Lettuce get along: effects of feeding enrichment on green sea turtle behavior and interactions

Green sea turtles (Chelonia mydas) are primarily solitary in the wild, although they may exhibit tolerance toward conspecifics within shared habitats. Adult males are particularly prone to agonistic interactions, especially when resources such as food, space, or mates are limited. In captive environments, factors including high population density, limited enclosure size, skewed sex ratios, and insufficient environmental enrichment may exacerbate aggressive behaviors, resulting in increased stress and risk of injury. Appropriate husbandry practices such as the provision of large, structurally complex habitats with numerous resting zones, multiple feeding stations, and environmental enrichment have been shown to mitigate aggression and improve welfare outcomes. This case study implemented scan and all-occurrence sampling methods to document the behavior and space use of two adult male green sea turtles preceding, during, and following three feeding conditions: baseline, depth pole, or floating ice block. Behavioral variation was driven primarily by feeding phase and time of day rather than by enrichment condition, with turtles exhibiting anticipatory activity pre-feeding, active engagement during feeding, and increased resting post-feeding. Agonistic interactions were relatively infrequent but occurred most often after feeding compared to preceding and during feeding only for the depth pole enrichment condition. These results contribute to the limited body of research on the effects of feeding enrichment on captive male green sea turtle interactions, and can inform aquarium management aimed at minimizing aggression and improving animal welfare.

Commonly thought of as being relatively solitary in the wild (Mullaney et al., 2024), recent research reveals the capacity for more complex social behaviors in Green sea turtles (Chelonia mydas) than previously expected. For example, they have been observed engaging in consistent and passive interactions, demonstrating tolerance toward conspecifics when foraging in the wild (Mullaney et al., 2024; Thomson et al., 2015). However, dominance behaviors, including chasing, snapping jaws, biting of the flippers, neck, or plastron, and raking with the beak (Thomson and Heithaus, 2014; Thomson et al., 2015) can emerge when resources are contested. Agonistic behaviors tend to increase during mating and nesting periods (Jessop et al., 1999). Adult males, relative to females or juveniles, are particularly likely to exhibit intraspecific aggression, and consequently face a higher risk of injury (Jessop et al., 1999), such as bite wounds, shell damage, and subsequent infection, as well as chronic stress if not addressed (Kawazu et al., 2022). For turtles under managed care, prompt veterinary care is required in the case of aggressive interventions, as medical issues such as localized swelling from bite wounds, lethargy, poor appetite, excessive surfacing (as a sign of stress) or progressive shell damage can occur (Kawazu et al., 2022).

In captivity, agonistic encounters may be exacerbated by space limitations and restricted access to resources (Schofield et al., 2022) due to factors including high population density, limited tank size, skewed sex ratios, and inadequate enrichment (Kawazu et al., 2022). Insufficient enclosure size may prevent individuals from dispersing, thereby increasing competition and interaction frequency. Additionally, the presence of females or a higher ratio of males to females can elevate stress and aggression levels (Wood, 2022). Along with sex imbalances, size disparities may result in larger individuals dominating and injuring smaller conspecifics (Jessop et al., 1999). Managing population composition by separating individuals based on sex, size, and age can reduce conflict (Wood, 2022). Particularly when housing male turtles together, it is beneficial to keep group sizes low and avoid large size disparities (Schofield et al., 2022). Accordingly, appropriate husbandry practices such as provision of large, structurally complex exhibits, multiple feeding stations, environmental enrichment and close monitoring, are essential for mitigating these interactions (Hill and Broom, 2009).

Enrichment and environmental complexity can reduce injury rates and promote more natural behaviors in captive animals (Diggins et al., 2022; Escobedo-Bonilla et al., 2022; Kanghae et al., 2021). Sufficient usable space is critical because it allows for natural movement and behavior and reduces forced, stressful interactions between co-housed animals (Browning and Maple, 2019; Shishis et al., 2022). Inadequate environments are linked to behavioral and physiological welfare problems, whereas appropriate exhibit size improves welfare (Ross et al., 2009). Poor water quality, inadequate diet, and lack of enrichment have all been associated with increased aggression and stress-related behaviors in captive sea turtles (Wood, 2022). Rest areas are of particular importance, as space use throughout an exhibit is not always evenly distributed (Gübert and Dierkes, 2024). Structural complexity, including the addition of rocks, shelters, ledges, and visual barriers, allows individuals to avoid one another and establish separate activity zones (Escobedo-Bonilla et al., 2022; Smith, 2023). Habitat complexity has been linked to changes in spatial use and behavior, with turtles utilizing a greater range of their environment when structural features are present (Zug, 1971). The availability of individual resting zones and basking areas further supports physiological and behavioral needs.

