Counting Nemo: anemonefish Amphiprion ocellaris identify species by number of white bars

Many animals, including mammals, birds, reptiles, amphibians, fish and insects, use color patterns to identify individuals, species and mates (e.g. Cuthill et al., 2017; Endler et al., 2005; Kohda et al., 2015; Losos, 1985; Siddiqi et al., 2004; Siebeck, 2004; Siebeck et al., 2010; Wiernasz and Kingsolver, 1992). Most animal colors are produced by pigment cells, and fish, especially coral reef fishes, possess more pigment cell types (e.g. three main types, melanophores, xanthophores and iridophores, as well as other less frequent types, leucophores, erythrophores, cyanophores, etc.) than any other vertebrate (e.g. Fujii, 1993; Kelsh, 2004; Salis et al., 2019). Fishes are known to use this variety of color pattern for interspecific and intraspecific recognition and individual identification (e.g. Kohda et al., 2015; Siebeck, 2004; Siebeck et al., 2010). For example, the Lake Tanganyika cichlid, Neolamprologus pulcher, can accurately distinguish between individuals based on facial color patterns (Kohda et al., 2015). The Ambon damsel, Pomacentrus amboinensis, has intricate patterns, visible only under UV light, which can be used to distinguish the face of a mate or to distinguish it from a congeneric species of similar appearance, P. moluccensis (Siebeck, 2004; Siebeck et al., 2010). As these species are highly social and have a territorial system, the color pattern around the face serves as an important cue to identify individuals and species (Kohda et al., 2015; Siebeck, 2004; Siebeck et al., 2010).

Anemonefishes have a distinctive color pattern among fishes, with conspicuous white bars on a black, red, orange or yellow background (Laudet and Ravasi, 2022). The ecological function of color patterns in anemonefishes is poorly understood, but at least three adaptive hypotheses have been proposed. The first is the species recognition hypothesis, as different species tend to have different numbers of white bars, especially among species living in the same areas (Hayashi et al., 2022b; Salis et al., 2018). In addition, Mitchell et al. (2023) showed experimentally that the orange and white bars of anemonefishes are highly UV reflective, and that the contrast and intensity of this UV reflection affected the determination of the anemonefish's social status. The second hypothesis is the disruptive coloration hypothesis, which suggests that white bars function to hide the fish silhouette (Salis et al., 2018). Lastly, the third hypothesis is the aposematism hypothesis, according to which the conspicuous color patterns serve to advertise the toxicity of the host anemone (Merilaita and Kelley, 2018). Although the first and second or first and third hypotheses can occur simultaneously, the second and third are mutually exclusive and cannot occur simultaneously.

Anemonefishes include 28 species in the genus Amphiprion within the family Pomacentridae (Dunn, 1981; Fautin and Allen, 1997; Laudet and Ravasi, 2022; Ollerton et al., 2007). Aside from a pelagic larval stage, anemonefish inhabit a host anemone throughout their life and fiercely defend their host as their territory (e.g. Fautin and Allen, 1997; Hattori, 2002; Moyer and Sawyers, 1973; Ross, 1978). They usually form a colony on an anemone that includes one female, one male and several immature individuals, and there is a hierarchy within the colony based on size, with the largest (and most dominant) individuals being females, the second largest being males and all other individuals being immature fish (Fautin and Allen, 1992, 1997; reviewed in Laudet and Ravasi, 2022). However, in some cases, colonies can be formed by only one to several immature individuals (e.g. Buston et al., 2022; Dunn, 1981; Fautin and Allen, 1992, 1997; Hattori, 2011). Each anemonefish species lives with only certain species of host anemones, ranging from generalists, such as A. clarkii, who utilize 10 different host species, to specialists, such as A. frenatus, who utilize only one species of host (Allen, 1972; Fautin and Allen, 1992; Hayashi and Reimer, 2022). In areas where many species of anemonefish co-occur, there may be cohabitation and/or competition over common host species (Camp et al., 2016; Elliott and Mariscal, 2001; Hayashi et al., 2018, 2021; Srinivasan and Jones, 2022). Thus, anemonefish compete for or cohabit in hosts not only among conspecifics but also among heterospecifics, and this interference varies depending on the density and species composition in each locality (Camp et al., 2016; Elliott and Mariscal, 2001; Hayashi et al., 2018, 2021; Srinivasan and Jones, 2022). Because the shelter of a host anemone is essential for anemonefish survival, and given these constant interferences between species, it is critically important for anemonefish to accurately identify and repel competitors both before and after they have established themselves in the host anemone (Fautin and Allen, 1992, 1997; reviewed in Laudet and Ravasi, 2022).

