Current Zoology 56 (1): 18−35, 2010

Factors influencing the establishment of dominance hierar- chies of the grey capriscus

David W. CLEVELAND, Kari L. LAVALLI*

College of General Studies, Division of Natural Sciences & Mathematics, Boston University, 871 Commonwealth Avenue, Boston, MA 02215, USA

Abstract Unlike other balistids, grey triggerfish Balistes capriscus occur in social groups in subtropical reef assemblages and have been noted to cooperate in capturing large crustacean prey. The objective of this study were to determine the structure of dominance hierarchies of these social groups and the factors that influence hierarchies of wild-caught grey triggerfish in a natu- ralistic setting. From observations of four groups of triggerfish (n = 19 ) in both dyad and group (4 – 5 fish) settings, we pro- vide a description of triggerfish behaviors and coloration patterns and an explanation of the social context in which suites of be- haviors are used by dominant, middle-ranking, and subordinate fish. Sixteen behaviors and nine coloration patterns were noted for grey triggerfish. Grey triggerfish groups form linear hierarchies in both dyads and groups as measured by Landau’s Index of Linearity (h = 1.0 for Groups 1, 3, and 4 and h = 0.95 for Group 2 in dyads; h = 1.0 for all groups in group settings). Dyadic hier- archies, however, were not necessarily good predictors of the hierarchies found in larger group settings, as they only predicted two of the four group hierarchies. Sex played no role in influencing status or behavior. Size had the greatest influence on domi- nance status, with larger fish being more dominant than smaller fish. An individual’s dominance ranking influenced both body coloration and posture. These results suggest that color patterns and body postures may also be used by observers as an indicator of an individual’s social status in groups [Current Zoology 56 (1): 18–35 2010]. Key words Grey triggerfish, Balistes capriscus, Balistids, Dominance hierarchies, Status, Rank, Behavior

Many social systems, in which group rela- While an observed hierarchy may initially appear to be tionships may be long-term, are based on a dominance linear, individuals of middle rank may repeatedly sup- hierarchy, whereby some members of the group physi- plant each other over time, resulting in a dynamic re- cally dominate other members in a relatively orderly shuffling of ranks (de Vries, 1998). Because of these and long-lasting fashion (Wilson, 1975) to gain a dis- possible alterations in hierarchies, Drews (1993) proportionate share of available resources (Metcalfe, re-synthesized the concept of dominance as a pattern of 1986; Grant et al., 1989; Metcalfe et al., 1989; Forrester, repeated, agonistic interactions between pairs of indi- 1990; Holbrook and Schmitt, 1992; Kroon et al., 2000; viduals, with a predictable outcome in favor of the same Webster and Hixon, 2000). Such hierarchies help pro- dyad member and a lack of response from its opponent vide structure to the social system such that agonistic rather than an escalation of aggression. The consistent interactions are reduced between group members, even winner is defined as the “dominant” whereas the con- when individuals possess weapons capable of inflicting sistent loser is defined as the “subordinate”. damage upon conspecifics. Hierarchies are either linear Individual success in competitive interactions and the or nonlinear. Linear hierarchies are transitive in nature, determinants of a dominance hierarchy formation have with a top-ranking (alpha, α) individual that dominates been hotly debated (de Vries, 1998; Jameson et al., 1999) all others, a second-ranking (beta, β) individual that but are distilled down to two main ideas: intrinsic and dominates all but the alpha, a third-ranking (gamma, γ) extrinsic factors. Intrinsic factors are those traits or at- individual that dominates all others besides the alpha tributes of an individual that correlate with fighting and beta, and so on (Lehner, 1996). Nonlinear hierar- ability, such as physical size, weapon size, and/or level chies are intransitive (circular) in nature, such that the of aggression (Landau, 1951a,b; Huntingford and alpha dominates the beta, the beta dominates the gamma, Turner, 1987; Archer, 1988; Barki et al., 1991; Beaug- and the gamma dominates the alpha (Lehner, 1996). rand et al., 1991; Ranta and Lindstrom, 1992, 1993;

Received Sep. 29, 2009, accepted Dec. 07, 2009 ∗ Corresponding author. E-mail: [email protected] © 2010 Current Zoology D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 19

Drews, 1993; Huntingford et al., 1995; Nakano, 1995; Linearity, provides a measure of the degree to which a Pavey and Fielder, 1996; Rutherford et al., 1995, 1996; dominance hierarchy is linear. This test of linearity is Vye et al., 1997; Webster and Hixon, 2000). Extrinsic then used to reorganize a dominance matrix to find an factors arise not from any physical attributes of the order that is most consistent with a linear hierarchy (de animal but from the social dynamics or social context in Vries, 1998). which the animal has participated. They include such As an alternative to a rigid linear hierarchy determi- things as prior fighting experience, prior residency, or- nation, Dugatkin’s (1997) computer model examining der of introduction to the group, and presence of by- winner and loser effects on the development of domi- standers during previous encounters. These are gener- nance hierarchies demonstrated that, when winner ef- ally termed winner and loser effects (Parker, 1974; fects alone were important, individual rank was clearly Rubenstein and Hazlett, 1974; Burk, 1979; Chase, 1982; defined. When only loser effects were important, how- Figler and Einhorn, 1983; Franck and Ribowski, 1987; ever, a clear alpha emerged, but the ranks of the subor- Evans and Shehadi-Moacdieh, 1988; Chase et al., 1994; dinate members were often unclear due to a lack of ag- Peeke et al., 1995; Beaugrand et al., 1996; Hsu and Wolf, gressive interactions. If, however, individuals are capa- 1999; Dugatkin, 2001). Both intrinsic and extrinsic fac- ble of assessing their own fighting abilities relative to tors can operate simultaneously (e.g., Chase et al., 2002), those observed for other group members by acting as particularly when differences in intrinsic factors are bystanders during encounters between other group small (Webster and Hixon, 2000). However, it may be members, then a more complex picture arises, as by- more difficult to determine the relative importance of standers change their assessment of the protagonists' extrinsic factors in determining a hierarchical pattern, fighting abilities along with their own assessment of especially when examining established social groups. In their ability to defeat the protagonist(s). In these cases, such groups, behavioral patterns expressed by group the assessment of rank within small groups (e.g., 5 indi- members may be the result of prior interactions that viduals) becomes increasingly difficult and nonlinear established the hierarchical pattern and successive be- (Dugatkin, 2001) and is subject to frequent changes, haviors. particularly within middle ranks. Most methods for determining hierarchy structures We investigated the relative importance of intrinsic are based on laboratory studies investigating dominance individual characteristics (size and sex) on the estab- relations between pairs (“dyads”, c.f., Drews, 1993) or lishment of dominance hierarchies of wild-caught grey triads of , rather than small groups (which would triggerfish Balistes capriscus in outdoor, naturalistic, provide dominance ranking information). Recording the large arenas. These fish are commonly found on both outcome of competition over resources, or types of con- sides of the Atlantic and are very common in the Gulf of tact upon initial and repeated meetings, for each pair in Mexico (Samuelson and Einarsen, 1996); they inhabit a group reveals that one individual often supplants all tropical to semi-tropical waters and are important others, while middle-ranking and bottom-ranking indi- members of reef fish assemblages. Their preferred prey viduals displace each other in an intransitive manner. consists of bivalves, , and sand dollars (Vose However, such studies may be misleading, particularly and Nelson, 1994; Kurz, 1995), but they are capable of in cases where both social (extrinsic) and physical (in- feeding on gastropods, urchins, crustaceans, and other trinsic) attributes play a role in hierarchy formation hard-prey items. Social groups are rare for balistid (Basquil and Grant, 1998; Beaugrand and Goulet, 2000) triggerfish, as most patrol territories and are aggressive such that fitness benefits and costs of social rank are towards conspecifics; however, groups of grey trigger- overlooked (Sloman and Armstrong, 2002). fish, highly variable in size, can be found in abundance Thus, some studies have examined groups larger than on reefs in subtropical waters (Vose and Nelson, 1994). two or three individuals. In such studies, a matrix is In studies on the benefits of grouping behavior or the used to simplify the relationships among the individuals benefits of weaponry versus defensive posturing in (Jameson et al., 1999), and to derive a dominance hier- spiny, clawed, and species, grey trigger- archy of the group (Martin and Bateson, 1993). The fish have been observed cooperating to subjugate spiny, dominant individual is placed at the top of the matrix, clawed, and slipper lobsters via coordination of attack while the most subordinate individual is at the bottom. (Herrnkind et al., 2001; Lavalli and Spanier, 2001; Bar- Linearity may be determined using one of several shaw et al., 2003). In spite of this cooperation, only one available models. One such model, Landau’s Index of fish appeared to consume the majority of the prey ob- 20 Current Zoology Vol. 56 No. 1 tained, while driving off other fish that had participated with sloped sides) of the KML or to the outdoor corrals in coordinated attack and ultimate subjugation (Lavalli at the NMFS. They were then netted and measured to et al., 20001). the nearest 1 mm standard length (snout to caudal pe- To better understand the group dynamics of the grey duncle) and randomly tagged through the caudal pedun- triggerfish, we examined their behaviors in several dif- cle just below the with colored Floy Tags ferent social settings. Specific questions addressed were: (white, red, yellow, or blue). One fish in each group was (1) What is the dominance structure of a triggerfish not tagged but pierced through the caudal peduncle in group? (2) Is the hierarchy linear or circular?; (3) Does the same position and manner as the tagged fish. the size and/or sex of the fish influence the dominance After tagging, all fish were isolated from each other status? (4) Is the group dominance structure related to both physically and visually for 48 h. At the KML facil- the dominance status of an individual within a dyad? (5) ity, oblong isolation pens were constructed of black What behavioral predictors are present that communi- plastic, diamond mesh aquaculture netting (3/4 cate the social status of an individual within a dyad or in.-diameter), and cable ties. The black plastic was larger group? Finally, we considered the direct conse- placed along the ends and middle of the divided pens to quences of the hierarchy on individual fish within the prevent physical and visual interaction of the fish. These group. pens were then placed in the non-experimental sections 1 Materials and Methods of the KML runways. At the NMFS facility, isolation pens consisted of a fenced-in section 6 m long × 1 m 1.1 Collection and handling of animals wide × 1 – 1.5 m deep (due to tidal fluctuations) adja- Members of intact schools of grey triggerfish were cent to the observation platform and north of the obser- fished from St. John’s Bay, Florida (29º48' N 84º26' W), vation sections. This 6 m-long section was divided into U.S.A. in the northeastern Gulf of Mexico via hook and six compartments 1 m long × 1 m wide × 1 –1.5 m deep. line, using squid as bait. The depth from which the fish Three individual fish from one group were placed into were reeled (~ 20 m) resulted in their swimbladders every other compartment to prevent visual and physical expanding, and inserting an 18-gauge needle into the interactions among individuals of the same group; a lateral post-opercular surface to penetrate the swim- second set of three individuals from another group were bladder relieved this expansion. All fish appeared unin- placed into the empty compartments and thereby sepa- jured following this procedure, and all resumed normal rated from each other visually and physically. The re- swimming activities immediately following the penetra- maining four fish (two from each group) were isolated tion of the swimbladder. The fish were then transported by placement into opposite sections of two 2 to the Florida State University Marine Laboratory m-diameter pens, divided into quarters. These pens were (FSUML) in Sopchoppy, Florida and held in outdoor placed approximately 3 – 4 m apart south of the obser- pens with ambient, flowing seawater for several weeks vation arenas. (temperature = 23 – 25 ºC; salinity = 30 – 32 ppt). After All fish groups were composed of five individuals this recovery period, they were transported via a live initially. In one of our fish groups, however, one of the fish transport container (~ 122 cm long × ~ 122 cm wide original five fish became sickly shortly after transporta- × ~ 124.5 cm high) with aeration to either the Keys Ma- tion from the FSU Marine Laboratory in Sopchoppy to rine Laboratory (KML) on Long Key, Florida or to the the NMFS research center in Panama City. Thus, this National Marine Service (NMFS) facility in fish was not used in pair-wise interactions or in later Panama City, Florida. During holding and transport group interactions, and in this particular group, only times, the triggerfish were held in their original groups four fish constituted the final grouping. of five fish total to maintain prior group ties and hierar- 1.2 Experimental conditions chies. All fish were fed a daily diet of frozen squid or The experimental facility at the Keys Marine Labo- crustaceans. ratory consisted of a large concrete-lined waterway Upon reaching their final destination, the fish groups (~ 300 m) supplied with water from the adjacent Florida were allowed two days of acclimation to runways (300 Bay. This runway was partitioned into sections, one of m-long, concrete-lined, continuously flowing waterways which measured 5 m long × 5 m wide × 1 m deep and