Beyond structural requirements, environmental enrichment is a critical component of welfare in captive sea turtles. Enrichment may include objects to explore, variable substrates, and feeding or foraging activities that encourage exploration and problem solving (Escobedo-Bonilla et al., 2022). These interventions have been shown to decrease agonistic encounters between sea turtles while increasing natural behaviors (Escobedo-Bonilla et al., 2022; Kanghae et al., 2021). Wild green sea turtles are primarily herbivorous, grazing on algae and seagrass beds (Hearne et al., 2019). Feeding strategies that allow these natural behaviors play a key role in reducing competition. The use of target feeding, a husbandry technique in which animals are trained to associate a visual cue with individual food delivery, allows for controlled, individualized feeding and reduced competition. Dispersed feeding methods, such as scatter feeding, has also been shown to decrease food-related aggression by minimizing direct competition (Costa et al., 2023; Escobedo-Bonilla et al., 2022). Similarly, enrichment-based feeding approaches can promote natural foraging behaviors while reducing antagonistic interactions. Experimental studies on sea turtle enrichment have demonstrated that novel stimuli and foraging opportunities increase activity levels, encourage diverse behaviors, and reduce aggression (Monreal-Pawlowsky et al., 2017; Therrien et al., 2007). These findings are consistent with broader captive animal welfare research, which shows that enrichment can reduce stress, improve behavioral outcomes, and support overall health (Hill and Broom, 2009; Mellor and Reid, 1994).

Despite research on sea turtle ecology, the social dynamics of captive male green sea turtles remain less understood. However, given that many captive sea turtles remain in human care for life due to non-releasable medical conditions, improving husbandry practices is essential. In particular, there is limited research on how co-housed males interact over extended periods in controlled environments, and how various enrichment strategies may impact these interactions. Behavioral studies have demonstrated that observable changes in activity and interaction patterns are key indicators of welfare (Miller et al., 2020; Watters et al., 2009), specifically in captive sea turtles (Inoue, 2025), reinforcing the importance of systematic behavioral monitoring in captive environments. This case study aims to provide preliminary information to address this gap by examining the behavioral interactions between two co-housed captive male green sea turtles at the Sea Life Michigan Aquarium. We observed the effects of two feeding enrichments in the form of supplementary food placed on a single depth pole to chase or frozen in two ice blocks, in comparison to their daily standard target feeding. We recorded the frequency of various individual behaviors, space use, and social interactions before feeding, during feeding, and post feeding. We hypothesized a reduction in adverse behaviors during and following the two enrichment conditions compared to the pre-feeding condition, but not for the baseline condition.

Study site and subjects

The subjects were two adult male green sea turtles, Benson and Carr, permanently co-housed at SEA LIFE Michigan Aquarium in Auburn Hills, Michigan, USA. Both individuals are non-releasable due to prior sustained boat-related injuries resulting in positive buoyancy syndrome (Flint et al., 2017). They reside in the Open Ocean Exhibit, a large, multi-species tank designed to simulate naturalistic oceanic conditions and support long-term animal welfare. In addition to the two focal subjects, the tank includes roughly 200 teleost fish (including goldentrevallies (Gnathanodon speciosus), tangs (Naso sp.), a marbled grouper hybrid (Epinephelus microdon x Epinephelus sp), a spot-fin porcupinefish (Diodon hystrix), 20 elasmobranchs, including eight cownose rays (Rhinoptera bonasus), two southern stingrays (Hypanus americanus), a female shovelnose guitarfish (Glaucostegus typus), six black tip reef sharks (Carcharhinus melanopterus), a male banded hound shark (Triakis scyllium), and a female nurse shark (Ginglymostoma cirratum). The turtles have been housed together in this dynamic mixed-species environment for approximately eight years. The aquarium is open to the public seven days a week from 10:00 to 17:00. The Open Ocean Exhibit is viewable from public viewing areas.

