Feeding ecology of the hawksbill turtle (Eretmochelys imbricata)


Material Information

Feeding ecology of the hawksbill turtle (Eretmochelys imbricata) spongivory as a feeding niche in the coral reef community
Physical Description:
vi, 118 leaves : ill., map ; 28 cm.
Meylan, Anne Barkau
Publication Date:


Subjects / Keywords:
Sea turtles   ( lcsh )
Sea turtles -- Caribbean Area   ( lcsh )
Sea turtles -- Ecology   ( lcsh )
Turtles -- Ecology   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1984.
Includes bibliographical references (leaves 107-117).
Statement of Responsibility:
by Anne Barkau Meylan.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000491140
notis - ACQ9642
oclc - 11941200
System ID:

Full Text








I would like to thank the members of my supervisory committee, Dr.

Archie Carr (chairman), Dr. Carmine Lanciani, and Dr. John J. Ewel, for

their guidance and encouragement throughout the study. Discussions and

correspondence with K. Ruetzler, R. Garrone, S. Bloom, S. Pomponi, H.

Reiswig, and G. Schmahl were particularly helpful in developing my

interest and background in sponge biology. I would like to thank J.

Ottenwalder, C. Sanlley, S. Inchaustegui, and N. Garcia for arranging

the collection of digestive tract samples in the Dominican Republic, and

Dr. M. Goodwin for those from Carriacou, Grenada. P. Jesse aided me

greatly in obtaining samples at Bocas del Toro, Panama. A. Ruiz, L.

Richardson, R. Witham, J. Fletemeyer, and N. Rouse also kindly provided

me with material for study. Field work in Panama was facilitated by the

logistic support provided by Y. Hidalgo, M. Panezo, and A. Ayala of

Recursos Naturales Renovables. I thank D. Galloway, A. Ruiz, and P.

Meylan for their assistance in the field in Panama. Specimens were

brought into the United States under U.S. Fish and Wildlife Service

Permit PRT 2-4481. I am grateful to Dr. S. Pomponi, B. Causey, and C.

Curtis for their help in collecting sponges in the Florida Keys. Col-

lections made in Looe Key National Marine Sanctuary were carried out

under National Marine Sanctuary Permit KLNMS and LKNMS-04-83. I would

like to give special thanks to Dr. K. Ruetzler, of the U.S. National

Museum, for his assistance in the identification of sponges. I am also

grateful to S. Blair, Dr.N. Eiseman, and P. Hall, of Harbor Branch

Foundation, for algae identifications; K. Auffenberg and D. Robinson,

Florida State Museum, for mollusk identifications; Dr. F. Maturo,

Department of Zoology, for identification of bryozoans and other inver-

tebrates; and G. Burgess, Florida State Museum, for identifying fish

eggs. Dr. E. Jacobsen provided technical assistance in histological

work. I thank Dr. J. Fiskell for the use of the IFAS Forest Soils

Laboratory to carry out nitrogen determinations. I thank M. McLeod for

teaching me the procedures. I am grateful to Dr. K. Bjorndal for many

helpful consultations concerning laboratory procedures for nutritional

analyses. Dr. J. Ewel, Dr. F. Putz, Dr. J. Anderson, Dr. F. Maturo, and

Dr. F. Nordlie kindly loaned me equipment. I thank Dr. L. Berner for

his advice on slide preparation and microphotography. I thank Dr. J.

Mortimer and Dr. W. Rainey for allowing me to use their unpublished data

on hawksbills. H. Kochman provided expert advice on formatting my data

for computer analyses. I thank my husband Peter for support of many

kinds throughout the study, including the preparation of the figures. I

am grateful to C. Barkau and G. Russell for their support and encourage-

ment. I thank Adele Koehler for typing the final version of the disser-

tation. Financial support for the project was provided primarily by the

World Wildlife Fund/International Union for the Conservation of Nature

and Natural Resources (Gland, Switzerland); supplemental funding was

provided by the U.S. National Marine Fisheries Service (Purchase Orders

03-78-D08-0025 and NA 80-GA-C-00011, A. Carr, Principal Investigator)

and the Caribbean Conservation Corporation. I thank the Department of

Zoology for its generous support of many kinds.



. ii

S. v


. 10




METHODS . . . .

Diet Analyses. . . .
Laboratory Analyses of Fresh Sponges . .

RESULTS . . . .

Composition of the Diet. . .
Structural Characteristics of Prey Sponges .
Toxicity and Antibiotic Activity of Prey Sponges .
Nutritional Characteristics of Prey Sponges. .

S. 25

DISCUSSION. . . ... .... 67

Composition of the Diet. . ... 67
Feeding Selectivity. . .... 77
Role of Feeding Deterrents . .... ... 86
Nutritional Characteristics of Prey Sponges. ... 100
Spongivory as a Feeding Niche. . 101

SUMMARY ... . . 105



Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy



Anne Barkau Meylan

December, 1984

Chairman: Archie Carr
Major Department: Zoology

The feeding ecology of the reef-dwelling hawksbill turtle was

investigated in Caribbean Panama, the Dominican Republic and five

countries of the Lesser Antilles. The high percentage of sponges in

digestive tract contents (x = 94.2% of dry weight) and the high degree

of homogeneity among samples from turtles of different sexes, sizes

(over 23 cm carapace length), and geographic origins provide strong

evidence that the species is a strict spongivore. Widespread occurrence

of spongivory in Eretmochelys is proposed.