1 Lavalli KL, Cleveland D, Herrnkind W, Childress M, 2000. Attack strategies of grey triggerfish Balistes capriscus pitted against solitary Caribbean spiny lobsters Panulirus argus. Abstracts of the 30th Annual Marine Benthic Ecology Meeting, University of New Hampshire, Durham, NH, 15–18 March 2001, p. 106. D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 21 acted as an observation arena for the fish dyads and other fish away from it; the behaviors and colorations groups. A 2 m-high platform was placed at one end of that such fish displayed while dominating the food re- this arena to allow observations and videotaping of in- source were considered more agonistic and were scored teractions with no disturbance to the fish. Trials were as such as part of the determination of rank (see section run on two separate groups of triggerfish in summer 1.5). Once every possible pairing of fish was completed, months from June to July, 2001. Water temperatures the pairings were randomly repeated for a second run to ranged from 29.5 to 34 °C in this arena. Large caged confirm the dyad hierarchy. corrals located at the National Marine Fisheries Service 1.4.2 Group hierarchies Observations of the dyad research center in St. Andrew Bay, Panama City, FL hierarchy acted as a control for the group study. Group were used for observations of two fish groups in win- observations began after the establishment of the dyad ter/spring months from March to April, 2002. Water hierarchy was confirmed via the second set of pairings. temperatures ranged from 14.5 to 22 °C. These corrals Fish comprising the study group were placed into the were partitioned to match the size of the observation observation arena (where they stayed for the remainder arena at the Keys Marine Laboratory (5 m wide × 5 m of the experiment) and observed ad libitum for 1 h. Fish long × 1 – 1.5 m deep). Observations and video re- were then fed from the observation platforms, and ad cordings were conducted from a 2 m-high observation libitum observations continued for an additional hour. platform (adjacent dock) to minimize disturbance to the All interactions between fish were video-recorded fish. (SONY DCR-TRV9 camera), and vocal descriptions of 1.3 Ethograms of behaviors displayed during the behaviors were also recorded either directly onto the conspecific encounters videotapes or onto a microcassette recorder. The group Grey triggerfish behaviors and coloration patterns observations were repeated daily until the fish hierar- were catalogued during dyad interactions. These behav- chy remained stable for a period of three consecutive iors and colorations were described in all contexts and days. A stable hierarchy was considered to be one in were scored after the winner of the encounter had been which a particular pattern of behaviors was maintained determined. Winner behaviors and colorations were over time (c.f., Nelissen, 1985). At the conclusion of scored as dominant patterns, whereas loser behaviors the group trial, individual fish were sacrificed for sex- ing by gonadal examination (FSU Animal Care and and colorations were scored as submissive patterns. Use Committee Protocol #0105, “Lobster/Triggerfish Where possible, photographs of the behaviors and/or Behavior”). colorations were obtained. 1.5 Data Analysis 1.4 Hierarchical determination Four replicate trials were conducted with fish from 1.4.1 Dyad hierarchies After the isolation period, four different social groupings caught from the north- random pairs of tagged triggerfish were introduced into eastern Gulf of Mexico. Videotapes were analyzed to the observation arena and observed for 30 min. Only produce an ethogram of dominant, subordinate, and two pairings of fish per group per day were possible neutral behaviors and coloration patterns. After testing (four of the five total fish within a group) to allow fish for homogeneity between pre-food and post-food dyad to recover from their interactions before pairing them trials, these trials were combined. Three 5 × 5 and one 4 with other fish; one fish per group was not observed on × 4 chi-square contingency tables (for Groups 1, 2, and any given day. Observations consisted of cataloguing 4, and Group 3, respectively) were used to determine if aggressive interactions, displays, and responses (e.g., the frequency of colorations displayed was independent behaviors, trigger and body positions, and coloration of the status of the fish for each of the four groups of patterns) for 15 min before feeding and 15 min after fish (Zar, 1999). Dyadic hierarchies were then examined feeding. Interactions between individuals remained fre- by scoring the frequency of agonistic behavior and col- quent if feeding occurred mid-observation; otherwise, oration by a particular fish (the winner or loser of the after an initial period of interaction between two fish, food resource) and constructing a dominance matrix. the fish tended to decrease their activities and interac- This matrix was tested for linearity using Landau’s In- tions. During the feeding period, one or two pieces of dex of Linearity (h): food (squid) were introduced into the observation arena n 2 from the observation platform. Fish were considered ⎛⎞12 ⎡⎤vna −−(1) h = ⎜ ⎟ ⎢⎥ ⎜ 3 ⎟∑ ⎢⎥ dominant if they secured the food resource and kept the ⎝⎠nn− a=1 ⎣⎦2 22 Current Zoology Vol. 56 No. 1 where: sis. The significant level for all tests was α=0.05. n = number of animals in the group 2 Results va = number of animals that individual ‘a’ dominates This index ranges from 0 to 1.0, with an h value of 2.1 Behavioral ethograms 1.0 indicating perfect linearity. Values of h greater than Sixteen behaviors and eight color patterns were de- 0.9 denote a strong linear hierarchy. Each individual’s fined from observations of 19 individual triggerfish linearity index value is its rank in the hierarchy. How- within the four separate groups. These behaviors and ever, if two individuals were assigned an identical index color patterns can occur simultaneously and may, there- value based on Landau’s Index of Linearity, it would be fore, convey different meanings. Nevertheless, some problematic to determine individual rank within the behaviors and colors typified dominant or subordinate group. Thus, we used an individual’s calculated domi- individuals, and behaviors and colorations could be di- nance index (0 – 1.0) to sort the fish according to their vided into agonistic, subordinate, or neutral displays. alpha to omega position within the group hierarchy. 2.1.1 Agonistic behaviors The behaviors that most Behavioral data analyses were performed using pub- often typified aggression included biting, chasing, and lic-domain Java Applets for the analysis of behavioral other physical contact such as veering into another fish. data (available on the Internet at http://caspar.bgsu.edu/ Biting involved one fish making physical contact with ~software/java/1hierarchy.html) developed by Hemel- another via teeth, resulting in the attacked individual rijk (1990). usually fleeing to avoid physical injury. Aggressive fish The hierarchies resulting from the dyadic pairings also displayed chasing behaviors that could be were compared to those determined for the group, for short-term (linear) or long-term (involving circular which the frequency of agonistic behavior and colora- repetitions). A simple Chase (Fig. 1A) consisted of one tion by a particular fish (winner and losers of the food fish quickly approaching another and pursuing it as it resource) was scored and used in the dominance matrix. fled. A Circle-chase was an escalated version of a sim- The same dominance structure applet was used to cal- ple chase (Fig. 1A-C) involving two fish swimming in culate the group Landau’s Index of Linearity and domi- repetitive circles until the chased individual exhibited nance index to determine the specific ranking of an in- some other submissive behavior or coloration pattern dividual in a group setting. Specific rankings were then (such as the Attack Inhibition Display described below used to assign fish into the categories of dominant (in- and illustrated in Fig. 1C) besides fleeing. Chases could dex value 1 – 0.6), middle-ranking (index value 0.59 – result from a prior behavior, such as Veer-into (Fig. 1B), 0.4), and subordinate (index value 0.39 – 0). The fre- in which the aggressor made a sudden rush towards an- quency of behaviors displayed by dominant (α, β), mid- other fish during an otherwise seemingly passive situa- dle-ranking (δ), and subordinate (γ, ω) individuals was tion. Veer-into could also lead to biting by the aggressor tallied for individuals in all groups, arcsin-transformed, and flight or Attack Inhibition Display by the attacked and compared via a two-factor ANOVA (with rank and fish. sex as the factors) to determine if mean frequency of all 2.1.2 Submissive behaviors Submissive behaviors behaviors (agonistic, neutral, and submissive) differed involved either action, such as flight, or positioning of among the different ranks and between the sexes of fish the body in a specific vertical plane and a change in (Zar, 1999). Where significant differences were found, color (see Fig. 2A). Flight (Flee) involved quickly Bonferroni-Dunn post-hoc tests were used to determine swimming away from an approaching fish to reduce which ranks and/or sexes differed in the behavior. Simi- potential harm or to stop a chase sequence. Subordinate larly, the frequency of different coloration patterns dis- body positions included Head-down (Fig. 2B), in which played was also tallied for dominant, middle-ranking, the subordinate fish positioned its head at a downward and subordinate individuals, arcsin-transformed, and angle, and Head-up (Fig. 2C), in which the head posi- compared via a two-factor ANOVA to determine if tion was at an upward angle from a horizontal plane. In mean frequency of colorations differed among the dif- contrast to a normal feeding position, a head-down dis- ferent ranks and between the sexes of fish. Bon- play was seen in conjunction with some combination of ferroni-Dunn post-hoc tests were used to determine a Trigger-up display, light subordinate coloration, and a which ranks and/or sexes differed in the coloration pat- Hover. Likewise, head-up could include one or all of terns expressed. Size of fish was examined as a potential these additional displays. If all three conditions existed factor in the determination of rank by regression analy- with head-down or head-up, this combination of dis- D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 23 plays described an Attack Inhibition Display or AID are most likely to continually use the trigger-up display, (Fig. 2D; Fig. 3A-C). often as part of an AID. It is notable that the dorsal The trigger-up display (Fig. 2D) involved an upward trigger spine has not been observed as a weapon but and forward movement of the most anterior mostly as a tool for communication. In contrast, in a spine. This modified spine, for which triggerfish have trigger-down display (Fig. 2E), the most anterior dorsal acquired their name, has many functions, e.g., fin spine is lowered after being in an up position. Nor- self-defense in the event of being swallowed or, when mally, the trigger is down when swimming, hovering, or used in conjunction with a smaller ventral spine, resting inactive. Alpha fish rarely displayed trigger-up followed or sleeping while wedged in reef structures. A solitary by trigger-down to subordinate fish unless seriously fish with a trigger-up display is often observed to be challenged. involved in an activity that requires some amount of An Attack Inhibition Display describes a submissive focus on the task at hand, e.g., predatory feeding be- behavior or sequence of submissive behaviors designed havior or being alert to a possible threat in its vicinity. to avoid confrontation and unnecessary energy expen- In a paired or group situation with conspecifics, the diture in defending one’s place in a dominance hierarchy. trigger-up position signals subordination. In a poten- An alpha fish never displayed an AID. The AID fre- tially escalating situation in the presence of a dominant quency increased with greater levels of subordination, fish, a subordinate fish will often raise its trigger either as evidenced by the increasing frequency of pale colora- partially or fully. The α-ranked fish rarely raised its tion, trigger-up, head-up, or head-down behaviors. It trigger in an established dominant/subordinate relation- was most often displayed while in a hover position ship. Middle-ranking individuals may or may not raise (usually) perpendicular to the dominant fish for which their trigger, depending on the situation and the stage in the display was meant and while using colorations of the establishment of the relationship. Subordinate fish speckled, grey, or white (Fig. 3A-C).