Exhibit

The Open Ocean tank has a volume of approximately 473, 000 L and a depth of up to 4.5 m. Environmental parameters are maintained within narrow ranges, with the temperature held at approximately 23–24.5 °C, salinity at 28–30 ppt, dissolved oxygen near saturation (~98%), and a controlled photoperiod of approximately 14:10 (light:dark). The tank contains structural complexity, including rock formations, ledges, and open swimming areas, allowing for both horizontal and vertical movement throughout the water column. These features provide opportunities for resting, avoidance, and spatial separation between individuals. The tank includes varied viewing areas, such as multiple windows for guests to view the animals. This includes a submerged visitor tunnel, around 2.1m in height, where visitors walk through the tank. This tunnel also provides tank inhabitants the ability to rest on the tunnel substrate and vary their swimming depth when swimming over it, contributing to differences in space use within the habitat.

The habitat also includes an out-of-sight Med Bay area, accessed by the turtles and other tank inhabitants through an opening in the tank wall. The Med Bay is oval in shape with a volume of approximately 11, 400 l and a depth of about 1.2 m. It did not contain any permanent structures or enrichment but enrichment could be provided there if desired. The turtles have regular, but not constant, access to the Med Bay. Access is controlled by a gate, which can be raised or lowered to allow or block access. For the majority of the observations conducted during this study, the Med Bay gate remained closed, allowing for almost constant visibility of the turtles. On rare occasions, one or both turtles were taken into the Med Bay area during observation sessions, during which turtle behavior was recorded as ‘out-of-sight’.

Procedure

Prior to the start of data collection, the observers (I. B. and L. Z.), who were not involved in the care of the animals, participated in six practice observation sessions simultaneously to establish reliability. Following training, behavioral observations were conducted three times per week each week from January 20, 2026 to April 9, 2026. A total of 108 observations were conducted (36 for each condition with 12 for each phase within each condition). Each observation lasted 30 minutes and included 15 two-minute intervals for both turtles simultaneously. Scan sampling occurred within two-minute intervals to record species typical behaviors; swimming, pacing, breathing, resting, sleeping, feeding, feeding from enrichment (Table 1), water column location, and spatial position.

All-occurrence sampling documented social interactions, particularly agonistic behaviors, as they occurred (Table 2). These behavioral interactions included the approach of one turtle and the resulting displacement of the other turtle, chasing and corresponding withdrawal behaviors, and contact aggression such as ramming, snapping, or biting. Neutral or social interactions, if present, were also recorded such as gentle contact between the turtles or time spent in close proximity in the absence of aggression. A total of 3, 350 raw data points were collected between both focal subjects.

Data were collected systematically to ensure that each experimental condition (baseline [no enrichment]), depth pole, and floating ice blocks] was observed equally often during morning (AM) and afternoon (PM) sessions over two-week blocks in which the order of observations differed. Observations occurred from either 11:00-13:30 (categorized as an AM observation) or 14:00-15:30 (categorized as a PM observation). Thus, all observations occurred during opening hours. On each test day, three observations were conducted; one preceding feeding, one during feeding and one following feeding (pre, during and post-feeding phases). Target feeding occurred once during every observation session. Target feeding involved the focal subjects each being fed via tongs at an individual controlled target location (Costa et al., 2023). Enrichment occurred following a pseudo random schedule such that the same enrichment did not always occur on the same weekdays directly following target feeding. For the depth pole condition, a PVC pole with evenly-spaced holes (approximately 122cm in height) was filled with lettuce and held suspended in the tank by a SEA LIFE Michigan aquarist, encouraging turtles to swim up and forage from or chase the pole as it was moved side to side. The depth pole enrichment typically lasted for 2–4 minutes of active feedings. For the floating ice block enrichment, two vegetable ice blocks (containing ~250 grams of different vegetables frozen in ice) were dropped into the tank to encourage each focal subject to individually engage, chase and exhibit foraging behaviors. The ice blocks typically lasted for about 5 minutes before the food was consumed.

Data were recorded using the ZooMonitor program (Ross et al., 2016) for species-typical behaviors and behavioral interactions, water-column usage, and animal location via heat map. The location of both turtles according to the level of the water column (upper or lower) and position within the tank was recorded using ZooMonitor’s mapping function to document the location of each turtle within the exhibit during scan behaviors (X/Y axes). Vertical positioning (Z-axis) was recorded as upper vs. lower water column. A map of the Open Ocean Exhibit was uploaded into ZooMonitor to document the turtles’ location within the aquarium (Figure 1). The resulting heat maps were used to identify areas of overlap, avoidance zones, and preferred locations within the exhibit. For all three observation sessions, both observers moved throughout the guest observation area to maintain continuous viewing of the turtles.