The number and shape of white bars differ among anemonefish species, and can be divided into five categories (Klann et al., 2021; Salis et al., 2018, 2022): (1) no vertical bars (Amphiprion sandaracinos, A. akallopisos, A. ephippium), (2) one bar on the head (e.g. A. frenatus, A. perideraion, A. leucokranos), (3) two bars on the head and trunk (e.g. A. chrysopterus, A. sebae, A. akyndinos), (4) three bars on the head, trunk and the caudal peduncle (e.g. A. ocellaris, A. percula, A. biaculeatus), and (5) polymorphic bars (e.g. A. clarkii, A. polymnus, A. melanopus). Also, Amphiprion akallopisos, A. sandaracinos, A. pacificus and A. perideraion have a pattern with horizontal stripe, with a vertical head bar in some but not all of these ‘skunk’ species (Fautin and Allen, 1992, 1997; Salis et al., 2018). The relationship between phylogeny and bar patterns has been investigated by Salis et al. (2018). That study suggested that the ancestral anemonefish species had three bars, followed by species with two bars, one bar and no bars and stripe (Salis et al., 2018). Salis et al. (2018, 2022) suggest that interspecific differences in the number of white bars may function to distinguish conspecifics from other species in coexisting habitats. This is because the bar patterns are more different among sympatric species than among species living in different habitats. However, no direct empirical studies on the behavior of anemonefish have been conducted to determine whether they are able to accurately distinguish the different numbers of bars observed in other anemonefishes.

Field experiments conducted in the Ryukyu Archipelago, Japan, revealed that the duration of aggressive behavior of A. ocellaris was significantly longer when a model with white vertical bars was presented than with a model with white horizontal stripes (Hayashi et al., 2022b). Consistent with this, A. ocellaris colonies are sometimes temporarily intruded on by other species of fish (cardinalfish, damselfish and wrasses) with plain or striped color patterns, but never by fish with vertical bar patterns (Hayashi et al., 2022a,b). Amphiprion ocellaris may therefore recognize fish with bar patterns as competitors and frequently attack and chase them out to defend their host anemone. These previous studies indicated that the white bars could be an important color pattern for distinguishing competitors for territory in anemonefish. Furthermore, it is also possible that the anemonefish use finer patterns, such as the number of bars, as cues to determine whether they should chase their congenerics from the colony. According to this hypothesis, we can formulate and therefore test a specific prediction: the frequency of aggressive behavior should be higher toward potential competitors for the host, i.e. toward members of the same species, rather than members of different species, and this should be possible if the fish can identify differences in the number of white bars, as it is a prominent characteristic of each species.

The symbiotic relationship between host anemones and anemonefish is one of the most iconic phenomena in coral reef ecosystems, where multiple species are intricately intertwined (Laudet and Ravasi, 2022). Anemonefish are space limited and not recruitment limited; therefore, species recognition and aggression toward other species or the same species of anemonefish are important for understanding the mechanisms of colony formation and coexistence of multiple species. Also, if certain color patterns trigger aggressive behavior in anemonefish, then examining their behavior in response to color patterns may help us understand the mechanisms of interspecific and/or intraspecific competition. Although the conspicuous pattern of anemonefish is well known, the ecological role of the white bar pattern in anemonefish has not yet been formally elucidated (Salis et al., 2019). The purpose of the present study was to test the species recognition hypothesis, that differences in the number of white bars help anemonefish to distinguish species among their competitors. In the present experiment, we used A. ocellaris (which has three bars) as our target species. In its native range, Heteractis magnifica and Stichodactyla gigantea, the host anemones used by A. ocellaris, may also be used by several anemonefish species with one or two bar patterns that have not been reported to cohabit with A. ocellaris (Camp et al., 2016; Hayashi and Reimer, 2022; Srinivasan and Jones, 2022). Therefore, to understand the fundamental behavioral cues used in host defense, we conducted the following experiments in captivity without the intervention of other species. First, we compared whether the frequency of aggressive behavior displayed by A. ocellaris differed towards intruder fish of the same and different species. Next, we determined whether differences in the number of white bars on a model affected the frequency of aggressive behavior, i.e. whether anemonefish use the number of white bars as a cue to elicit their aggressive behavior.