Fig. 1 Sequence of behaviors involved in a typical Chase scene A. Approach of aggressor fish. B. Rapid swimming as the approached fish flees and the aggressor Chases. C. Veer-into by aggressor fish.

Fig. 2 Fish behaviors A. Top, right fish (darker) Veers-into bottom, right fish (white coloration). B. Left fish displays Head-down in conjunction with White coloration and Trigger-up. C. Right fish displays Head-up. D. Trigger-up display in conjunction with White coloration. E. Trigger-down behavior. F. Resting fish displaying Flatten behavior with Mottled coloration. 24 Current Zoology Vol. 56 No. 1

Fig. 3 Sequence of behaviors involved in a typical Attack Inhibition Display A. Approach of dominant fish from above (out of view) while a middle-ranking fish (bottom) Passes a subordinate fish that is displaying Trigger-up behavior and an off-White coloration. B. Subordinate fish intensifies blanching to achieve a pure White coloration and repositions body into a Head-down posture in response to approaching dominant fish. C. Subordinate fish intensifies Head-down posture while dominant fish examines it.

2.1.3 Neutral behaviors Several behaviors were approached a prey item and is determining where to seen in all fish, regardless of rank. When in pairs or in strike the prey (Lavalli and Spanier, 2001), but it was group settings, all fish typically Approach other indi- also used in conjunction with an AID by subordinate viduals. Generally, an approach towards another fish is fish. Backward-swimming, swimming in reverse, usu- benign, but it could be a prelude to a chase or veer-into. ally occurred when subordinate fish attempted to get out An approach may or may not elicit a response from ei- of the way of an approaching dominant fish. It was also ther party, but typically it elicited a passing behavior in often used in conjunction with hover while feeding. Fish which the approaching fish would simply examine the also displayed a Blowing-water behavior in which they other fish and swim past it (Pass). Another behavior would forcibly blow water at an object (usually food) to performed by all fish was Hover, which consisted of flip it over or blow sediment off it. After gaining control fluttering the dorsal and anal fins to remain suspended over a food item, fish could display a Wobble behavior in place. Hovering typically occurs after the fish has in which they rocked side-to-side while swimming