Data analysis

For space use, we conducted two chi-square analyses to examine the association between spatial location and the category of behavior occurring within the interval; one for the horizontal and one for the vertical coordinates according to the ZooMonitor maps.

For the interval behaviors, we conducted a series of linear mixed models in SPSS v. 30 with session as a random effect and fixed effects of subject, condition (baseline, depth pole, floating), phase (pre, during, post) and time of day (AM, PM) along with all two-way interactions. We modeled session as a random intercept nested within subjects with variance components for the variance structure. We used a Satterthwaite approximation for the relatively small number of observations. We included subject as a fixed effect because, with only two subjects, the random effect variance estimates would not be stable, making tests unreliable. We conducted mixed models with the same effects for each of the following outcomes: proportion of time spent on the bottom of the tank, time spent breathing, pacing, resting/sleeping, swimming and out of sight. There were not enough observations of self-grooming to include. We examined pairwise comparisons to evaluate main effects and interactions.

We compared these models to null models and simpler models without the interactions, using corrected AIC to compare the models.

For feeding and feeding from enrichment outcomes, we ran mixed models for only the observations during feeding with fixed effects of subject, condition (depth pole and floating ice block only), and time of day using the same model specifications as above.

There were relatively few instances of all occurrence behaviors across the observations, making it inappropriate to analyze specific behaviors independently. Therefore, we evaluated whether any adverse behavior had been perpetrated by each subject, coding this as 1. We coded all intervals in which no adverse behavior occurred as 0. However, dichotomous generalized linear mixed models would not converge given the unbalanced nature of the data. Therefore, we conducted chi square tests of independence to evaluate whether adverse events were distributed evenly across phases for each condition.

Reliability

Inter-observer reliability was tested across six study sessions utilizing the Reliability Analysis feature in ZooMonitor. Across the six sessions, the percent agreement averaged 82.35% for space use, 95.84% for depth within the tank, and 95.42% for the scan interval behaviors, and. Both observers continued to collect data throughout the observation period.

Spatial location of behaviors

Chi-square analyses was conducted to examine the association between spatial location and the type of behavior occurring within the interval. There was a significant association between horizontal space use and behavior, χ²(70) = 742.32, p <.001, as well as between vertical space use and behavior, χ²(90) = 866.32, p <.001. These results indicate that behaviors did not occur evenly throughout the tank. This was confirmed by visual inspection of heatmaps generated from ZooMonitor. Figure 2 presents the turtles’ use of space across the three phases for each of the three conditions. It appears that Carr, in particular, used more of the habitat during and following enrichment compared to baseline and pre-feeding conditions.

Spatial position

Subjects differed in how much time they spent at the bottom of the tank, F (1, 166.62) = 12.35, p <.001. Estimated marginal means reveal that Carr (M = 28.48, SEM = 1.90) spent more time at the bottom of the tank compared to Benson (M = 19.87, SEM = 1.90). Subjects differed in where they spent their time across phases, F (2, 104.71) = 55.51, p <.001, and time of day F (1, 141.98) = 5.71, p =.018. The turtles spent more time at the bottom during (M = 17.68, SEM = 1.84, MD = -11.04, p <.001, 95% CI: [-17.38, -4.69]) and post feeding (M = 48.21, SEM = 3.49, MD = -41.57, p <.001, 95% CI: [-51.19, -31.94]) compared to pre feeding (M = 6.64, SEM = 1.86). They also spent significantly more time at the bottom of the tank in the mornings (M = 27.65, SEM = 2.06) compared to the afternoons (M = 20.71, SEM = 2.06, MD = 6.94, SEM = 2.91, p =.02, 95% CI: [1.20, 12.69]). There was no significant main effect of condition and no significant interactions. The corrected AIC for the full model was 1845.90, which was an improvement over the model without the interaction terms (AIC = 1877.92) and the null model (AIC = 1978.51).