Experiments were conducted at the Okinawa Institute of Science and Technology Marine Science Station in Okinawa Prefecture, southern Japan. All experimental procedures described were approved by the Okinawa Institute of Science and Technology (OIST application no. 2021-343). Fish used in these experiments were hatched and raised at this facility, and 50–150 immature anemonefish were housed and reared together in one large tank (50×70×55 cm). Thus, these individuals had seen only their own species up until the time of the experiment. Once an individual had completed the experiment, it was quarantined to ensure that it would not be used again in the experiment. Individuals used in the experiment were placed in a small observation case with a scale, and their total length (TL) was read to one decimal place before transferring them to the experimental tank. The experimental tank was surrounded by black curtains, and a 13.6 klx electric light (LED Liner Black 600, NISSO) on the surface of the tank was turned on from 07:00 to 18:00 h. Water temperature ranged from 25 to 28°C and salinity ranged from 34.7 to 35.3‰. Feeding was daily at 10:00 and 16:00 h (Probiotic Megabyte Red S).

Fifty immature A. ocellaris (TL: 3.2–5.4 cm), 10 immature A. clarkii (TL: 3.2–4.6 cm), 10 immature A. sandaracinos (TL: 3.3–4.3 cm) and 10 immature A. polymnus (TL: 2.9–4.9 cm) hatched from adult pairs (3–5 pairs) originating from Okinawa Island were used in this experiment. One individual immature A. ocellaris was acclimated for 3 days in four separate tanks (30×30×30 cm), each with one small flower pot as shelter (Fig. 1A). On the fourth day, one individual of each of the four anemonefish species (A. ocellaris, A. clarkii, A. sandaracinos and A. polymnus), the intruder fish, was placed in a small transparent case (14.5×5.0×2.0 cm) filled with water with the lid closed to block exchange of olfactory information. This small case was sunk in the center of the tank, standing vertically. To eliminate the influence of body size, individuals were introduced with a total length difference of less than 5 mm from A. ocellaris (Table S1). There were no species differences in the total lengths of the individuals used in the experiment (Kruskal–Wallis test, χ²=4.53, P=0.21). A small case containing each species of anemonefish was placed in each of the four tanks with A. ocellaris and the behavior was video recorded for 5 min (Fig. 1A). This experiment was repeated 10 times from 12 September to 1 December 2022.

Previous studies have defined aggressive behaviors such as biting and chasing in anemonefish (Hayashi et al., 2020; Mitchell et al., 2023; Wong et al., 2013), but in this experiment, the fish were separated by the case, so they could not directly bite or chase each other. Therefore, we instead defined the following behaviors as aggressive behaviors: (1) rushing towards the fish in the small case, and (2) following the fish in the small case while facing the fish frontally (frontal display) (Fricke, 1979; Hayashi et al., 2019b; Wong et al., 2013). During footage review, the video was stopped every second to see whether any aggressive behavior was observed, and the number of times (frequency) that aggressive behavior was observed during the 5-min period was counted. The duration of a single aggressive behavior and the number of times a series of aggressive behaviors was performed were also recorded.

If the behavior differs among the intruding fish species, the frequency of aggressive behavior of A. ocellaris in the tank may be affected. To determine whether the movement of intruding fish differed among species, we used Kinovea^® motion analysis software (ver. 9.3) to record the position of the intruder in the small case every 10 s (da Silva Souza et al., 2020). We manually stopped the video every 10 s on the Kinovea software and recorded the position of the eyes of the fish introduced in the small case. Using the software's measurement function, we read how much the eye position changed in 10 s. The movement of A. ocellaris in the tank was also recorded using the method described above. This was done 30 times over a 5-min period, and the average value was calculated as a measure of the amount of movement.