Fig. 4 Fish coloration patterns A. Top left fish displaying Light banding coloration. B. Dark banding coloration during feeding in conjunction with Head-down and Trigger-up behaviors. C. Upper fish displaying Grey coloration with a partial Trigger-up behavioral display to lower fish that is also displaying Trigger-up. D.Upper fish displaying Black coloration. E. White (blanching) coloration in conjunction with Trigger-up behavior. F. Mottled coloration. G. Speck- led coloration. H.Upper fish displaying Olive coloration in conjunction with Head-down and a partial Trigger-up behavioral display. D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 25 slowly. Flatten (Fig. 2F) was used in two contexts: when A Dark-banded coloration was observed in all fish resting during the day on the bottom of the mesocosm engaged in frenzied feeding or a predatory attack mode and when swimming over prey with weaponry (e.g., and was an amplification of the bands seen in the spiny lobsters) to reach a better angle for attack (Lavalli light-banded display (Fig. 4B). Otherwise, in a group and Herrnkind, 2009). When resting, flatten was always situation, α-ranked fish used it as coloration pattern ~2 combined with a mottled coloration pattern. Fish also to 20% of the time (Fig. 5A), while middle-ranking fish engaged in schooling behavior when searching for food displayed dark banding to a more dominant fish ~17 to or exploring their environment. 58% of the time (Fig. 5B). The dark-banded coloration 2.2 Coloration ethograms was used by the lowest ranked fish ~7 to 33% of the B. capriscus is typically found as a solitary fish and time (Fig. 5D), although in one of our fish groups it was displays a Light-banded coloration (Fig. 4A). In this the primary (69% of the time) color pattern of the coloration, most of the fish is a grey speckled color, but ω-ranked fish (Fig. 5D). several (two to three) bands of darker grey appear from A solid Grey coloration (Fig. 4C) was often displayed the dorsal edge of the body near the trigger and dorsal momentarily in conjunction with an AID by a subordi- fin and continue down the sides of the body no further nate fish as a dominant fish approached and passed by. than the level of the pectoral fins. The belly is typically High level middle-ranking (β and γ) fish used this col- white. In this set of experiments, the light-banded col- oration pattern ~0 to 37% of the time (Fig. 5B), while oration was displayed by all fish in a “normal” state, i.e., lower level middle-ranking (δ) fish used it 0 to 33% of not harassed or otherwise occupied. In a group situation, the α-ranked fish remained light-banded in most situa- the time (Fig. 5C). The ω-ranked fish used this pattern tions, except feeding and resting, when it could take on a ~7 to 63% of the time (Fig. 5D). By contrast, α-ranked dark-banded or mottled coloration, respectively. The fish never used this color pattern. Likewise, White (Fig. light banded coloration used 77 to 98% of the time, was 4E) was never used by α-ranked fish, but was used by all most commonly displayed by individuals of high hier- other ranks: 2 to 22% by high middle-ranking (β and γ) archical position (Fig. 5A). fish (Fig. 5B), 2 to 38% by low middle-ranking (δ) fish

Fig. 5 Color patterns of fish in each group (A –D), illustrating differences in color pattern usage among highest-ranking fish (α), higher middle-ranking fish (β and γ), lower middle-ranking fish (δ), and lowest-ranking fish (ω) Color patterns are not independent of rank (χ2 analysis, P < 0.001 for all groups of fish). Blk = Black, DkB = Dark banding, G = Grey, LtB = Light banding, S = Speckled, and W = White. 26 Current Zoology Vol. 56 No. 1

(Fig. 5C), and 2 to 48% by ω-ranked fish (Fig. 5D). based on a dyad analysis. The behaviors and coloration Four rare coloration patterns were Black (Fig. 4D), patterns observed during the dyad observations (de- Mottled (Fig. 4F), Speckled (Fig. 4G), and Olive (Fig. scribed above) were categorized into agonistic (e.g., 4H). Black, a dull charcoal coloration, was seen mostly Bite, Chase, Circle chase, Veer-into, Light-banded, Dark in middle-ranking fish (0 to 9% of the time) (Fig. 5B, C) banded, Black), submissive (e.g., Flee, Head-down, and very rarely in ω-ranked fish (Fig. 5D). Speckled, an Head-up, Trigger-up, Grey, White, Speckled), and neu- intermediate coloration between grey and white with tral (e.g., Approach, Pass, Flatten) displays and tallied iridescent flecks of blue or green, was seen in mid- for each fish pair interaction to determine dominant or subordinate status. Observations of the behaviors dis- dle-ranking (0 to 17% of the time, Fig. 5B, C) and played by paired fish were used to construct a domi- ω-ranked fish (~2% of the time, Fig. 5D) and was used nance matrix for each group of fish (Table 1) by scor- during an AID or while schooling. Mottled, a pattern ing the average frequency of agonistic behaviors/color similar to dark banded but with disrupted bands, and patterns of the fish during the four 15-min observation olive, a solid light brown, were seen less than 1% of the periods per pairing. The dominance matrix was tested time in these fish and were used when the fish were rest- for linearity via Landau’s Index of Linearity (h), and ing in a flatten position or when schooling, respectively. all ties in rank were sorted via a dominance index cal- 2.3 Hierarchical determinations culation for each individual (as per the Java Grinder 2.3.1 Dyad hierarchies Paired interactions between program of Dr. Robert Huber of Bowling Green Uni- fish comprising a full group were undertaken before, versity, Ohio, USA, at http://caspar.bgsu.edu/ during, and after feeding to determine the hierarchy %7esoftware/Java/1Hierarchy. html).

Table 1 Dominance matrix constructed for fish during dyad interactions Fish code NT W B R Y Totals D.I. Rank Sex Size (SL) Group 1 NT-1 — 96 98.2 97.3 94.2 385.7 0.651 α M 30.7 W-1 48.68 — 98.73 78.95 90.79 317.15 0.577 β M 27.3 B-1 72.45 86.36 — 95.04 90.9 344.75 0.561 γ M 26.0 R-1 37.74 39.31 17.48 — 59.49 154.02 0.331 δ F 25.2 Y-1 47.11 10.34 55.36 38.65 — 151.46 0.311 ω F 22.5 Totals 205.98 232.01 269.77 309.94 335.38 1353.08 Group 2 NT-2 — 75 37.5 54.54 100 267.04 0.428 δ F 21.1 W-2 100 — 100 57.89 76.6 334.55 0.498 γ F 23.6 B-2 100 100 — 77.77 67.64 345.41 0.576 β F 21.9 R-2 100 86.95 83.33 — 100 370.28 0.66 α M 25.1 Y-2 55.55 74.07 33.33 0 — 162.95 0.321 ω F 22.8 Totals 355.55 336.02 245.16 190.2 344.3 1480.23 Group 3 NT-3 — 93.75 100 100 293.75 0.706 α F 28.6 W-3 48 — 100 40 188 0.492 γ M 25.2 B-3 7.14 0 — 26.66 33.8 0.108 ω F 21.3 Y-3 66.66 100 77.84 — 244.5 0.594 β F 27.6 Totals 121.8 193.75 277.8 166.66 760.05 Group 4 NT-4 — 35.48 13.04 26.92 38 113.44 0.274 ω M 23.8 W-4 68.18 — 34.32 37.5 0 140 0.384 δ M 25.4 B-4 32.14 38.37 — 15.25 0 85.759 0.29 γ M 24.4 R-4 100 50 61.76 — 18.18 229.94 0.57 β F 27.0 Y-4 100 100 100 93.33 — 393.33 0.875 α M 33.7 Totals 300.32 223.85 209.12 173 56.18 1353.08 Frequency of dominant behaviors scored to determine dominance index for each individual fish and to assess rank. (df = 12, n = 20 for Groups 1, 2, and 4; df = 6, n = 12 for Group 3). Dominance index (D.I.) and Landau’s Index of Linearity calculated by the program at http://caspar.bgsu.edu/ %7esoftware/Java/1Hierarchy.html. SL = Standard length in cm. D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 27

