Directly observing organisms in wild settings is fundamental to behavioral research (Schofield et al. 2006). However, marine vertebrate ethology, particularly at depths exceeding 15 m, is plagued by a host of logistical challenges (Hooker and Baird 2001). As a result, marine turtle researchers have relied on various remote sensing technologies such as time-depth recorders, animal-borne cameras, ultrasonic tracking devices, multisensor archival tags, and even mandibular movement sensors to probe in situ habitat use and behavior (Hochscheid et al. 2005; Houghton et al. 2008; Wilson et al. 2008; Blumenthal et al. 2009; Fuller et al. 2009; Okuyama et al. 2013). Still, direct observation remains the only way to ground-truth these remotely sensed data and simultaneously gain access to the details of in situ behavior and real-time ecological interaction.

Hawksbill turtles (Eretmochelys imbricata) are globally distributed sea turtles that maintain significant populations in the Caribbean Sea, the Gulf of Mexico, Florida, and Bermuda (Witzell 1983; Lund 1985; Meylan and Redlow 2006). They are listed as Endangered under the US Endangered Species Act, and Critically Endangered worldwide by the International Union for Conservation of Nature (Baillie and Groombridge 1996). Though a number of studies examining movements, dive profiles, diet selection, and in-water behavior have improved our understanding of Caribbean hawksbill ecology (León and Bjorndal 2002; Cuevas et al. 2007; Vélez-Zuazo et al. 2008; Blumenthal et al. 2009; Rincón-Díaz et al. 2011b; Stringell et al. 2016), and several others have assigned broad behavioral categories such as swimming, feeding, resting, and surfacing to estimate in situ activity budgets (van Dam and Diez 1997; Houghton et al. 2003; Schofield et. al. 2006; Dunbar et al. 2008; Blumenthal et al. 2009; Proietti et al. 2012), none have focused on the process of prey acquisition itself, which for hawksbills appears to be focused on a fairly narrow range of marine sponges, that is, “spongivory” (León and Bjorndal 2002; Rincón-Díaz et al. 2011a; Stringell et al. 2016).

Spongivory is highly unusual among vertebrates (Burns and Ilan 2003), and although hundreds of sponge species are found in the Caribbean region, relatively few are known to be consumed by hawksbills (Meylan 1988; van Dam and Diez 1997; León and Bjorndal 2002; Rincón-Díaz et al. 2011a). Although empirical techniques to study diet (e.g., stomach lavage and fecal analysis) can reveal prey choice, they do not necessarily show the entire diet, nor do they address search behavior or effort, prey recognition, capture rate, or incidental effects on the surrounding community. Because poriferans can be dangerously well protected by both chemical and physical defenses (Sara and Vacelet 1973), hawksbills may have developed important behavioral strategies that ensure the intake of only the most desirable species.

In order to establish direct links between behaviors that appear to be related to foraging and the actual act of prey consumption, the relative frequency of each behavior leading to the focal act of ingestion can be measured and scored (Bakeman and Gottman 1997; Nowacek 2002). In a consistent step-wise progression, each behavior leading up to a focal act should become less frequent and more dependent on a specific prior behavior (Nowacek 2002). In other words, if a given behavior in a repeated sequence always leads to the focal act, its relative “importance” in the process is assumed to be high, and vice versa. A generalist, for example, might employ a relatively short sequence of investigative behaviors with a high rate of acceptance (intake), while a specialist might instead employ a longer sequence of behaviors with a higher rate of rejection per area or item explored.

No other routine daily behavior involves more direct physical interaction with the microenvironment than foraging, which in the case of hawksbill turtles, may be playing an important role in reef ecology (León and Bjorndal 2002). In this study, an ethogram was developed from video recordings of wild hawksbill turtles foraging off the east coast of South Florida to generate the most systematic first-person characterization of hawksbill feeding behavior to date. Our findings reveal novel insights concerning search behavior, prey choice, and habitat use, which can fill important gaps in our overall understanding of this species and help strengthen and focus conservation and management policies (Blumenthal et al. 2009).