We used 120 immature individuals of A. ocellaris (TL: 2.1–4.0 cm) hatched from five pairs originating from Okinawa Island. These fishes were different from those used in the previous experiment. Three immature A. ocellaris were placed in each of four tanks (30×30×30 cm) (Fig. 1B). All of these individuals were grown in the same tank, and we did not distinguish which parent the three fish came from among the five pairs. An 8-day acclimation period was allowed to establish social relationships among the three individuals. During this period, aggressive behavior between individuals in the colony was recorded for 1 min at 12:00 h (to avoid feeding time in the morning and evening) daily from day 2 to day 8. A total of 40 tanks were recorded for 7 days each, but we forgot to record in all four tanks on day 1 of the first trial (Table S2). Because the frequency of aggressive behavior did not tend to increase or decrease as the days progressed, the average of the 7-day period (6-day for 4 tanks) was used for the analysis. Inside each tank, to avoid too strong intraspecific aggressivity, there was a slight difference in body size among three individuals. The largest of them was designated as the α individual with a total length of 3.17±0.33 cm (mean±s.d.), the second largest one as the β individual with a length of 2.74±0.22 cm, and the smallest as the γ individual with a length of 2.49±0.21 cm. The mean (±s.d.) difference of total length between α and β was 0.43±0.21 cm, α and γ was 0.68±0.25 cm, and β and γ was 0.25±0.13 cm.

On day 8, a different model was placed in each of the four tanks and the behavior of anemonefish was recorded for 3 min. This was carried out 10 times from 12 January to 3 June 2022, using different anemonefish each time. The models were made of resin clay (nos 4979909946404 and 4979909946398, Daiso, Japan) and the white bars and black lines were drawn with a paint marker (Mitsubishi Paint Marker PX20 and PX21). The color patterns of the model used in this experiment were no bars (plain), 1 bar, 2 bars and 3 bars, and the size of the model was 3.0 cm (Fig. S1). The reflectance spectra of background orange color were determined using a Konica Minolta CM-5 spectrophotometer (Tokyo, Japan), which was equipped with a pulsed xenon lamp and diffused illumination (geometry d/8, illuminant D65, measurement conditions SCE, measurement step 10 nm, spectral range 360–740 nm using 3.0 mm diameter) (Fig. S1). Similarly, the reflectance spectrum of the white bars was measured by the spectrophotometer Konica Minolta CM-700D (Tokyo, Japan), with a spectral range of 400 to 700 nm. Also, the three attributes of color (L: lightness, C: chroma and H: hue) of the model were measured by a colorimeter (NR-11A; Nippon Denshoku Industries, Tokyo, Japan). Measurements were conducted three times and the mean value of the base color was 48.0, 71.6 and 59.0 (L, C and H), and that of bars was 71.6, 6.7 and 111.7 (L, C and H). These models were dangled by a transparent fishing line and hung from a mobile to move erratically around the tank. Because the fish behaviors were not limited by obstacles in this experiment, ‘chasing’ and ‘biting’ were defined as aggressive behaviors, as in previous studies (Hayashi et al., 2019b, 2020, 2022b; Mitchell et al., 2023; Wong et al., 2013). During review, the video was stopped every second to see whether any aggressive behavior was observed, and the number of times aggressive behavior occurred (frequency) during the 3-min period was counted per individual.

Statistical analyses were performed on: (1) the frequency of aggressive behaviors toward other/same species of anemonefish, (2) the frequency of aggressive behaviors toward other individuals during acclimation periods, (3) the frequency of aggressive behaviors toward the model by the entire colony, and (4) the frequency of aggressive behaviors toward the model by each individual. All statistical tests were conducted using SPSS statistics ver. 25 and Microsoft Excel 2013.

In the following experiments, the Kruskal–Wallis test and post hoc multiple comparison test (Steel–Dwass method) were used following results of a test of normality indicating a lack of normality. Differences in the total instances of aggressive behavior, the duration of a single aggressive behavior, and the number of times a series of aggressive behaviors was peformed were compared among species. We tested whether there was a difference in swimming distance between different species of intruders and whether there was a difference in swimming distance of A. ocellaris in the tank when different species of intruders were introduced.

Linear regression analysis was used to determine whether there was a correlation between the average swimming distance of intruders and the frequency of aggressive behavior of A. ocellaris toward intruders, and a correlation between the average swimming distance of A. ocellaris and the frequency of aggressive behavior of them.