Despite attempts to increase interactions by feeding to the γ position (higher rank), while B-2, scored as the midway through the 30-min observation period of the second-most dominant fish (β) in the dyad interactions, dyads, frequency of interactions varied greatly among moved to the lowest rank (ω) in the group hierarchy. paired individuals and groups. Group 1 fish displayed a Similarly, NT-4 moved from the lowest rank (ω) in the high frequency of agonistic interactions, particularly dyad interactions to the second-highest rank (β) in the chasing behavior, and a low frequency of trigger-up group interactions, while B-4 moved from the γ position behavior. Likewise, Group 3 fish were highly interactive, to the ω position and R-4 moved from the β position to with a high frequency of veer-into and chasing behav- iors in the dominant fish and high frequencies of trig- the γ position (Table 2). Hierarchies in all groups were ger-up and head-down body positions in the subordinate found to be perfectly linear (h = 1). fish. Group 2 fish had a low frequency of interactions 2.4 Effects of size and sex on behavior, colora- between individuals within a pairing, but clear submis- tion, and hierarchical rankings sive behaviors were present that allowed ranking of the Following the behavioral observations, the fish were subordinates. In Group 4, neutral behaviors were the sacrificed, measured for standard length, and sexed. most commonly displayed, but dominant individuals Size influenced ranking of fish in dyad trials such that chased more frequently, while subordinates displayed the largest fish in each pairing always occupied the head-up or head-down body positions with white or highest rank (α). However, size did not always influ- grey colorations patterns. Coloration patterns displayed ence the position of the middle and lower ranks (Fig. by fish in these pairings were not independent of fish 6A), such that the most subordinate fish (ω was not al- rank (χ2 = 160.27, df = 12, P < 0.001 for Group 1; χ2 = ways the smallest fish. Despite this, regression analysis 156.63, df = 12, P < 0.001 for Group 2; χ2 = 443.98, df demonstrated a strong relationship between rank and = 6, P < 0.001 for Group 3; and χ2 = 252.898, df = 12, P size (r2 = 0.716, P < 0.001; Table 3). Likewise, size was < 0.001 for Group 4; see Fig. 5A-D). Dominant fish related to rank in the group setting (Fig. 6B), but not as were more often light-banded and dark-banded than strongly as in the dyad setting (r2 = 0.5912, P < 0.001; subordinate fish. Subordinate fish showed more white, Table 3). The largest fish was always the dominant (α), dark-banding, gray, and black and less light-banding. but the most subordinate fish (ω) was not always the The hierarchies for Groups 1, 3, and 4, based on the smallest fish. dyadic pairings, was perfectly linear (h = 1), while that Both behaviors and coloration patterns were exam- for Group 2 was highly linear (h = 0.95). ined via a two-factor analysis of variance (ANOVA) to 2.3.2 Group hierarchies In the same manner that determine if either rank or sex (or both) influenced the behaviors and colorations defining dominant and subor- frequency of displays by a particular fish (see Tables 4 dinate were determined in dyad observations, and 5). Frequency data were arcsin-transformed prior to group data were compiled from Day 1 pre-food obser- ANOVA analysis to adjust the data to a normal distribu- vations, which occurred immediately after fish from a tion. All fish from all groups were divided into dominants particular group were first placed together in the obser- (index of 1 – 0.60), middle-ranking individuals (index of vation arena. Again, behavioral and coloration patterns 0.59 – 0.40), and subordinates (index of 0.39 – 0). provided values for dominance matrices so that group Fish rank had a significant effect on the expression of dominance hierarchies could be compared to dyadic approach behavior (two-factor ANOVA, F2, 13= 19.71, P dominance hierarchies to determine if dyadic pairings < 0.001, Fig. 7). While there was a trend for an interac- were a good predictor of dominance patterns within a tion between rank and sex, this trend was not significant group of triggerfish. In the case of Groups 1 and 3, the (two-factor ANOVA, F 2,13= 2.851, P = 0.094) and is dyadic interactions were a good predictor of the struc- likely the effect of one sex being more common in each ture of the hierarchy. While the dominance index values of the groups of fish. Post-hoc tests showed that domi- for a particular individual changed, the actual rankings nant fish approached more frequently than did mid- did not (Table 2). In contrast, for Groups 2 and 4, the dle-ranking and subordinate fish (Bonferroni-Dunn, P = dyadic interactions were a good predictor of the 0.004 and P < 0.001, respectively), and middle-ranking α-ranked fish only; the dyadic interactions did not ac- fish approached more frequently than did subordinate curately predict the hierarchy for middle-ranking and fish (Bonferroni-Dunn, P < 0.011). Fish rank also had lower-ranking fish. For example, NT-2 was scored as a an effect on the expression of pass behavior (two-factor subordinate individual (δ) from the dyad data but moved ANOVA, F2,13= 7.645, P < 0.01, Fig. 7), with dominant 28 Current Zoology Vol. 56 No. 1

Table 2 Dominance matrix constructed for fish during group interactions

Fish code NT W B R Y Totals D.I. Rank Sex Size (SL)

Group 1

NT-1 — 100 100 95 100 395 0.897 α M 30.7

W-1 11.36 — 90.91 60 100 262.27 0.659 β M 27.3 B-1 27.27 27.5 — 85.71 88.89 229.37 0.531 γ M 26.0 R-1 6.452 7.692 11.54 — 16.67 42.354 0.149 δ F 25.2 Y-1 0 0 0 0 — 0 0 ω F 22.5 Totals 45.08 135.19 202.45 240.79 305.56 928.99 Group 2 NT-2 — 61.54 100 40 75 276.54 0.566 γ F 21.1 W-2 76.92 — 82.35 27.78 100 287.05 0.573 β F 23.6 B-2 35 24 — 33.33 11.11 103.44 0.212 ω F 21.9 R-2 83.33 89.47 100 — 100 372.8 0.782 α M 25.1 Y-2 16.67 38.46 100 2.381 — 157.51 0.355 δ F 22.8 Totals 211.92 213.47 382.35 103.49 286.11 1197.34 Group 3 NT-3 — 96.88 100 100 292.12 0.74 α F 28.6 W-3 47.37 — 82.35 60 189.72 0.49 γ M 25.2 B-3 0.602 0 — 0 0.602 0.002 ω F 21.3 Y-3 54.29 100 50 — 204.29 0.568 β F 27.6 Totals 102.26 196.88 232.35 155.24 686.73 Group 4 NT-4 — 85.71 100 100 100 385.71 0.61 β M 23.8

W-4 71.43 — 50 25 0 146.43 0.352 δ M 25.4

B-4 0 0 — 0 0 0 0 ω M 24.4

R-4 75 83.33 77.78 — 28.57 264.68 0.54 γ F 27.0

Y-4 100 100 100 100 — 400 0.756 α M 33.7

Totals 246.43 269.04 327.78 225 128.57 1196.82 Frequency of dominant behaviors scored to determine dominance index for each individual fish and to assess rank. (df = 12, n = 20 for Groups 1, 2, and 4; df = 6, n = 12 for Group 3). Dominance index (D.I.) and Landau’s Index of Linearity calculated by the program at http://caspar.bgsu.edu/%7esoftware/Java/1Hierarchy.html. SL = Standard length in cm.

Table 3 Regression analysis for the effects of size on ranking in dyads and groups

df Sum of squares Mean square F-value P-value

Dyad trials

Regression 1 27.872 27.872 42.780 0.0001

Residual 17 11.076 0.652

Total 18 38.947

Group trials

Regression 1 23.027 23.027 24.589 0.0001

Residual 17 15.920 0.936

Total 18 38.947 D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 29

Fig. 6 Fish size versus fish rank in (A) dyad and (B) group interactions Slopes are significantly greater than zero (P < 0.0001). Sizes are coded as 5 = largest, 4 = next largest, 3 = middle-sized, 2 = second smallest, 1 = smallest.

Table 4 ANOVA tables for effects of rank and sex on behavior df Sum of squares Mean square F-value P-value Approach Fish Rank 2 0.294 0.147 19.714 0.0001 * Fish Sex 1 0.001 0.001 0.125 0.73 Fish Rank × Fish Sex 2 0.043 0.021 2.851 0.094 T Residual 13 0.172 0.013 Veer-into Fish Rank 2 0.011 0.005 1.469 0.266 Fish Sex 1 0.001 0.001 0.1161 0.694 Fish Rank × Fish Sex 2 0.014 0.007 1.914 0.187 Residual 13 0.048 0.004 Pass Fish Rank 2 0.244 0.122 7.645 0.006 * Fish Sex 1 0.001 0.001 0.078 0.785 Fish Rank × Fish Sex 2 0.047 0.023 1.458 0.268 Residual 13 0.208 0.016 Trigger-up Fish Rank 2 0.311 0.156 13.110 0.001 * Fish Sex 1 0.006 0.006 0.547 0.473 Fish Rank × Fish Sex 2 0.076 0.038 3.190 0.075 T Residual 13 0.154 0.012 Chase Fish Rank 2 0.007 0.004 0.932 0.419 Fish Sex 1 0.0001 0.0001 0.023 0.882 Fish Rank × Fish Sex 2 0.012 0.006 1.464 0.267 Residual 13 0.052 0.004 Flee Fish Rank 2 0.009 0.005 0.432 0.658 Fish Sex 1 0.01 0.01 0.908 0.358 Fish Rank × Fish Sex 2 0.027 0.014 1.288 0.309 Residual 13 0.139 0.011 Bite Fish Rank 2 0.0001 0.0004 0.672 0.527 Fish Sex 1 0.00002 0.00002 0.364 0.557 Fish Rank × Fish Sex 2 0.0001 0.00005 0.984 0.399 Residual 13 0.001 0.00005 Head-down Fish Rank 2 0.189 0.095 11.857 0.001 * Fish Sex 1 0.01 0.01 1.296 0.276 Fish Rank × Fish Sex 2 0.001 0.0005 0.063 0.94 Residual 13 0.104 0.008 Head-up Fish Rank 2 0.003 0.001 0.574 0.577 Fish Sex 1 0.00004 0.00004 0.018 0.895 Fish Rank × Fish Sex 2 0.008 0.004 1.655 0.229 Residual 13 0.032 0.002 * Significance. T: trend for factors to exert an effect. 30 Current Zoology Vol. 56 No. 1

Table 5 ANOVA tables for effects of rank and sex on coloration patterns

df Sum of squares Mean square F-value P-value

Light banding Fish Rank 2 1.87 0.935 7.224 0.008 * Fish Sex 1 0.148 0.148 1.143 0.305 Fish Rank × Fish Sex 2 0.764 0.382 2.952 0.088 T Residual 13 1.683 0.129 Dark banding Fish Rank 2 0.699 0.349 2.992 0.085 T Fish Sex 1 0.092 0.092 0.79 0.39 Fish Rank × Fish Sex 2 0.348 0.174 1.491 0.261 Residual 13 1.518 0.117 Grey Fish Rank 2 0.247 0.123 3.881 0.047 * Fish Sex 1 0.001 0.001 0.017 0.899 Fish Rank × Fish Sex 2 0.008 0.004 0.132 0.878 Residual 13 0.414 0.032 White Fish Rank 2 0.486 0.243 3.515 0.06 T Fish Sex 1 0.005 0.005 0.079 0.783 Fish Rank × Fish Sex 2 0.028 0.014 0.203 0.819 Residual 13 0.899 0.069 Residual 13 0.032 0.002 * Significance. T: trend for factors to exert an effect.