METHODS

Handheld digital cameras were used to film hawksbill turtles on the 15–25-m-deep reefs of central and northern Palm Beach County (Florida) during daytime scuba dives. The turtles in this area have become accustomed to divers, and appear largely unaffected by their presence. Although various makes and models were used, the majority of the footage was recorded with a Canon Powershot SD960 IS Digital ELPH camera inside an Ikelite, Inc. (Indianapolis IN) underwater housing. No lights were used during the filming. When foraging behavior was observed, the photographer positioned the camera as close to the point of prey consumption as possible, often within 0.5 m of the turtle's head. Tag numbers, if present, were recorded. Filming sequences were terminated when either the observer was required to continue or terminate his or her dive owing to safety practices and air consumption, or when the turtle actively swam off.

Behaviors associated with foraging were grouped into categories as follows (Table 1): Scan: any combination of slow swimming or “walking” (flippers in contact with substrate); side-to-side head movement close to the substrate; rapid ocular movement; generally continuous forward momentum. Target: forward motion stopped; head often sharply angled down when on horizontal surfaces; rhamphotheca pointed directly at object. Nudge: rhamphothecal (closed-mouth) contact with item or substrate. Bite: open mouth contact with item, not necessarily resulting in detachment of item. Chew: detachment of item, followed by at least 2 mastications. Swallow: item ingested; ventral throat contraction observed.

Table 1. Hawksbill foraging ethogram.

View Fullscreen

Videos were viewed frame by frame using Quicktime Player (Apple), from which behavioral transitions were scored on a transition matrix (Table 2). Each transition from one of the above behaviors to the next was tallied in the corresponding box in the matrix; for example, if the subject transitioned from Bite to Chew, a mark was placed in the box where the Bite row intersects the Chew column. The transitions between each of these paired behaviors in a sequence are known as “lags” (Bakeman and Gottman 1997). The data from the videos were combined into one cumulative matrix. The relative frequency of each behavior prior to the focal act of swallowing the food were calculated by dividing the total number of times a given behavior was observed in proportion to the other behaviors in the sequence, and by the total number of observations (Bakeman and Gottman 1997). In addition, readily observable biotic and abiotic features of the turtles' surroundings (e.g., habitat type, current direction, presence or absence of flora and fauna), the morphological characteristics of targeted prey items, and various behavioral subtleties (e.g., buoyancy control, habitat manipulation) not addressed within transition matrix were hand-noted by time code for each video sequence.

Table 2. Cumulative transition matrix. Numbers in cells represent the number of times a behavior on the y axis transitioned to a behavior on the x axis, followed by the proportion of each behavior relative to the sum (Σ) of that row in parentheses. For example, Scan transitioned to Target 247 of 259 times, resulting in a frequency of 0.95, or 95% of the time.

View Fullscreen

RESULTS

Thirty videos totaling 141 min (range, 0.5–21.7; x̄ = 4.7) were evaluated, each containing footage for 1 turtle. A total of 1242 behavioral transitions were recorded (Table 2). Sequential transition frequencies are shown in Table 2 and Fig. 1. Aside from the onset of the feeding sequence where Target is the primary activity that follows Scan (95%), no other consistent step-wise sequence leading to prey consumption was found. Target led to any of 3 options: returning to scanning (24%), investigating further with nudging (27%), or transitioning straight to a bite (47%). Nudging occasionally led to a bite (36%), but usually transitioned back to scanning (60%), and bites only resulted in chewing about half (51%) of the time. Once the item was separated from the substrate and in the turtles' mouth, chewing resulted in ingestion a slight majority of the time (61%), but otherwise mostly transitioned back to targeting (27%). After swallowing the item, the turtles usually transitioned back to targeting (73%), or moved on to continue scanning (26%). Overall, there was close to a 50/50 chance that an item was rejected at each stage leading to the focal act of swallowing; that is, about half (53%) of the targets received a bite, about half of those bites (51%) provided something to chew, and only a little more than half (61%) of those items were swallowed (Table 2).

