Individual consistency in the behaviors of newly-settled reef fish

Introduction

There has been considerable interest in, and evidence for, consistent patterns in the behaviors of individual animals within a species over the last decade (Dall, Houston & McNamara, 2004; Sih, Bell & Johnson, 2004; Sih et al., 2004; Dingemanse & Réale, 2005; Bell, 2007; Réale et al., 2007; Smith & Blumstein, 2008; Bell, Hankison & Laskowski, 2009). Differences in the amount of aggressive, exploratory and bold behaviors among individuals have been shown to be widespread and heritable (Boake, 1994; Stirling, Reale & Roff, 2002; Kolliker, 2005; Van Oers et al., 2005; Réale et al., 2007) across a diverse array of taxa (Dingemanse & Réale, 2005; Smith & Blumstein, 2008) and to influence survival (Downes, 2002; Dingemanse et al., 2004), reproductive success (Both et al., 2005; Sih & Watters, 2005; Pruitt & Ferrari, 2011), resource acquisition (Webster, Ward & Hart, 2009) and growth. Adopting a certain behavioral phenotype can have both costs and benefits, for example, highly aggressive female fishing spiders (Dolomedes triton) are more successful at acquiring food in a competitive environment, but this aggression can be detrimental in another context such as when it leads to precopulatory sexual cannibalism (Johnson & Sih, 2005). Thus, consistent patterns in behavior among individuals can lead to trade-offs in aspects of fitness, which can ultimately influence population dynamics, community structure, and species diversity (Pruitt, Grinsted & Settepani, 2013; Mittelbach et al., 2014).

Variation in consistent behavioral patterns among individuals have been variously (and interchangeably) termed ‘behavioral syndromes,’ ‘temperament,’ ‘personality,’ and ‘coping styles’ (Dall, Houston & McNamara, 2004; Réale et al., 2007; Dingemanse et al., 2010; Sih et al., 2012), although some authors have argued for a more restrictive use of terminology (Bell, Hankison & Laskowski, 2009; Garamszegi & Herczeg, 2012). Here, we adopt the definitions of Garamszegi & Herczeg (2012), where consistency in single behaviors (e.g., individuals that display repeatedly higher or lower levels of boldness, exploration, or aggression than others in the population) are described as displaying ‘personality,’ and consistency in the relationship between two or more functionally different behaviors within the same individual is defined as a ‘behavioral syndrome.’ For example, a behavioral syndrome is evident in the correlation between boldness and aggression documented within individual sticklebacks (Gasterosteus aculeatus) (Bell, 2005) and funnel-web spiders (Agelenopsis aperta) (Riechert & Hedrick, 1993).

Although the ability to alter behavior to suit changing environmental conditions is likely to be advantageous (Kelley, Phillips & Evans, 2013), behavior is not infinitely plastic (DeWitt, Sih & Wilson, 1998). If a single optimal behavioral phenotype existed, natural selection should reduce genotypic variation over generations (Réale et al., 2007). Because behavioral phenotypes show heritable variation not eroded by selection (Penke, Denissen & Miller, 2007; Réale et al., 2007), different behavioral strategies are likely to have different associated costs and benefits (Kelley, Phillips & Evans, 2013). For example, larger, bolder and faster-growing phenotypes of rainbow trout (Oncorhynchus mykiss) are more likely to be captured by fishing gears (Biro & Post, 2008).

Estimating the consistency of a behavioral trait is necessary for measuring the repeatable characteristics of a focal organism, quantifying trait plasticity and determining trait heritability (Nakagawa et al., 2007). Historically, personality studies using a single assay were common, but it has been recently suggested that repeated tests are essential for any personality study (Réale et al., 2007) and the strength of behavioral syndromes are likely underestimated when based upon single assays of varying traits (Adolph & Hardin, 2007; Beckmann & Biro, 2013; White et al., 2013; White, McCormick & Meekan, 2013).

Clearly, there is a need to determine the consistency of behaviors before examinations of personality, behavioral syndromes and associated trade-offs of alternative behavioral strategies can be attempted. Here, we examine evidence for personalities in a juvenile tropical reef fish, the Ambon damselfish (Pomacentrus amboinensis), by establishing the consistency of commonly-used field and laboratory assays of activity, aggression and boldness over time scales ranging from minutes to days following settlement. Similar to many reef fishes, young of this species can be collected at the end of their larval phase immediately prior to settlement on the reef, when they are naïve to reef-based predators and behaviors learned after settlement (Meekan et al., 2010). In this immediate post-settlement phase of their life cycle, reef fishes typically experience very high mortality (Almany & Webster, 2006), with rates within the first 48 h of benthic life averaging 57% (Doherty et al., 2004; Almany & Webster, 2006). Because experience can influence behavioral phenotypes (Budaev, 1997; Bell & Sih, 2007; Dingemanse et al., 2009), the use of naïve study organisms allows us to control for variation and consistency in behavior associated with experience and to examine ecologically important behavioral traits at a critical ontogenetic boundary (McCormick & Meekan, 2010; Poulos & McCormick, 2014). Because field measurements are made directly by an observer on SCUBA (where visual and auditory presence is not easily concealed), we tested for an effect of observer presence by comparing observed behaviors to those recorded by video-camera. Specifically, we aimed to determine if juvenile damselfish behaviors were: (1) significantly altered by observer presence; (2) consistent over various time scales (minutes, hours, days) relevant to their major mortality bottleneck (first 48 h following settlement); (3) consistent in an aquarium setting; and (4) correlated between field and lab-based measurements. Based on our anecdotal previous experience with this system and study species, we predicted all behaviors to be moderately consistent in the field and laboratory.