The frequency of aggressive behavior depending on the social rank during the pre-experimental acclimation period was tested by the Kruskal–Wallis test and post hoc multiple comparison test (Steel–Dwass method). As it is known that differences in body size affect aggressive behavior among individuals in anemonefish (Branconi et al., 2020; Buston et al., 2022), a linear regression analysis was used in this experiment to test whether there was a correlation between differences in total length (dependent variable) and the frequency of aggressive behavior (independent variable). This analysis was performed for α, β and γ individuals. In addition, a linear regression analysis was used to determine whether there was any relationship between the frequency of aggressive behavior toward each social rank. For example, if the frequency of aggressive behavior from α to β increases, then the frequency of aggressive behavior from β to γ increases, or if the frequency of aggressive behavior from α to γ increases, then the frequency of aggressive behavior from β to γ increases. Using the non-parametric Mann–Whitney U-test, we compared the frequency of aggressive behavior from α individuals with that from β individuals and from α individuals with that from γ individuals.

Differences in the frequency of aggressive behavior towards the four types of models (no bars, 1 bar, 2 bars, 3 bars) in the entire colony was tested by the Kruskal–Wallis test and post hoc multiple comparison test (Steel–Dwass method). To test whether there was an effect of the number of bars even when differences in the potential aggressive behavior in the colony were excluded, a one-way ANCOVA and a post hoc analysis (Bonferroni tests) was conducted with the frequency of aggressive behavior in a colony during the pre-experimental acclimation period as a covariate.

To determine the factors affecting differences in aggressive behavior among individuals toward the model, two-way ANOVA using two factors, social rank (α, β, γ) and model type (number of bars), was performed. Where there were significant differences, a post hoc analysis using the Bonferroni method was performed. To test whether there was an effect of the number of bars even when size differences relative to the model were excluded, a one-way ANCOVA was performed with size difference as a covariate.

We first compared the aggressive behavior displayed by A. ocellaris toward other species of anemonefish that differ in terms of color pattern and, in particular, number of bars. Total frequency of aggressive behaviors by A. ocellaris significantly differed towards each intruder species (Kruskal–Wallis: χ²=13.66, d.f.=3, P<0.01), with the highest frequency being towards A. ocellaris (Fig. 2). The post hoc Steel–Dwass tests indicated that frequency of aggressive behavior toward A. ocellaris was significantly higher than toward A. sandaracinos (Table 1, Fig. 2). There was also a significant difference in the number of aggressive behaviors, with that toward A. ocellaris being significantly greater than that toward A. sandaracinos (Kruskal–Wallis test, χ²= 15.14, d.f.=3, P<0.01; Table 1), but there was no significant difference in the duration per attack (Kruskal–Wallis test, χ²=5.55, d.f.=3, P=0.14; Table 1). The duration of aggressive behavior by A. ocellaris was 1 s in 113 cases (60.8%), 2 s in 36 cases (19.3%) and more than 3 s in 37 cases (19.9%). The longest aggressive behavior, lasting 11 s, occurred in only one case. Although the differences were not statistically significant, the effect sizes between A. ocellaris and A. polymnus or A. clarkii were greater than 0.50, indicating a large effect (Table 1) (Mizumoto and Takeuchi, 2008; Maher et al., 2013). There was no significant species difference in the mean swimming distance of intruder individuals (Kruskal–Wallis test, χ²=2.62, d.f.=3, P=0.45). Similarly, there was no significant correlation between the mean swimming distance of intruder individuals and the frequency of aggressive behavior of A. ocellaris (R²=0.0004, N=40, P=0.90). There was no significant species difference in the mean swimming distance of resident A. ocellaris (Kruskal–Wallis test, χ²=1.92, d.f.=3, P=0.59), and there was a slight negative correlation between the mean swimming distance and the frequency of aggressive behavior (R²=0.01, N=40, P=0.05).