Fig. 7 Frequency of observed behaviors for dominant, middle-ranking, and subordinate fish in a group (4 - 5 fish together) interaction setting Bars indicate significant differences (two-factor ANOVA; see Table 4 for specific P values) between ranks of fish. n = 6 for dominant and mid- dle-ranking fish; n = 7 for subordinate fish; n = 9 for females; n = 10 for males. D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 31 fish passing more frequently than did subordinate fish behaviors (Fig. 7). (Bonferroni-Dunn, P < 0.003). As with some behaviors, there was a significant ef- Fish rank also influenced the expression of trigger-up fect of rank, but not sex, on the expression of various and head-down. Trigger-up was displayed more fre- coloration patterns (Fig. 8). The light-banded coloration quently by subordinate fish than by middle-ranking fish was expressed significantly more often by dominant fish

(two-factor ANOVA, F2,13= 13.11, P < 0.001, Bon- than either by middle-ranking or subordinate fish ferroni-Dunn, P < 0.014, Fig. 7) and more frequently by (two-factor ANOVA, F2,13= 7.22, P < 0.01, Bon- middle-ranking and subordinate fish than by dominant ferroni-Dunn, P = 0.045 and P < 0.01, respectively). fish (Bonferroni-Dunn, P = 0.038 and P = 0.0001, re- Middle-ranking and subordinate fish displayed the same spectively). There was a tendency for there to be an in- frequency of light-banded coloration. While there was teractive effect between rank and sex for the expression no significant effect of rank or sex on the expression of of trigger-up (two-factor ANOVA, F2,13= 3.19, P = dark banding, there was a trend for rank to have an ef-

0.075), but this trend was not significant and is likely fect (two-factor ANOVA, F2,13= 2.99, P = 0.085). Grey due to one sex predominating in each of the groups. coloration was expressed more frequently by subordi- Head-down, often used in conjunction with trigger-up, nate fish than by dominant fish (two-factor ANOVA, was displayed more frequently by subordinate fish than F2,13= 3.88, P = 0.048, Bonferroni-Dunn, P = 0.022), by middle-ranking fish (two-factor ANOVA, F2,13= but middle-ranking and subordinate fish expressed this 11.86, P = 0.001, Bonferroni-Dunn, P = 0.018, Fig. 7) coloration with the same frequency. Despite the frequent and by subordinate more than by dominant fish (Bon- pairing of white coloration with submissive behaviors ferroni-Dunn, P < 0.001). There was no difference in (head-down, trigger-up), there was no significant effect the frequency of expression of head-down by mid- of rank or sex on its expression, although there was a dle-ranking and dominant fish. trend for rank to exert some effect (two-factor ANOVA,

There was no effect of rank or sex on the expression F2,13= 3.52, P = 0.06). Speckled and black colorations of veer-into, chase, flee, bite, or head-up behavior. were so infrequently displayed by any rank of fish that These behaviors were infrequent compared to the other they could not be examined for rank and sex effects.

Fig. 8 Frequency of observed coloration patterns for dominant, middle-ranking, and subordinate fish in a group (4 – 5 fish to- gether) interaction setting Bars indicate significant differences (two-factor ANOVA; see Table 5 for specific P values) between ranks of fish. n = 6 for dominant and mid- dle-ranking fish; n = 7 for subordinate fish; n = 9 for females; n = 10 for males. 32 Current Zoology Vol. 56 No. 1

3 Discussion ners were very likely to win subsequent contests only if they fought opponents within a short period of time. If, The grey triggerfish B. capriscus uses a series of be- however, they fought opponents after a waiting period haviors, body postures, and coloration patterns to com- exceeding 1 h, they tended to have a lower-than-expected municate status to conspecifics within a social group. chance of dominating their second opponent (Chase et Dominant-ranking fish predominantly use a light- or al., 1994). These results suggest that winner effects may dark-banded coloration and tend to frequently approach be of only short-term importance. Such winner and loser and pass lower-ranking individuals as part of an overall effects have been found in blue gourami Macropodus behavioral sequence (Lavalli and Spanier, 2001). Both opercularis, three-spined sticklebacks Gasterosteus middle-ranking and subordinate fish use a wider variety aculeatus, swordtails Xiphophorus helleri, and pump- of coloration patterns, depending on the rank of the in- kinseed sunfish, with the effects lasting a variable pe- dividual with which they interact. These patterns in- riod of time depending on species (Francis, 1983; 1987; clude light-banded, dark-banded, grey, white, black, and Bakker and Sevenster, 1983; Beacham and Newman, speckled and are often combined with head-down and 1987; Beaugrand et al., 1996). In most studies, fish trigger-up behaviors that are not often seen in domi- memories of other individuals tend to be short-term if nant-ranking fish. Such signals may be very important separated (Johnsson, 1997; Miklosi et al., 1997), but to individuals within a social group, especially if an they may be constantly reinforced when fish are in individual can use these cues to deduce rank without long-term associations with each other. having to engage in agonistic encounters. Some fish Other studies focusing on visual communication be- species are capable of inferring hierarchical arrange- tween fish opponents have described similar color ments when acting solely as bystanders watching changes for winners and losers in combat (Beeching, fights between pairs of individuals comprising a larger 1995, Dawkins and Guilford, 1993). Beeching (1995) group (Grosenick et al., 2007). This ability would be described how defeated oscars Astronotus ocellatus un- most useful to new individuals joining a social group dergo a change in color pattern in which the normal in which the hierarchy is already established, and we olive green to brown body coloration darkens to have noted that triggerfish groups on the same reefs near-black with irregular white barring. This banded vary in number from as few as two individuals to as pattern inhibited further aggression by dominant tank many as 23 (Lavalli and Herrnkind, 2009), suggesting mates. Dawkins and Guilford (1993) described the rapid that there may be a high level of fusion and fission in and frequent changes of body coloration of the terminal such groups. phase of the adult male bluehead Thalassoma While dominant-ranking fish used only a banding bifasciatum. Body coloration changed from bright green coloration pattern that switched between light, almost when an individual was aggressive towards other fish to negligible bands, to an intensified banding (darker), opalescent when an individual was courting females and lower-ranking fish switched rapidly among a variety of spawning. Likewise, juvenile Atlantic Salmo colorations, presumably using a variety of chromato- salar typically adopted a uniformly pale coloration phores (melanophores, leucophores, xanthophores, and when on light-colored substrate but darkened both the iridophores) to do so. Fish chromatophores can be under body and sclera in response to losing an aggressive en- nervous and/or endocrine control (Fujii, 2000), and it is counter with a dominant fish (O’Connor et al., 1999). likely that the rapid responses seen in this study were Abbott et al. (1985) showed that subordinate rainbow the result of nervous activity stimulated by retinal input trout Oncorhynchus mykiss both darkened their body of information presented by other fish within the pair or coloration and altered their body postures when domi- group settings. Dominant fish may, however, have much nant fish were present. Subordinates generally retreated of their chromatophore (melanophore) activity under from the dominants and increased the curvature of their endocrine control, particularly after they have won en- dorsal outline in what was termed a “hunched posture.” counters with other fish. Such endocrine changes are In other fish species, darkened bars signal aggression known for “winners” of other taxa (c.f. crayfish, Huber (Neil, 1983, 1984; Zimmerer and Kallman, 1988; Hurd, and Delago, 1998; Huber, 2005; Oyegbile and Marler, 1997) and may also provide information about an indi- 2005) and result in behavioral differences among domi- vidual's size, since the number of bars and the extent of nant and subordinate individuals, at least in the short their area will be greater in larger specimens (Zimmerer term. In pumpkinseed sunfish Lepomis gibbosus, win- and Kallman, 1988). Our results demonstrate that D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 33 triggerfish are but another fish species that uses both an effective predictive tool for determining a dominance body coloration and posture to indicate current intention hierarchy in a group of four to five fish. Like Oliveira et as well as immediate future behavior, and that the al. (1998) and Chase et al. (2003), we conclude that for banding patterns displayed most frequently by the fish species found in groups, dominance hierarchies dominant and middle-ranking fish may provide a rapid need to be studied at the group rather than the paired assessment of size in this species. In addition, since level so that the effects of observation, eavesdropping, barred patterns are more conspicuous in water where and social context can be allowed to play their appro- light is scattered (Lythgoe, 1979), they may provide priate roles in the determination of relationships among longer-distance signaling among individuals on a reef. individuals. Triggerfish seem to use a combination of light colors What intrinsic factors seem to influence rank in with conspicuous bands to indicate the highest ranks triggerfish? In salmonids, body size is generally a good and a darker color with more intensified bands to indi- indicator of social status. Large individuals of Salmo cate middle ranks (Fig. 8). Lack of bands and a decrease salar often become dominant as they become larger and in or absence of pigmentation appear to indicate the win more and more contests; however, a smaller fish lowest ranks. We have noted in previous field experi- with a greater number of prior wins may emerge domi- ments that one or two triggerfish arriving at a reef will nant over a larger fish with fewer prior wins. Results often disappear and recruit other conspecifics to that suggest that prior wins and losses influence an individ- reef. Once a group of triggerfish is present, individuals ual’s fighting ability to a greater extent than size, will display different coloration patterns to each other thereby affecting its position in the hierarchy (Metcalfe while only one or two fish attempt to subdue tethered et al., 1992; Beaugrand and Cotnoir, 1996). In a study to decapods (Barshaw et al., 2003; Lavalli and Herrnkind, determine who picks on whom, Castro and Caballero 2009). We have also noted that banding and coloration (1998) found that dominant juvenile white seabream patterns differ among individuals engaged in subduing Diplodus sargus cadenati carry out aggressive attacks free-ranging spiny lobsters in large, naturalistic meso- selectively on fish whose subordination level is imme- cosms (Lavalli, unpublished data). Hence, it is likely diately inferior to their own to establish a that these coloration patterns, as well as body postures, peck-dominance hierarchy. The level of subordination is are important signals in this species. directly related to an individual’s body size. In trigger- A number of studies using vertebrates show that the fish, large size seems to be the major determinant of relationship between paired individuals is not necessar- which fish will occupy an alpha position, but size is not ily the same when those individuals are found in a larger coupled to the determination of the precise level of sub- group setting (Neilssen, 1985; Holekamp and Smale, ordination. Therefore, it seems likely that triggerfish 1991; Chapais, 1995; Oliveira et al., 1998; Chase et al., may use mainly extrinsic encounter effects (wins and 2003). This predictive failure of the dyad approach to losses) to determine the middle and lower ranks within a assessing dominance relationships may result because group. Sex does not seem to play a role in rank deter- interactions amongst many individuals may be influ- mination in this species. enced not only by intrinsic attributes of the individuals (size, weight, age, etc.) but also by extrinsic attributes Acknowledgements This study would not have been possi- relating to the social network in which the individuals ble without the help of many people. We thank the staff at the exist. With regard to social fish, individuals are capable Keys Marine Laboratory, Layton, FL, for providing services, of ascertaining the fighting abilities of other individuals space, and supplies for setting up experiments in their runways. that they observe in contests (McGregor, 1993; Johns- We also thank Dr. William Herrnkind from Florida State Uni- son and Akerman, 1998; Oliveria et al., 1998; Herb et versity for catching the fish and providing useful insights on al., 2003) and some of these bystanders even undergo triggerfish behavior, as well as his former student, Dr. Peter Bouwma, who helped acquire and handle the fish and video hormonal changes when observing such contests equipment. Dr. Bill Fable and the staff of the National Marine (Oliveira et al., 2001) or develop preferences for inter- Fisheries Service facility in Panama City, FL, USA provided acting with those who have not seen them lose an en- services and space for data collection. Thanks also to two counter (Herb et al., 2003). As with these studies, our anonymous reviewers who provided helpful comments on the results – wherein hierarchies determined by dyads were manuscript. This study was partially funded by a Research the same as hierarchies seen in group settings only 50% Enhancement Grant to KLL from Southwest Texas State Uni- of the time – indicate that dyadic relationships are not versity, San Marcos, TX, USA. 34 Current Zoology Vol. 56 No. 1