View Figure

• Download as Powerpoint
• Download Figure

Taken cumulatively, the data show a gradual decrease in the frequency of behaviors as the sequence progressed toward prey consumption, with the exception of nudging behavior (Fig. 2). Scan data are not included in the figure because Target is the behavior that initiates the feeding sequence. Targeting, biting, and chewing represented 33%, 21%, and 16% of the total number of observations respectively. Swallowing (n = 78), which is the focal act of the sequence, occurred 81% fewer times than targeting (n = 410) and represented only 6% of all behaviors recorded collectively. When plotted against video duration, the number of targeting behaviors increased at a near linear rate (R² = 0.772). In contrast, the number of items swallowed per video duration increased only slightly (R² = 0.201) (Fig. 3).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Aside from surfacing ascents, the hawksbills did not appear to exploit any part of the water column itself for feeding. The turtles maintained neutral to slightly positive buoyancy at depth, and often used the tips of all 4 flippers to gently push off of the substrate in an alternating walking gait when moving slowly over the reef. In exposed areas such as the reef top and slope, there was a preference (75% of the foraging sequences) for facing into the prevailing current while searching for prey. Spanish hogfish (Bodianus rufus), queen angelfish (Holocanthus ciliaris), and/or juvenile wrasses of the family Labridae were observed cofeeding with hawksbills in 7 of 30 videos.

When engaged in searching behavior, the turtles typically kept their heads angled down with the eyes focused forward, mouth slightly opening and closing, and the rhamphotheca within inches of, and often making contact with, the substrate (a.k.a. “nudging”). Considerable effort went into searching for concealed or semiconcealed prey, which required relatively slow forward movement and the close inspection of holes, small crevices and depressions, and small patches of accumulated sand. In patches dominated by a high density of tall (≥ 0.5 m) octocorals (e.g., sea rods and whips, sea fans), the turtles remained close to the substrate by pushing through and over the dense faunal cover. The turtles often paused to inspect the bases of vase (Callispongia sp.) and barrel (Xestospongia sp.) sponges with nudges and exploratory bites. Open-mouthed scraping of rock surfaces was common. In some cases, the turtles used their mouths to deliberately grasp and overturn rocks to gain access to their underside and/or the substrate beneath. Small patches of the reef with thin (approximately 2–6 cm) layers of sand were closely examined, and often bites reaching several centimeters deep into the substrate resulted in mouthfuls of well-concealed prey, likely Spheciospongia vesparium. Pushing upward away from the substrate with the front flippers was often required to separate mouthfuls of leathery-textured sponges. Mouthfuls were thoroughly masticated using multidirectional lower mandible movements during which sponge fragments and biotic and abiotic debris (e.g., sand, pebbles, algae, hydroids) were separated and rejected. It was common for the turtles to take several bites into the prey's newly exposed inner tissue before moving on, and though some were fragmented or displaced, no one individual sponge was fully consumed. On reef ledges and artificial reefs (e.g., scuttled vessels, concrete debris), turtles were seen actively seeking prey and scraping small bites from vertical and underside surfaces.

The difficulty in identifying the prey items from the videos precluded any realistic estimates of diet selection. However, all targets were sessile invertebrates, and only poriferans were observed being eaten. The only prey species that could be identified to the species level with confidence were the chicken-liver sponge Chondrilla caribensis (Demospongiae: Chondrosida: Chondrillidae), encrusting forms of the loggerhead sponge Spheciospongia vesparium (Demospongiae: Hadromeridae: Clionaidae), and the giant barrel sponge Xestospongia muta (Demospongiae: Haplosclerida: Petrosiidae).

DISCUSSION

During foraging bouts, the close examination of the substrate and of each item prior to ingestion suggests that the turtles were focused on finding specific prey items, many of which were fully or partially concealed by sand, epibiota, rocks, or crevices. Similar slow searching and probing behaviors were reported in juvenile hawksbills in Mona Island, Puerto Rico, and in Brazil (van Dam and Diez 1997; Proietti et al. 2012). As opposed to what would be expected of a nondiscriminate feeder, only a small proportion of the items that were initially investigated by the turtles were actually consumed, suggesting that, through multiple stages of scrutiny, the turtles actively sought and preferentially accepted a narrow range of prey items among the suite of potential alternatives that may also have been present. The relative abundance of preferred prey items does not necessarily reflect the distribution, abundance, or diet composition of Caribbean hawksbills (León and Bjorndal 2002; Rincón-Díaz et al. 2011a, 2011b), and dietary plasticity may be important to juvenile hawksbills as they move among habitat types during their development (Rincón-Díaz et al. 2011b). Persistently exploring and testing an array of potential food items could help young individuals establish a hierarchy of dietary alternatives or supplements which, under changing environmental or physiological circumstances, might necessarily gain or lose prominence in their individual diets over time.