The frequency of aggressive behavior toward other individuals in the colony during the pre-experimental acclimation period differed significantly with the social rank (Kruskal–Wallis test, χ²=97.41, d.f.=2, P<0.01). α individuals (the largest individuals) were significantly more likely to engage in aggressive behavior toward others than β (the second largest) and γ individuals (the smallest) (both P<0.01, Steel–Dwass test), and β individuals significantly more than γ individuals (P<0.01, Steel–Dwass test). For α and γ individuals, there was no significant correlation between total length difference and frequency of aggressive behavior toward others (α: R²=0.01, P=0.33, γ: R²=0.01, P=0.42; Fig. 3A,C). β individuals showed a significant positive correlation between length difference and the frequency of aggressive behavior, and tended to engage in aggressive behavior only when their own length was larger than other individuals, i.e. against γ individuals (R²=0.24, P<0.01; Fig. 3B). The frequency of aggressive behaviors from β to γ individuals was not influenced by the increase in frequency from α to β individuals (R²=0.08, N=40, P=0.08), nor that from α to γ individuals (R²=0.04, N=40, P=0.23). There was no significant difference in the frequency of aggressive behavior of α individuals toward β individuals (mean±s.d.: 5.10±4.18, N=40) and γ individuals (4.73±4.56, N=40) (Mann–Whitney U-test: U=863.5, P=0.54).

The total frequency of aggressive behavior per colony toward the model significantly differed by the number of white bars of the model (Kruskal–Wallis test, χ²=19.94, d.f.=3, P<0.01). The frequency of aggressive behavior toward 1-, 2- and 3-bar models was significantly higher than that toward the no-bar model (Table 2, Fig. 4). The frequency of aggressive behavior displayed by A. ocellaris throughout the colony was highest toward the 3-bar model (the same as themselves), but there was no significant difference between the 3- and 2-bar models. There was no significant difference in the frequency of aggressive behavior between the 3- and 1-bar models, but the effect size was greater than 0.5 (Table 2).

To test whether there was an effect of the number of bars even when differences in the potential aggressive behavior in the colony were excluded, a one-way ANCOVA was conducted using the mean frequency of aggressive behavior in the pre-experimental period as a covariate. There was a significant positive correlation between the mean frequency of aggressive behavior during the pre-experimental period and the frequency of aggressive behavior toward the model in both cases, one including all data and the other excluding outlier data (Fig. 5, Table 3). We then compared the frequency of aggressive behavior across model types and rejected the null hypothesis of no difference in the frequency of aggressive behavior among the four models (P<0.001; Table 3). Therefore, even excluding the effect of potential frequency of aggressive behavior, the number of bars in the model was a significant contributing factor to the difference in the frequency of aggressive behaviors in the colony. A post hoc analysis revealed that the frequency of aggressive behavior for the no-bar model was significantly smaller than that for the 1-, 2- and 3-bar models (Bonferroni, P<0.05, P<0.001 and P<0.001, respectively) and the frequency of aggressive behavior toward the 1- and 3-bar models was also significantly different (Bonferroni, P<0.05) (Fig. 5).

Next, we analyzed the frequency of aggressive behavior toward the model by individual members. A two-way ANOVA using two factors (social rank and model type) indicated that the frequency of aggressive behavior by individual fish was significantly affected by both variables (Table S3). There was no significant interaction between the two variables, rank and number of bars (F=1.42, d.f.=6, P =0.21), but significant differences were found in aggressive behavior for social rank (F=5.36, d.f.=2, P<0.01) and number of bars (F=15.22, d.f. =3, P <0.001). By multiple comparisons, there were significant differences in the frequency of aggressive behaviors between no-bar and 2-bar (Bonferroni, P<0.001), no-bar and 3-bar (P<0.001), and 1-bar and 3-bar models (P<0.01). By multiple comparisons, there was also a significant difference in the frequency of aggressive behaviors only between α and γ individuals (Bonferroni, P<0.01). There was no significant relationship between the total length difference and the frequency of aggressive behavior toward the models (F=0.74, d.f.=1, P=0.39).

Our study showed that A. ocellaris, raised in captivity and which had only seen individuals of their own species, exhibited aggressive behaviors toward their own species more frequently than toward other species of anemonefish, and the frequency of aggressive behavior toward the model with 3 bars was longest. This indicates that, at the very least, A. ocellaris may be able to distinguish species similar to those they have seen from species they have not seen. These results support one of the three hypotheses regarding the function of the white bar in anemonefish – species recognition – but the other two hypotheses (disruptive coloration hypothesis and aposematism hypothesis) were not tested in this experiment.