Methods

Results

Discussion

Our study is one of the most detailed assessments of behavioral consistency of a marine organism to date. It shows that shortly after entering a new habitat at the end of their larval phase fish approximately three weeks old already have a complex repertoire of behaviors that are displayed in a consistent way through time, indicative of the existence of individual personalities. Moreover, this personality appears to be established prior to or immediately upon metamorphosis and settlement. Factors that are likely to favor consistent over conditional behavior, and thus give rise to individual personalities are diverse and include: genetic, physiological or developmental limits, costs of flexibility, costs and availability of information acquisition, metabolism, body size, or constraints on behavioral plasticity (Sih, Bell & Johnson, 2004; Bergmuller, Schurch & Hamilton, 2010; Briffa & Greenaway, 2011). Stable behavioral states are hypothesized to be created when positive feedback loops form between state variables such as size, competitive ability, or condition and state-dependent behavioral decisions (Dall, Houston & McNamara, 2004; Sih & Bell, 2008). For example, individuals with higher body condition may be more cooperative compared to those in poorer condition because they can afford the energy expenditure. If cooperative behavior then led to increased energy gains, this feedback loop would maintain higher body condition (Bergmuller, Schurch & Hamilton, 2010). Naïve juvenile reef fish exhibiting personalities at settlement suggests a genetic component and strong trade-offs related to adopting alternative personalities. High mortality rates at this phase of their life cycle could provide very strong selective force and are most likely to be involved (McCormick & Meekan, 2010).

Generally, our study found moderate to highly repeatable behavioral scores for almost all behavioral measures. These ranged from 0.33 (height on the habitat patch across multiple days) to 0.89 (height across camera observations), values well within the range recorded by earlier studies. A recent meta-analysis by Bell, Hankison & Laskowski (2009) reported an average repeatability value of 0.37 in various behavioral traits across 114 studies and 98 species. They found mating, habitat selection and aggression to be the most repeatable traits; while activity, mate preference, and migration were the least repeatable. Consistency was generally higher for behaviors measured at closer time intervals, juveniles compared to adults and field studies versus laboratory settings (Bell, Hankison & Laskowski, 2009). Approximately 70% of this distribution was between 0.1 and 0.6 (see Fig. 1, Bell, Hankison & Laskowski, 2009). An additional 11 studies published more recently (Réale et al., 2000; Smith & Blumstein, 2008; Briffa & Greenaway, 2011; Marras et al., 2011; Couchoux & Cresswell, 2012; Carter et al., 2012; Beckmann & Biro, 2013; Neumann et al., 2013; Pruitt, Grinsted & Settepani, 2013; Kelley, Phillips & Evans, 2013; Burtka & Grindstaff, 2013) reported repeatability scores ranging from as low as 0.14 for a measure of aggression in male crested macaques (Macaca nigra) (Neumann et al., 2013) to as high as 0.92 for a measure of escape response in European sea bass (Dicentrarchus labrax) (Marras et al., 2011). Despite the wide range in these scores, they were cited as evidence of the consistency of behaviors and therefore personalities. On this basis, the repeatability scores we obtained suggest evidence for personality in the 3-week old damselfish that were the subjects of our study.

Large confidence intervals around a repeatability estimate suggest significant within-individual variation in behavior (Jones & Godin, 2009). While juvenile damselfish are known to adopt a wide range of behavioral strategies (White et al., 2013; White, McCormick & Meekan, 2013), some of the variation we recorded may be due to plasticity in the amount of habituation to the experimental protocol (Martin & Réale, 2008). Across repeated trials, an environment or test may become less novel and individuals may habituate to novelty in itself (Réale et al., 2007; Edwards et al., 2013), or alternatively become less responsive or sensitized (Budaev, 1997; Martin & Réale, 2008; Kelley, Phillips & Evans, 2013). In our study, the tests that involved an experimental set-up, such as laboratory-based measurements of boldness (e.g., latency to emerge), have some of the largest confident intervals. However, given our significant repeatability estimates, we are confident all the measures reported are reliable measures of an individual’s behavior within these contexts.

Variables that originated from the aggression assay (strike latency and mirror strike rate) were the only measurements found not to be repeatable through time or context. This suggests the moderate negative correlation found between field and laboratory measures of aggression strike latency is likely to be ecologically irrelevant. While a commonly-used test (Gerlai, 2003; Marks et al., 2005), these measures may be susceptible to the habituation effect discussed above. A closely-related species, P. moluccensis, has been shown to recognize threats after a single exposure (Mitchell et al., 2011). Perhaps P. amboinensis similarly learns to ignore the false threat of their reflection after repeated exposures.