The presence of pelagic species of the alga Sargassum, pelagic fish

eggs, and other flotsam in digestive tract contents of hawksbills

smaller than 23 cm carapace length provides evidence linking post-

hatchlings to the pelagic Sargassum raft community.

Twenty-three species (14 genera) of demosponges, all representa-

tives of the tetractinomorph orders Hadromerida, Astrophorida and Spiro-

phorida, account for 98.8% (dry weight) of all identified sponges.

Comparison of the sample distribution with the composition of well-

studied Caribbean sponge faunas indicates that the diet is narrowly

restricted. Four major orders of sponges with reef-dwelling representa-

tives are poorly, if at all, represented. Ten species account for 87.4%

(dry weight) of all identified sponges.

Prey sponges are characterized in terms of structural and bio-

chemical properties. The effectiveness of assumed defensive mechanisms

of sponges is evaluated. Spongin fibers are absent in prey sponges,

providing circumstantial evidence that they serve as a feeding deter-

rent. Prey sponges are rich in collagen fibrils; carbohydrate-rich

compounds associated with the fibrils may impart nutritional value.

Silica content varies widely among prey sponges (0-51.6%), suggesting

that siliceous spicules do not deter predation by hawksbills. Astro-

phorid sponges are among the most highly silicified demosponges. Samples

of intestinal contents consisted of up to 92% ash, which was largely

silica. Scanning electron micrographs of the intestinal epithelia show

numerous embedded spicules. Organic content, energy content, and

nitrogen content are determined for representative prey sponges.


The hawksbill turtle (Eretmochelys imbricata), one of seven species

of marine turtles, occurs in tropical and subtropical waters of the

Atlantic, Pacific and Indian oceans. It is widely distributed in the

Caribbean and western Atlantic, normally ranging from southern Florida

southward along the Central American mainland to Brazil, and throughout

the Bahamas and the Greater and Lesser Antilles. Two subspecies (E. i.

imbricata in the Atlantic Ocean and E. i. squamata in the Indo-Pacific)

have been described (Carr, 1952), on the basis of differences in colora-

tion and carapace shape. The criteria have proven to be unreliable in

distinguishing the two forms, however, and subspecific designations are

rarely used.

The affinities of Eretmochelys with other sea turtle genera are not

well established. Osteological evidence (Carr, 1942) and serum protein

analysis (Frair, 1979) suggest closer affinities with the loggerhead

(Caretta) and ridley (Lepidochelys), than with the green turtle

(Chelonia). On the basis of immunological distance, the genus.

Eretmochelys is estimated to have diverged from other turtles 29 million

years ago, in the Oligocene (Chen et al., 1980). Zangeri (1980) dates

the divergence time of the line leading to Eretmochelys as middle Mio-

cene, on the basis of morphological features.

The hawksbill is a small to medium-sized marine turtle; adult

females in the Caribbean range from 62.5-91.4 cm straight carapace

length. Nearly all published size data are for females, because of


limited access to males. The heaviest hawksbill ever recorded was a 127

kg individual caught at Grand Cayman, in the West Indies (Lewis, 1940).

Since 1970, the hawksbill has been listed as an endangered species

by the International Union for the Conservation of Nature and Natural

Resources (Honneger, 1970). International trade in tortoiseshell, the

translucent epidermal scutes of the carapace, is the single greatest

threat to the species (Groombridge, 1982). Throughout its circum-

tropical range, the hawksbill is also subject to intense exploitation

for meat and eggs. Immature animals are harvested in great numbers for

the taxidermy trade in the Far East. The diffuse distribution of the

species in both nesting and foraging habitats has impeded effective

conservation action.

Life history data on the hawksbill have been slow to accumulate,

partly because of the depleted status of populations throughout the

world, but also because of logistic difficulties inherent in the study

of highly mobile, marine animals. The tendency of hawksbills to nest

diffusely, rather than in large aggregations, has hindered the effec-

tiveness of land-based tagging programs, which, in the study of other

marine turtles, have been very useful. With few exceptions (Diamond,

1976; Hirth and Latif, 1980; Limpus, 1980; Limpus et al., 1983; Brooke

and Garnett, 1983) most data on the nesting biology of the hawksbill

have been collected incidental to investigations of other species.

Whether hawksbills undertake periodic migrations to distant nesting

beaches, as other sea turtles do, has not been determined. Tag re-

coveries indicate that some long-distance travel does occur (for review

see Meylan, 1982). Evidence to support the commonly held theory that

hawksbills nest on beaches adjacent to their feeding grounds is largely


Coral reefs are widely recognized as the resident foraging habitat

of the hawksbill (Babcock, 1937; Carr et al., 1966; Carr and Stancyk,

1975; Alcala, 1980; Nietschmann, 1981; Carr et al., 1982). Homing

records (Nietschmann, 1981) and sightings of tagged individuals (Alcala,

1980; Boulon, 1983) suggest a relatively parochial existence on the

reef. Other habitats--such as rocky outcrops and, along the Pacific

coast of Central and South America, mangrove-bordered bays and estu-

aries-are occupied to a limited extent when coral reefs are absent.