The orange and white band pattern of anemonefish reflects UV light well, and anemonefish color vision has been reported to discriminate between orange and white contrasts under UV light (Mitchell et al., 2023; Stieb et al., 2019). The frequency of attacks by anemonefish A. akindynos on individuals smaller than themselves was greater under UV filters than under neutral density (ND) filters that reduce the intensity of all wavelengths equally, and dominance was determined by the degree of UV reflection of the white bar, suggesting that the white bar controls social behavior in the colony (Mitchell et al., 2023). According to Hayashi and Reimer (2022), when white horizontal and white vertical bar models were shown to A. ocellaris colonies in the wild, the frequency of aggressive behavior toward the white vertical bar was significantly higher than that toward the white horizontal bar, suggesting that the anemonefish can discriminate between horizontal and vertical bars. These previous studies have shown that anemonefish recognize white bars, but it is unclear whether they recognize the number of bars or even their shape. The present study suggested that anemonefish may recognize the number and shape of white bars, at least in LED lighting with no UV or ND filters.

A characteristic of the social groups of anemonefishes is that breeding pairs allow non-breeding individuals to coexist, but can regulate their growth to avoid competition (Buston et al., 2022). Also, this experiment shows that aggression may be used to maintain their own species in a non-breeding stage, but not to evict an individual from the anemone. The size ratio of individuals in A. percula colonies was shown to be maintained by precisely controlling the growth of subordinate individuals, and dominant individuals may evict or kill subordinate individuals (Buston, 2003; Buston and Cant, 2006). Buston et al. (2022) considered that non-breeding individuals continue to persist despite severe attacks from dominant individuals because the number of host anemones is limited, and migration is risky and unlikely to be successful. Immature anemonefish sometimes enter the hosts already occupied by mature pairs of their own species, but there have also been cases of single or multiple immatures settling in vacant hosts with no mature pairs, and in hosts living with other species of mature pairs (Hayashi et al., 2019a). During the formation of social groups in A. percula, individuals coexisting with similar-sized rivals grew more than individuals living alone (Buston, 2002). Field removal and introduction experiments have shown that anemonefishes are sensitive to small differences in body size and change their aggressive behavior accordingly (Branconi et al., 2020; Buston et al., 2022). Our study supports these findings by showing that immature fish with little size difference (total length difference was only 0.43 cm between α and β, and 0.25 cm between β and γ) established a ranking relationship during a short acclimation period. In our study, the ranking relationship was established by the following simple but reliable mechanism. Individuals with the largest size (α) showed the most frequent aggressive behavior toward other individuals and attacked regardless of the size of the opponent. Individuals with intermediate size (β) accurately judged and attacked only individuals smaller than themselves, and the smallest individuals (γ) rarely attacked.

During the pre-experimental period, there were differences in the frequency of aggressive behavior between colonies, and colonies with higher levels of aggressiveness tended to show higher frequencies of aggressive behavior toward the model. Furthermore, the largest α individual most frequently engaged in aggressive behavior toward the model, whereas the smallest γ individuals showed almost no aggressive behavior toward the model. Interestingly, the frequency of aggressive behavior did not vary with size difference from the model, and α individuals showed aggressive behaviors most frequently regardless of size difference, suggesting that the behavior of α individuals is activated by their social rank. Using 3-month-old individuals and in accordance with our results, Chen and Hsieh (2017) reported that α individuals attacked other individuals in the colony most frequently. Even taking these effects (potential aggressiveness within a colony and social rank) into account, there were significant differences in the frequency of aggressive behavior depending on the number of bars of the model. Thus, the number of white bars function as an important cue for immature A. ocellaris to control their aggressive behaviors.

Competition over a host anemone occurs most frequently among individuals within a species (Buston et al., 2022). Therefore, it makes sense for A. ocellaris to identify and behave aggressively toward their own species (having three bar patterns). However, multiple species of anemonefish are often living in sympatry, sometimes sharing the same host species (e.g. Camp et al., 2016; Elliott and Mariscal, 2001; Hayashi et al., 2018, 2021; Srinivasan and Jones, 2022). Competition for the host anemone thus occurs not only among individuals of the same species, but also among individuals of different congeneric species.