Behaviors

Swimming

Subjects differed significantly in swimming behavior, F (1, 180.43) = 55.36, p <.001, with Benson (M = 60.03, SEM = 1.98) spending more time swimming than Carr (M = 39.70, SEM = 1.98). There was a significant main effect of phase, F (2, 134.08) = 7.37, p <.001. Pairwise comparisons revealed that swimming was significantly more frequent during pre-feeding (M = 57.99, SEM = 2.75) compared to during-feeding (M = 45.89, SEM = 1.95, MD = 12.10, SEM = 3.37, p =.001, 95% CI: [3.92, 20.29]) and post-feeding (M = 45.71, SEM = 2.66, MD = 12.28, SEM = 3.83, p =.005, 95% CI: [3.00, 21.56]). There was no significant main effect of condition, or time of day, and no significant interactions. The corrected AIC for the full model was 1846.51, which was an improvement over the model without the interaction terms (AIC = 1899.32) and the null model (AIC = 1936.69).

Pacing

The subjects did not differ significantly in pacing behavior. There was a significant main effect of phase F (2, 128.62) = 28.03, p <.001. Pairwise comparisons revealed that pacing was significantly higher during pre-session periods (M = 30.35, SEM = 3.12) compared to both during (M = 9.12; SEM = 1.36, MD 21.24, p <.001, 95% CI: [12.94, 29.53]) and post-session periods (M = 4.70; SEM = 1.43, MD = 25.65, p <.001, 95% CI: [17.30, 34.01]). There was no significant main effect of condition, or time of day and no significant interactions. The corrected AIC for the full model was 1731.74, which was an improvement over the model without the interaction terms (AIC = 1781.55) and the null model (AIC = 1846.48).

Breathing

The subjects differed in how much time they were observed breathing, F (1, 183.29) = 5.06, p =.03 with Benson (M = 2.75, SEM = 0.40) observed breathing more often than Carr (M = 1.50, SEM = 0.40). There were significant main effects of condition F (2, 181.04) = 3.79, p =.02, and time of day, F (1, 181.04) = 12.90, p <.001. Pairwise comparisons indicated that breathing was higher in the depth pole condition (M = 3.18, SEM =.51) compared to baseline (M = 1.22; SEM =.51, MD = -1.96, p =.02, 95% CI: [-3.69, -0.23]). Additionally, subjects spent more time breathing during morning sessions (M = 3.18, SEM =.41) than afternoon sessions (M = 1.08; SEM =.41, MD = 2.10, SEM =.59, p <.001, 95% CI: [0.95, 3.26]). There was no significant main effect of phase and no significant interactions. The corrected AIC for the full model was 1206.52, which was an improvement over the model without the interaction terms (AIC = 1235.63) and the null model (AIC = 1264.46).

Resting/sleeping

The subjects differed significantly in resting and sleeping behavior, with Carr (M = 18.35, SEM = 1.43) spending more time resting and sleeping than Benson (M = 12.07, SEM = 1.43). There was a significant main effect of phase, F (2, 95.04) = 48.35, p <.001. Pairwise comparisons revealed that resting and sleeping were significantly more frequent post-feeding (M = 33.94, SEM = 3.14) compared to during feeding (M = 9.13, SEM = 1.48; MD = 24.81, SEM = 3.47, p <.001, 95% CI: [16.36, 33.27]) and pre-feeding (M = 2.55, SEM = 0.96; MD = 31.39, SEM = 3.28, p <.001, 95% CI: [23.38, 39.41]). Resting and sleeping were also significantly more frequent during compared to pre-feeding (MD = 6.58, SEM = 1.76, p <.001, 95% CI: [2.30, 10.87]). There was no significant main effect of condition or time of day and no significant interactions. The corrected AIC for the full model was 1691.66, which was an improvement over the model without the interaction terms (AIC = 1740.01) and the null model (AIC = 1828.23).

Out of sight

The subjects did not significantly differ in time spent out of sight. There was a significant main effect of condition F (2, 171.69) = 6.89, p =.001, phase, F (2, 113.68) = 4.37, p =.02, and time of day, F (1, 171.69) = 20.20, p <.001. However, these main effects were qualified by significant interactions between condition and phase, F (4, 113.68) = 3.20, p =.02, and between condition and time of day, F (2, 177.11) = 3.24, p =.04. Pairwise comparisons revealed, only for the baseline condition, turtles were out of sight more post-feeding (M = 24.85, SEM = 3.90) compared to pre-feeding, (M = 3.95, SEM = 2.83; MD = -20.91, p <.001, 95% CI: [9.21, 32.60],) and during the feeding, (M = 11.38, SEM = 2.40, MD = 13.47, p =.01, 95% CI [2.34, 24.60]. See Figure 3. Only for the baseline condition, turtles were out of sight more often during the afternoon (M = 21.49, SEM = 2.47) compared to the morning, (M = 5.30, SEM = 2.47, MD = -16.19, p <.001, 95% CI [-22.90, -9.47]). See Figure 4. The full model had a corrected AIC of 1708.63, which was an improvement over the model without the interaction terms (AIC = 1770.68) and the null model (AIC = 1817.02).