Ectosomal (outer) and choanosomal (inner) tissue layers in known hawksbill prey sponges Anthosigmella varians and Geodia sp. are differentiated, with the strongest chemical and structural defenses allocated to the ectosomal parts of the sponge, as those are most likely to be encountered by predators (Hill 1999; Hill and Hill 2002; Rohde and Schupp 2012). Indeed, the hawksbills observed in this study did appear to seek exposed choanosome of target species, including the giant barrel sponge X. muta, and were frequently seen feeding on irregularly shaped, encrusting, concealed forms of S. vesparium. Recent satellite tracking data from hawksbills in the area reveal small (< 1 km²) overlapping home ranges (Wood et al. 2017), where certain preferred specimens could be reencountered fairly frequently.

Marine turtles are equipped with complex visual systems that include the capacity to discriminate colors (Levenson et al. 2004; Eckert et al. 2006; Young et al. 2012), as well as olfactory systems to perceive chemical cues (Manton et al. 1972; Grassman and Owens 1982; Endres and Lohman 2012). Although our observations strongly imply that hawksbills were visualizing their surroundings at multiple scales, they also suggest that olfaction was playing a prominent role in foraging. Most striking was the turtles' ability to locate concealed prey. Several sponge taxa are known to produce odors (Pawlik et al. 2002), making olfaction a potentially effective tool for locating concealed items among the benthic community. While the turtles were closely inspecting the substrate (frequently facing into the prevailing current), they continuously performed small jaw movements known as buccal oscillations (Walker 1959). Common in reptiles and amphibians, variations on these movements are known to increase nasal ventilation without involving the lungs, and likely aid in olfaction (Jorgensen 2000). Hochscheid et al. (2005) found similar jaw movements ceased during rest and amplified just prior to feeding in captive loggerhead turtles (Caretta caretta), and suggested they function in chemoreception during foraging. The frequent nudging behavior observed in the hawksbills may further facilitate chemoreception, and is similar to the “touch with muzzle” behavior observed in the four-eyed turtle (Sacalia quadriocellata) (Liu et al. 2009), “nose explore” in the desert tortoise (Gopherus agassizii) (Ruby and Niblick 1994), and “nose” behavior in various emydid turtles (Davis 2009). Additionally, applying pressure with the rhamphotheca may serve an additional tactile function for hawksbills by providing information on the consistency or compaction (a.k.a. “hardness”) of an item of interest.

Close scrutiny of the substrate while searching for food is by nature time-consuming, and the small differences in food ingestion rates observed between short and long video sequences suggest that the turtles probably forage fairly consistently throughout the day to acquire sufficient caloric intake. The long-term use of small areas inevitably results in a familiarity with the terrain, and it is likely that the turtles become aware of the location of the most productive patches among and within habitat types. In Palm Beach, Florida, it is apparent that turtles find sufficient populations of suitable prey species within small, repeatedly patrolled areas, but invest a considerable amount of time and energy to locate and positively identify those they deem most desirable.

Detailed in-water behavioral observations hold considerable potential to better understand the basic survival strategies of sea turtles and the species with which they interact (Houghton et al. 2003; Schofield et al. 2006, 2007; Blumenthal et al. 2009). Where possible, researchers should take opportunities to collect photos and video footage of turtles in their natural environment, which are becoming easier to obtain as small handheld underwater cameras become increasingly available. These data will broaden our understanding of habitat preferences, diet, and social behavior among sea turtles, and in the case of hawksbills, help scientists examine the effects of spongivory on coral reef community structure and the adaptations required to maintain this unusual trophic interaction.