References Drews C, 1993. The concept and definition of dominance in animal behaviour. Behaviour 125: 283–313. Abbott JC, Dunbrackand RL, Orr CD, 1985. The interaction of size Dugatkin LA, 1997. Winner and loser effects and the structure of and experience in dominance relationships of juvenile steelhead dominance hierarchies. Behav. Ecol. 8: 583–587. trout Salmo gairdneri. Behaviour 92: 241–253. Dugatkin LA, 2001. Bystander effects and the structure of dominance Archer J, 1988. The Behavioural Biology of Aggression. Cambridge: hierarchies. Behav. Ecol. 12: 348–352. Cambridge University Press. Evans DL, Shehadi-Moacdieh M, 1988. Body size and prior residency Bakker T, Sevenater P, 1983. Determinants of dominance in male in staged encounters between female prawns Palaemon elegans sticklebacks Gasterosteus aculeatus. Behaviour 86: 55–71. Rathke (Decapoda: Palaemonidae). Anim. Behav. 36: 452–455. Barki A, Karplus I, Goren M, 1991. Morphotype-related dominance Figler MH, Einhorn DM, 1983. The territorial prior residence effect in hierarchies in males of Marcrobrachium rosenbergii (Crustacea, convict cichlids Cichlasoma nigrofasciatum gunther: Temporal Palaemonidae). Behaviour 117: 145–160. aspects of establishment and retention, and proximate mechanisms. Barshaw DE, Lavalli KL, Spanier E, 2003. Is offence the best defence: Behaviour 85: 157–181. The response of three morphological types of lobsters to predation. Forrester GE, 1990. Factors influencing the juvenile demography of a Mar. Ecol. Prog. Ser. 256: 171–182. . Ecology 71: 1666–1681. Basquil SP, Grant JWA, 1998. An increase in habitat complexity re- Francis RC, 1983. Experiential effects on agonistic behavior in the duces aggression and monopolization of food by zebra fish Danio paradise fish Macropodus opercularis. Behaviour 85: 292–313. rerio. Can. J. Zool. 76: 770–772. Francis RC, 1987. The interaction genotype and experience in the Beacham JL, Newman JA, 1987. Social experience and the formation dominance success of paradise fish Macropodus opercularis. Biol. of dominance relationships in the pumpkinseed sunfish Lepomis Behav. 12: 1–11. gibbosus. Anim. Behav. 35: 1560–1563. Franck D, Ribowski A, 1987. Influences of prior agonistic experiences Beaugrand JP, Cotnoir P, 1996. The role of individual differences in on aggression measures in the male swordtail Xiphophorus helleri. the formation of triadic dominance orders of male green swordtail Behaviour 103: 217–240. fish Xiphophorus helleri. Behav. Proc. 38: 287–296. Fujii R, 2000. The regulation of motile activity in fish chromatophores. Beaugrand JP, Goulet C, 2000. Distinguishing kinds of prior domi- Pigment Cell Res. 13: 300–319. nance and subordination experiences in males of green swordtail Grant JWA, Noakes, DLG, Jonas KM, 1989. Spatial distribution of fish Xiphophorus helleri. Behav. Proc. 50: 131–142. defence and foraging in young-of-the-year brook charr Salvetinus Beaugrand JP, Goulet C, Payette D, 1991. Outcome of dyadic conflict fontinalis. J. Anim. Ecol. 58: 773–784. in male green swordtail fish Xiphophorus helleri: Effects of body Grosenick L, Clement TS, Fernald RD, 2007. Fish can infer social size and prior dominance. Anim. Behav. 41: 417–424. rank by observation alone. Nature 445: 429–432. Beaugrand JP, Payette D, Goulet C, 1996. Conflict outcome in male Hemelrijk CK, 1990. A matrix partial correlation test used in investi- green swordtail dyads Xiphorsus helleri: Interactions of body size, gations of reciprocity and other social interaction patterns at a prior dominance/subordination status experience and prior resi- group level. J. Theor. Biol. 143: 405–420. dency. Behaviour 133: 303–319. Herb BM, Biron SA, Kidd MR, 2003. Courtship by subordinate male Beeching SC, 1995. Colour pattern and inhibition of aggression in the Siamese fighting fish Betta splendens: Their response to eaves- cichlid fish Astronotus ocellatus. J. Fish Biol. 47: 50–58. dropping and naïve females. Behaviour 140: 71–78. Burk TE, 1979. An analysis of social behaviour in crickets. Ph.D. Herrnkind WF, Childress MJ, Lavalli KL, 2001. Cooperative defense dissertation, University of Oxford. and other benefits among exposed spiny lobsters: inferences from Castro JJ, Caballero C, 1998. Dominance structure in small groups of group size and behavior. Mar. Freshwater Res. 52: 1113–1124. juvenile white seabream (Diplodus sargus adenati de la Paz, Bau- Holbrook SJ, Schmitt, RJ, 1992. Causes and consequences of dietary chot and Daget 1974). Aggr. Behav. 24: 197–204. specialization in surfperches: Patch choice and intraspecific com- Chapais B, 1995. Alliances as a means of competition in primates: petition. Ecology 73: 402– 412. evolutionary, developmental, and cognitive aspects. Yrbk. Phys. Holekamp KE, Smale L, 1991. Dominance acquisition during mam- Anthropol. 38: 115–136. malian social development: The “inheritance” of maternal rank. Chase ID, 1982. Dynamics of hierarchy formation: the sequential Amer. Zool. 31: 306– 317. development of dominance relationships. Behaviour 80: 218–240. Hsu YY, Wolf LL, 1999. The winner and loser effect: Integrating Chase ID, Bartolomero C, Dugatkin LA, 1994. Aggressive interac- multiple experiences. Anim. Behav. 57: 903–910. tions and inter-contest interval: How long do winners keep win- Huber R, 2005. Amines and motivated behaviours: A simpler systems ning? Anim. Behav. 48: 393–400. approach to complex behavioural phenomena. J. Comp. Physiol. A Chase ID, Tovey C, Spangler-Martin D, Manfredonia M, 2002. Indi- 191: 231– 239. vidual differences versus social dynamics in the formation of ani- Huber R, Delago A, 1998. Serotonin alters decisions to withdraw in mal dominance hierarchies. Proc. Nat. Acad. Sci. 99: 5744–5749. fighting crayfish Astacus astacus: The motivational concept revis- Chase IA, Tovey C, Murch P, 2003. Two's company, three's a crowd: ited. J. Comp. Physiol. A 182: 573–583. differences in dominance relationships in isolated versus socially Huntingford F, Turner A, 1987. Animal Conflict. London: Chapman embedded pairs of fish. Behaviour 140: 1193–1217. and Hall. Dawkins MS, Guilfod T, 1993. Colour and pattern in relation to sexual Huntingford FA, Taylor AC, Smith IP, Thorpe KE, 1995. Behavioural and aggressive behaviour in the bluehead wrasse Thalassoma bi- and physiological studies of aggression in swimming . J. Exp. fasciatum. Behav. Proc. 30: 245–252. Mar. Biol. Ecol. 193: 21–39. de Vries H, 1998. Finding a dominance order most consistent with a Hurd PL, 1997. Cooperative signaling between opponents in fish linear hierarchy: A new procedure and review. Anim. Behav. 55: fights. Anim. Behav. 54: 1309–1315. 827–843. Jameson KA, Appleby MC, Freeman LC, 1999. Finding an appropri- D. W. CLEVELAND, K. L. LAVALLI: Grey triggerfish hierarchies 35