Feeding

There was a significant main effect of time of day, F (1, 65) = 4.61, p =.04 Pairwise comparisons revealed that feeding was significantly higher during the morning (M = 22.32, SEM = 1.12) compared to the afternoon (M = 18.92, SEM = 1.12). The subjects did not differ significantly in feeding behavior. There was no significant main effect of condition, and there was no significant interaction between condition and time of day. The full model had a corrected AIC of 454.08, which was an improvement over the model without the interaction term (AIC = 468.48) and the null model (AIC = 486.31).

Feeding through enrichment

The subjects did not differ in feeding through enrichment. However, there was a significant effect of time of day, F (1, 43) = 14.64, p <.001, which was qualified by a significant interaction with condition, F (2, 43) = 8.61, p =.005. There was no effect of time of day for the depth pole condition; however, turtles fed more from the floating ice block in the morning (M = 18.44, SEM = 1.84) compared to the afternoon (M = 5.97, SEM = 1.84; MD = 12.47, SEM = 2.61, p <.001, 95% CI: [7.21, 17.73]). See Figure 5. The full model had a corrected AIC of 297.94, which was identical to the model without the interaction term and improved over the null model (AIC = 328.29).

Adverse behaviors

The frequency of adverse behaviors did not differ significantly across conditions. The frequency of adverse events significantly differed across phases only for the depth pole condition, χ²(2, N = 72) = 9.60, p =.01. Pairwise comparisons of observed frequencies indicated that adverse interactions were more frequent during the post-feeding phase (n = 10) compared to earlier phases (Pre-feeding: n = 2; During: n = 3). See Figure 6. In contrast, there was no significant effect of phase for the baseline condition, or the floating ice block condition.

We examined how two feeding enrichment strategies influenced behaviors, space use, and agonistic interactions between two co-housed male green sea turtles relative to baseline feeding. We found that turtles adjusted their behavior around feeding times; however, the method of delivery of food did not exert strong effects on their behavior. Rather than concluding that the enrichment was ineffective for these two turtles, the lack of effect indicates that providing the turtles with enrichment garnered no additional benefits above and beyond the standard method of feeding and that there were no ill effects resulting from the enrichment strategies. Although, the fact that all three feeding conditions resulted in appropriate changes in the turtles’ behaviors is a positive sign, it is important to recognize that these results may not generalize to other members of this species. Here, post-feeding periods were characterized by an increase in resting behavior and time spent at the bottom of the tank. The turtles exhibited an adaptive progression from anticipation to engagement to recovery surrounding feeding times. This pattern is consistent with energy and digestion cycles and habits typically seen in other captive and wild animals (Wetherbee et al., 2004). The significant increase seen in resting post-feeding sessions further indicates that feeding events structure activity levels.

The turtles engaged in more pacing preceding, compared to during and following feeding, suggesting that the feeding conditions were effective in redirecting the turtle’s behavior. This likely shows anticipatory behavior, commonly observed in captive animals in response to predictable feeding schedules (Anderson et al., 2020; Waitt and Buchanan-Smith, 2001). Pacing is sometimes interpreted as a stress-related or stereotypic behavior (Burgener et al., 2008); however, it has also been viewed as an expression of more positive emotional states reflecting anticipation of imminent food provision (e.g., Rolls, 2005). It is unlikely that these turtles experienced frustration (Anderson et al., 2020) given that food was regularly delivered when anticipated. Nonetheless, it may be additionally beneficial to introduce less predictable feeding times. Both turtles also engaged in feeding more often in the morning, which may suggest diurnal rhythm and preferences (Ballorain et al., 2013).

It is likely that the variety in feeding methods kept things more interesting for the turtles than if food was delivered in the same manner every day. We observed that the turtles fed more from the floating ice blocks during morning compared to afternoon sessions, suggesting that turtles may not benefit equally from different types of enrichment at different times of day. This observation may be helpful when structuring feeding activities to maximize their benefits.

