Koelreuteria M.G. for YGWD1820 Asikeng, Yingge Dist., New Taipei City 2018/6/19 M A T -- 2 elegans YGWD1821 Koelreuteria M.G. for YGWD1816 Asikeng, Yingge Dist., New Taipei City 2018/6/19 M A T -- 2 elegans YGWD1817 Koelreuteria YGWD1804 44L Asikeng, Yingge Dist., New Taipei City 2018/6/19 M A T 18.69 3 elegans Koelreuteria YGWD1803 40R Asikeng, Yingge Dist., New Taipei City 2018/6/19 M A T 9.455 1 elegans Koelreuteria YGWD1822 Asikeng, Yingge Dist., New Taipei City 2018/6/19 F A T -- 0 elegans Koelreuteria YGWD1821 Asikeng, Yingge Dist., New Taipei City 2018/6/19 F A T -- 0 elegans

54

Koelreuteria YGWD1817 Asikeng, Yingge Dist., New Taipei City 2018/6/19 F A T -- 0 elegans Koelreuteria YGWD1805 Asikeng, Yingge Dist., New Taipei City 2018/6/21 M A T 9.58 1 elegans Koelreuteria M.G. for YGWD1818 Asikeng, Yingge Dist., New Taipei City 2018/6/22 M A T -- 2 elegans YGWD1819 Koelreuteria YGWD1819 Asikeng, Yingge Dist., New Taipei City 2018/6/22 F A T -- 0 elegans YGWD1813 Donghu, Yingge Dist., New Taipei City 2018/6/26 M A L 11.41 2 YGWD1812 40L Donghu, Yingge Dist., New Taipei City 2018/6/26 M A L 9.34 1 YGWD1811 41L Donghu, Yingge Dist., New Taipei City 2018/6/26 M A L 18.11 3 YGWD1810 46L Donghu, Yingge Dist., New Taipei City 2018/6/26 M A L 12.03 2 M.G. for Koelreuteria YGWD1814 YGWD1809 43L Asikeng, Yingge Dist., New Taipei City 2018/6/26 M A T 16.53 3 elegans from 2018/06/22 YGWD1808 Donghu, Yingge Dist., New Taipei City 2018/6/26 M A L 10.24 2 YGWD1807 41R Donghu, Yingge Dist., New Taipei City 2018/6/26 M A L 17.93 3 Koelreuteria YGWD1806 43R Asikeng, Yingge Dist., New Taipei City 2018/6/26 M A T 15.96 2 elegans GSWD1807 37R Guilun Mt., Guishan Dist., Taoyuan City 2018/6/26 M A T Citrus spp. 18.31 3 GSWD1806 42R Guilun Mt., Guishan Dist., Taoyuan City 2018/6/26 M A T Citrus spp. 13.71 2 GSWD1805 45R Guilun Mt., Guishan Dist., Taoyuan City 2018/6/26 M A T Citrus spp. 12.6 2 M.G. for GSWD1803 38L Guilun Mt., Guishan Dist., Taoyuan City 2018/6/26 M A T Citrus spp. 9.685 2 GSWD1804

GSWD1801 37L Guilun Mt., Guishan Dist., Taoyuan City 2018/6/26 M A T Citrus spp. 18.33 3 M.G. for

55

GSWD1802 Koelreuteria YGWD1814 Asikeng, Yingge Dist., New Taipei City 2018/6/26 F A T -- 0 elegans GSWD1804 Guilun Mt., Guishan Dist., Taoyuan City 2018/6/26 F A T Citrus spp. -- 0 GSWD1802 Guilun Mt., Guishan Dist., Taoyuan City 2018/6/26 F A T Citrus spp. -- 0 M.G. for NHWD1801 39R Niuchouwei Mt., Neihu Dist., Taipei City 2018/6/29 M A T Citrus spp. 7.92 1 NHWD18/02 NHWD1802 Niuchouwei Mt., Neihu Dist., Taipei City 2018/6/29 F A T Citrus spp. -- 0 GSWD1808 45L Guilun Mt., Guishan Dist., Taoyuan City 2018/7/2 M A T Citrus spp. 11.98 2 WLWD1802 Fushan, Wulai Dist., New Taipei City 2018/7/10 M A L 13.13 2 WLWD1803 Fushan, Wulai Dist., New Taipei City 2018/7/11 M A L 14.94 2 Koelreuteria M.G. for XDWD1808 Sishifen, Xindian Dist., New Taipei City 2018/7/12 M A T 7.65 1 elegans XDWD1809 Koelreuteria XDWS1809 Sishifen, Xindian Dist., New Taipei City 2018/7/12 F A T -- 0 elegans PLWD1801 Beiyi Rd., Pinglin Dist., New Taipei City 2018/7/12 F A L -- 0 ML: mandible length (mm); MT: mandible type (3: alpha male, 2: beta male, 1: gamma male, 0: female); status (A: adults, L: larvae); collecting method (L: light trap, W: wood chopping, T: tree sap); M.G.: mate guarding.

56