ate order for a hierarchy based on probabilistic dominance. Anim. O’Connor KI, Metcalfe NB, Taylor AC, 1999. Does darkening signal Behav. 57: 991–998. submission in territorial contests between juvenile Atlantic salmon Johnsson JL, 1997. Individual recognition affects aggression and Salmo salar. Anim. Behav. 58: 1269–1276. dominance relations in rainbow trout Oncorhynchus mykiss. Oliveira RF, McGregor PK, Latruffe C, 1998. Know thine enemy: Ethology 103: 267–282. Fighting fish gather information from observing conspecific inter- Johnsson JL, Akerman A, 1998. Watch and learn: Preview of the actions. Proc. R. Soc. Lond. B 265: 1045–1049. fighting ability of opponents alters contest behaviour in rainbow Oliveira RF, Lopes M, Carneiro LA, Canário AVM, 2001. Watching trout. Anim. Behav. 56: 771–776. fights raises fish hormone levels. Nature 409: 475. Kroon FJ, de Graaf M, Liley NR, 2000. Social organisation and com- Oyegbile TO, Marler CA, 2005. Winning fights elevates testosterone petition for refuges and nest sites in Coryphopterus nichosii levels in California mice and enhances future ability to win fights. (Gobiidae), a temperate protogynous reef fish. Env. Biol. Fish 57: Horm. Behav. 48: 259–267. 401–411. Parker GA, 1974. Assessment strategy and the evolution of fighting Kurz RC, 1995. Predator-prey interactions between grey triggerfish behaviour. J. Theor. Biol. 47: 223–243. (Balistes capriscus Gmelin) and a guild of sand dollars around ar- Pavey CR, Fielder DR, 1996. The influence of size differential on tificial reefs in the northeastern Gulf of Mexico. Bull. Mar. Sci. 56: agonistic behaviours in the freshwater crayfish Cherax cuspidatus 150–160. (Decapoda: Parastacidae). J. Zool. 238: 445–457. Landau HG, 1951a. On the dominance relations and the structure of Peeke HVS, Sippel J, Figler MH, 1995. Prior residence effects in animal societies: I. Effects of inherent characteristics. Bull. Math. shelter defense in adult signal crayfish Pacifastacus leniusculus Biol. 13: 1–19. (Dana) – results in same-sex and mixed-sex dyads. Crustaceana 68: Landau HG, 1951b. On the dominance relations and the structure of 873–881. animal societies: II. Some effects of possible social factors. Bull. Ranta E, Lindstrom K, 1992. Power to hold sheltering burrows by Math. Biol. 13: 245–262. juveniles of the signal crayfish Pacifastacus leniusculus. Ethology Lavalli KL, Herrnkind WF, 2009. Defensive strategies of Caribbean 92: 217–226. spiny lobsters: Effects of group size and predator group size. N.Z. Ranta E, Lindstrom K, 1993. Body size and shelter possession in J. Mar. Freshwater Res. 43: 15–28. mature signal crayfish Pacifastacus leniusculus. Ann. Zool. Fenn. Lavalli KL, Spanier E, 2001. Does gregariousness function as an 30: 125–132. antipredator mechanism in the Mediterranean slipper lobster Scyl- Rubenstein DL, Hazlett BA, 1974. Examination of the agonistic be- larides latus? Mar. Freshwater Res. 52: 1133–1143. haviour of the crayfish Orconectes virilis by character analysis. Lehner, PN, 1996. Handbook of Ethological Methods. 2nd edn. Cam- Behaviour 50: 193–216. bridge: Cambridge University Press. Rutherford PL, Dunham DW, Allison V, 1995. Winning agonistic Lythgoe JN, 1979. The Ecology of Vision. Oxford: Clarendon Press. encounters by male crayfish Orconectes rusticus (Girard) (Deca- Martin P, Bateson P, 1993. Measuring Behaviour: An Introductory poda, Cambaridae) – chela size matters but chela symmetry does Guide. 2nd edn. Cambridge: Cambridge University Press. not. Crustaceana 68: 526–529. McGregor PK, 1993. Signalling in territorial systems: A context for Rutherford PL, Dunham DW, Allison V, 1996. Antennule use and individual identification, ranging and eavesdropping. Proc. R. Soc. agonistic success in the crayfish Orconectes rusticus (Girard, 1852) Lond. B 340: 237–244. (Decapoda, Cambaridae). Crustaceana 69: 117–122. Metcalfe NB, 1986. Intraspecific variation in competitive ability and Samuelson TJ, Einarsen JS, 1996. The grey triggerfish (Balistes caro- food intake in salmonids: Consequences for energy budgets and linensis Gmelin), a balistid fish new to the Norwegian fauna. growth rates. J. Fish. Biol. 28: 525–531. Fauna Norv. Ser. A. 17. Metcalfe NB, Huntingford FA, Graham WD, Thorpe JE, 1989. Early Sloman KA, Armstrong JD, 2002. Physiological effects of dominance social status and the development of life-history strategies in At- hierarchies: Laboratory artifacts or natural phenomenon? J. Fish. lantic salmon. Proc. R. Soc. Lond. B 236: 7–19. Biol. 61: 1–23. Metcalfe NB, Wright PJ, Thorpe JE, 1992. Relationships between Vose FE, Nelson WG, 1994. Gray triggerfish (Balistes capriscus social status, size at first feeding and subsequent growth in Gmelin) feeding from artificial and natural substrate in shallow Atlantic salmon Salmo salar. J. Anim. Ecol. 61: 585–589. Atlantic waters of Florida. Bull. Mar. Sci. 55: 1316–1323. Miklosi A, Haller J, Csanyi V, 1997. Learning about the opponent Vye C, Cobb JS, Bradley T, Gabbay J, Genizi A, Karplus I, 1997. during aggressive encounters in paradise fish (Macropodus oper- Predicting the winning or losing of symmetrical contests in the cularis L.): When it takes place? Behav. Proc. 40: 97–105. American lobster Homarus americanus (Milne-Edwards). J. Exp. Nakano S, 1995. Individual differences in resource use, growth and Mar. Biol. Ecol. 217: 19–29. emigration under the influence of a dominance hierarchy in fluvial Webster MS, Hixon MA, 2000. Mechanisms and individual conse- red-spotted masu salmon in a natural habitat. J. Anim. Ecol. 64: quences of intraspecific competition in a coral reef fish. Mar. Ecol. 75–84. Prog. Ser. 196: 187–194. Neil SJ, 1983. Field studies of the behavioral ecology and agonistic Wilson, EO, 1975. Sociobiology: The New Synthesis. Cambridge, behavior of Cichlasoma meeki (Pisces: Cichlidae). Env. Biol. Fish MA: Harvard University Press. 10: 59–68. Zar JH, 1999. Biostatistical analysis. 4th edn. New Jersey: Prentice Neil SJ, 1984. Color pattern variability and behavioral correlates in the Hall. firemouth cichlid Cichlasoma meeki. Copeia 1984 (2): 534–538. Zimmerer EJ, Kallman KD, 1988. The inheritance of vertical bar- Nelissen M, 1985. Structure of the dominance hierarchy and domi- ring (aggression and appeasement signals) in pygmy swordtail nance determining “group factors” in Melanochromis auratus (Pi- Xiphophorus nigrensis (Poeciliidae, Teleostei). Copeia 1988: sces, Cichlidae). Behaviour 94: 85–107. 299–307.