Whereas increases in foraging are sometimes linked to reduced aggression (da Silva et al., 2020), importantly, the frequency of aggressive interactions here was overall very low, indicating a lack of need for mitigation. The relatively low frequency of aggressive interactions suggests that current husbandry practices are largely effective in maintaining stable cohabitation between these two male turtles. However, these adverse behaviors occurred most frequently following the depth pole enrichment feeding condition, compared to during or preceding it. A similar effect was not observed after baseline or floating ice block feeding, perhaps because these latter feeding conditions allowed the turtles to avoid competition by providing food in more than one location (Costa et al., 2023; Escobedo-Bonilla et al., 2022). Specifically, during target feeding, the turtles were fed at separate locations, and during the floating ice block, two ice blocks were provided, allowing the turtles to separate from each other. During the depth pole condition, there was a single depth pole provided for both turtles to interact with. Although the depth pole condition, which encouraged the turtles to “chase” their food, may have increased their movement, making it more likely that they would interact with each other, this explanation for the increased adverse behaviors is less likely because we did not observe effects of condition for swimming, pacing, or feeding, which one would expect if this condition simply increased activity levels. It may be useful to consider enrichments that ensure that food is offered in different locations; for example, using one depth pole per animal, to reduce conflict further. In addition, both feeding enrichments lasted for a relatively brief period of time (estimated 2–5 minutes) before the food was consumed. Given that the floating ice block enrichment lasted, on average, longer than the depth pole enrichment, this could be another reason for the increase in adverse behaviors following the depth pole condition relative to the other conditions. Longer lasting enrichments might be even more beneficial.

A possible positive impact of the enrichment feeding conditions was that the turtles were less likely to be out of sight, or not visible to guests, following the two enrichment feeding conditions compared to baseline feeding. This finding suggests that the enrichment feeding activities are engaging the turtles, making them more active and visible to aquarium visitors, relative to baseline feeding. As enrichment activities are often designed to increase use of space and ensure the visibility of captive animals (Puehringer-Sturmayr et al., 2023), the enrichments used here appear to be successful in this regard.

Clear individual differences in behavior and space use were also observed between the two turtles. Carr spent more time at the bottom of the tank and more frequently engaged in resting behavior, compared to Benson. However, both turtles spent more time at the top of the tank compared to the bottom. The tendency to be found in the upper portion of the water column may be in part due to the fact that both subjects experience positive-buoyancy syndrome, making it difficult for them to dive and forage (Campbell, 1996; Schmitt et al., 2013). These turtles were treated using weights affixed to their shells, which did allow them to reach the bottom. However, it is still possible that their use of space may have been impacted by this condition.

It is important to note some limitations with the current study, namely its sample size of two individuals, which limits generalizability. Additionally, the length of this study was relatively short, stretching over a three-month period. Following the turtles for a longer period of time could generate further meaningful results. In addition, observation periods for this study were isolated to pre-scheduled feeding times. It is possible that feeding in itself is sufficiently enriching for the turtles, and that other manipulations outside of feeding times may exert more pronounced effects on their behavior. More comprehensive studies could examine the impact of enrichment outside of these specific times. Future research should also explore the relationship between resting site availability and aggression more directly, and examine behavioral changes in response to habitat complexity, such as evaluating impacts of adding additional resting zone spaces. Unfortunately, we did not record the space in which aggressive behaviors occurred due to a technical failure. Future studies should ensure that space use is recorded for all occurrence behaviors as well as scan sampling intervals.

These findings provide insight into the mechanisms influencing behavior of captive male sea turtles, and highlight the importance of spatial resource use and careful enrichment structure in managing social interactions. This study demonstrates that feeding phase appears to be the primary driver of behavioral change in these captive male green sea turtles, creating a cycle of anticipation, activity, and rest. More importantly, the results suggest that, under appropriate housing conditions, which include effective feeding strategies, sufficient space, and environmental complexity, co-housing of adult male green sea turtles can be feasible with relatively low levels of conflict. As the male sea turtles in this study are non-releasable, monitoring their welfare long-term is essential. Furthermore, continuous research on captive conditions is necessary to evaluate current practices and investigate means for improvement. This research aims to support aquariums in optimizing welfare outcomes by identifying practical, ethical, and cost-effective strategies such as enrichment implementation, feeding modifications, and habitat design improvements. Ultimately, enhancing captive conditions not only benefits the animals themselves but also supports conservation, rehabilitation efforts, and public education initiatives. These findings contribute to the limited literature on male sea turtle cohabitation. This work could inform welfare-based husbandry approaches in public aquariums, improving the lives of animals under human care through evidence-based research.