Although A. ocellaris do not cohabit with different species of anemonefish in the Ryukyu Archipelago (Hayashi et al., 2018, 2021), A. clarkii cohabit with other species of immature anemonefish (A. sandaracinos and A. perideraion), a situation referred to as ‘cohabitation’ (e.g. Camp et al., 2016). When models of the same shape, color and size, differing only in the number of white bars, were presented to A. ocellaris, the frequency of aggressive behavior toward the model without bars was the lowest. This is consistent with the lowest frequency of aggressive behavior being toward A. sandaracinos, a species without white bars, in the experiment. It has been shown that immature A. sandaracinos inhabit the same host anemone as adult A. clarkii (Camp et al., 2016; Hayashi et al., 2018, 2021). Species lacking white bars such as A. sandaracinos might have an advantage to intrude upon the host anemone used by other species of anemonefish, and this could inform future investigation for understanding the function of white bars.

According to Hattori (2002), interspecific aggression was observed more frequently than intraspecific aggression in α individuals of A. clarkii. According to Camp et al. (2016), however, colonies with a mixture of A. clarkii and A. perideraion have a lower frequency of aggressive behavior by dominant individuals on other individuals. These field studies addressed interspecific and intraspecific aggression in a mixed colony of already established A. clarkii and A. perideraion (Camp et al., 2016; Hattori, 2002). However, many interspecific and intraspecific relationships, although not studied in the field, are thought to occur during the immature stage, prior to establishment in the host. Our study demonstrates that species recognition is possible even in inexperienced individuals of immature fish, an important trait in the colony establishment of anemonefish.

Because individuals used in this experiment were hatched and raised in an environment where they had seen only the same species, it was not possible to conclude whether this behavior was innate or acquired. Amphiprion ocellaris use the same species of host anemone as A. perideraion and A. clarkii in the wild (e.g. Burke da Silva and Nedosyko, 2016; Hattori, 2011; Hayashi and Reimer, 2022). Therefore, it is expected that under natural conditions, in addition to A. ocellaris, A. clarkii and A. perideraion often approach the host anemone that A. ocellaris inhabit. Unlike A. ocellaris kept in tanks where other fish were not present, the frequency of aggressive behavior of A. ocellaris toward A. clarkii and A. perideraion in the wild may be more frequent than that toward conspecifics. Future research is needed to determine whether this aggressive behavior toward their own species is innate or acquired, both in the wild and in captivity.

In the present experiment, A. ocellaris showed a higher frequency of aggressive behavior toward models with 2 or 3 bars, which may be related to the developmental process of white bars in A. ocellaris. The white bars in A. ocellaris do not form one by one: A. ocellaris has two white bars visible from the early stages of body coloration (stage 5: about 11 days after hatching), and three white bars are formed after about 14 days (stage 7) after hatching (Roux et al., 2022; Salis et al., 2018, 2021). Amphiprion ocellaris takes about 3 days to develop from two white bars to three white bars, and is characterized by having the same number of white bars early in development as in adulthood (Roux et al., 2022; Salis et al., 2018, 2021, 2022). It is thought that A. ocellaris attacked more frequently 2- and 3-bar models in this experiment than no-bar and 1-bar models because during their developmental stage they have often seen individuals with 2 or 3 bars as competitors (Roux et al., 2022; Salis et al., 2018, 2021, 2022). Some species of anemonefish have different body coloration when they are immature and as adults; for example, A. frenatus (but also other species such as A. ephippium, A. allardi or A. melanopus) have more white bars during the larval stage than in subadult and adult stages (Salis et al., 2018). Because the number of white bars on the models that would elicit aggressive behavior in A. frenatus may differ between the immature and adult periods, similar experiments need to be done using this species.

The present experiment showed that immature anemonefish exhibit more frequent aggressive behavior toward their own species than toward other species, and differences in the number of white bars caused differences in the frequency of aggressive behavior. Our results support the notion that anemonefish are able to discriminate between different numbers of white bars, and the differences in the number of white bars helps anemonefish distinguish their own species.

Lilian Carlu, Hiroki Takamiyagi, James Hutasoit and Germain Salou (OIST: Laudet Unit) are thanked for very efficient help in taking care of fish. We also thank Marleen Klann, Manon Mercader (Laudet Unit) and Natacha Roux (OIST: Reiter Unit), who gave advice during this study. Marcela Herrera Sarrias and Laurie Mitchell are thanked for reading and checking an earlier version of the manuscript. Okinawa Institute of Science and Technology allowed us to use their breeding facility.