Observations repeated over short time scales (4 min apart, simultaneous video vs. video observations) had the highest repeatability scores. Measures conducted over longer (30 min apart and 3 times daily over 3 days) time periods had similar, but lower scores. This agrees with results from a meta-analysis, which showed higher estimates of repeatability for behaviors measured at shorter time intervals (Bell, Hankison & Laskowski, 2009). Our results suggest juvenile damselfish quickly adopt stable behavioral phenotypes regarding foraging and activity rates following settlement and remain consistent throughout the intense predation pressure experienced during the first few days on the reef.

There was a trend for repeatability estimates obtained in the laboratory to be lower compared to field-based measurements. This same pattern was found in Bell, Hankison & Laskowski’s (2009) meta-analysis. If there are advantages to behaving consistently (Dall, Houston & McNamara, 2004; McElreath & Strimling, 2006), then the greater environmental variance in the field might create micro-niches, increasing repeatability by allowing individual expression of behavioral variations (Bell, Hankison & Laskowski, 2009). Also, because juveniles are exposed to innately higher predation pressure in the field, this could act as a directional or stabilizing selection on behavior (Bell, Hankison & Laskowski, 2009). However, in this study fish are initially naïve and neophobic upon introduction to the field (Meekan et al., 2010; Chivers et al., 2014; Ferrari et al., 2015), so perhaps the greater sensory input in the field environment is enough to act as a stabilizing influence. A recent study found three-spined sticklebacks (Gasterosteus aculeatus) adopted stable boldness-aggressiveness correlations once exposed to predators (Bell & Sih, 2007). Juvenile damselfish quickly learn about predators (Mitchell et al., 2011) and are likely to swiftly adopt a consistent behavioral phenotype when faced with the variations and challenges of their natural habitat. Given the few and weak correlations found between field and laboratory measures, and lower consistency for laboratory studies suggests inferences about natural behaviors in the field derived from laboratory studies need to be made cautiously (White, McCormick & Meekan, 2013). The lack of predators and increased novelty of the laboratory environment may enable juvenile damselfish to exhibit a great variability of behaviors or prompt different behavioral responses that have little bearing on likely behavior under natural conditions. This implies behavioral studies have limited predictive ability across situations, in particular using laboratory measures to predict behaviors in the field (White et al., 2013; White, McCormick & Meekan, 2013).

Interestingly, Beckmann & Biro (2013) reported repeatability values almost identical to ours for the same laboratory-based boldness measure. They tested two species of juvenile damselfish (P. wardi and P. amboinensis) and showed repeatability in the emergence latency test in home tanks (R = 0.42 for P. amboinensis on the third observation), but no correlations when compared against the same and different behavioral tests in different contexts. Others have also argued for the use of multiple measures of boldness in order to obtain an ecologically relevant assessment of this behavioral trait (White et al., 2013), and have also found a lack of behavioral consistency across situations (White, McCormick & Meekan, 2013) for juvenile damselfish. While Beckmann & Biro (2013) argue the lack of correlation across contexts means this assay is inadequate to measure boldness, their study likely had issues with habituation (Edwards et al., 2013). In contrast, we found latency to emerge behavior to be significantly repeatable within a single context and moderately positively correlated with an emergence test in the field.

Another important result of our study was that the presence of observers seemed to have no significant impact on fish behavior. While fishes are the focus of much behavioral research, they are rarely observed in their natural environments (Réale et al., 2000; Bell, Hankison & Laskowski, 2009). Typically, observations in a field situation would be conducted from behind a blind (Martin & Bateson, 2007), a luxury not afforded to a noisy bubble-blowing SCUBA diver. While the simultaneous observer and video observations had slightly lower repeatability scores for bite rate and height compared to the comparison of the two video scores (difference of 0.23 and 0.13, respectively), this is most likely an artifact of the difficulties associated with observing detailed behavior via camera. Even with high resolution video, it was difficult to distinguish between feeding strikes and the natural stop–start swimming of these fish. Also, fish leaving the field of view of the camera for a short duration was not an issue for the diver who could maintain visual contact with the target fish at all times. Overall, discrepancies between the methods of observation may have resulted in a slight over-counting of bite rates in the video. This suggests video data is less useful for subjects such as these small damselfish that are quick moving and very mobile. As long as slow, deliberate movements are employed and the observer remains a least a meter away, juvenile damselfish seem indifferent to human presence thus diver observations provide useful records of behavior.

In summary, our results demonstrate that measures of boldness and activity, both in the field and the laboratory, are highly repeatable over time scales relevant to this species during a key period of their life history. These stable behaviors indicate that these 3-week old juvenile fish already have personalities. From a methodological perspective, our results indicate that an initial 3 min assessment of their behavior provides a useful record of an individual’s personality. However, caution is required when comparing field and laboratory based behaviors (White et al., 2013). Future studies with this species can reasonably use a single (i.e., unrepeated) assay to reduce animal stress, which can then be correlated with physical measures of performance and success to determine how individual characteristics combine to affect fitness. Future research will investigate if adult P. amboinensis retain this behavioral consistency through ontogeny.

Supplemental Information