Despite the association of the hawksbill with the well-studied

coral reef community, the species' ecological niche has never been

investigated. The present study of feeding ecology was initiated as an

approach to filling this gap in knowledge. The feeding biology of the

hawksbill has received little previous scientific study. A considerable

number of anecdotal accounts exist in the literature, reporting the

stomach contents of single individuals (for review see Witzell, 1983).

Although they provide useful information, their qualitative nature makes

it difficult, if not impossible, to construct a profile of the diet.

The authors seldom give any quantitative information on the relative

importance of the various food categories. The accounts suggest wide

variety in the hawksbill's diet, and include such diverse food items as

mollusks, sponges, gorgonians, fish, seagrasses, crustaceans, sea

urchins, mangrove fruits and leaves, tunicates, jellyfish, algae and

cephalopods--to name only a few.

Current knowledge of the feeding habits of the hawksbill is based

largely on a study by Carr and Stancyk (1975). Theirs was one of the

few detailed studies of the hawksbill's diet and apparently the only

quantitative one. Stomach contents of 20 mature turtles caught off the

nesting beach at Tortuguero were examined. On the basis of frequency of

occurrence, sponges and tunicates were ranked as the most important

components of the diet. Small amounts of seagrass, algae, mollusks and

bottom material were also found. The authors concluded that "the

hawksbill is a relatively indiscriminate feeder whose food consists

mainly of benthic invertebrates" (p. 165).

Another study, which is useful because of its detail, was that of

Den Hartog (1980), who examined the contents of the entire digestive

tract of a single small hawksbill (36.2 cm carapace length) caught in

the Salvage Islands, eastern Atlantic. He identified two species of

sponges, the actinian Anemonia sulcata, at least two species of pelagic

coelenterates, fragments of marine algae, a spider crab, and some

gastropod mollusks. No attempt was made to quantify the various food

items and the total amount of food examined was not reported. Den

Hartog (1980) concluded from his analysis that the hawksbill was

essentially carnivorous but did not make any inferences about specific

food preferences.

The present study was influenced and, to a degree, channelized by

the discovery that the hawksbill feeds almost exclusively on sponges--at

least at 19 localities in the Caribbean where digestive tract samples

were obtained. This was an unexpected finding. Sponges were an impor-

tant component of diet samples examined by Carr and Stancyk (1975), but

they concluded that the hawksbill is an opportunistic omnivore, with a

preference for benthic invertebrates,and this view is widely accepted.

Spongivory is an unusual feeding niche, occupied by relatively few

animal groups the world over. The list of animals that occasionally

feed on sponges includes diverse phyla--mollusks, echinoderms, annelids,

nematodes, crustaceans and vertebrates (for review see Sara and Vacelet,

1973). Relatively few species, however, subsist primarily on sponges.

Sponge-feeding is particularly rare among vertebrates. De Laubenfels

(1950b) commented on the extreme paucity of sponge-feeding records for

reptiles, birds, and mammals. Numerous surveys of the feeding habits of

marine fish, some involving over 200 species, have revealed very few

true spongivores (Dawson, Aleem, and Halstead, 1955; Hiatt and

Strasburg, 1960; Randall, 1967; unpub. references in Bakus, 1969;

Vivien, 1973; Hobsen, 1974; and Green, 1977). Angelfishes belonging to

the genera Holacanthus and Pomacanthus are among the few exceptions.

They have been identified as spongivores at numerous localities, in-

cluding the West Indies (Randall, 1967; Randall and Hartman, 1968),

Guyana (Lowe, 1962), Veracruz, Mexico (Green, 1977), Hawaii (Hobsen,

1974), and Madagascar (Vivien, 1973). Other sponge-feeding fish include

certain species of filefishes (Monacanthidae), trunkfishes (Ostracion-

tidae), puffers (Tetraodontidae), and the moorish idol (Zanclidae).

Among invertebrates spongivory is somewhat more common--although by

no means widespread. Certain species of dorid nudibranchs are appar-

ently obligate spongivores. A number of sponge associates--e.g., poly-

chaetes, isopods, shrimp, etc.--consume sponge, but the extent to which

sponges contribute to their diet has not been determined. Asteroid

echinoderms are major predators of sponges at McMurdo Sound, Antarctica

(Dayton et al., 1974). Sponge predation by sea urchins is reviewed by

Lawrence (1975). The food chains in which the majority of sponge


predators are involved tend to be side chains, which do not lead to

higher trophic levels (Vacelet, 1979).

Spongivores tend to be highly specialized morphologically and, in

some cases, behaviorally. The highly evolved relationships of dorid

nudibranchs and their sponge prey are well known. Many nudibranchs form

species-specific feeding relationships with sponges. Some incorporate

secondary metabolites (including pigments) and spicules from their prey

and use them for their own defense. Spongivorous angelfishes (Chaeto-

dontidae), filefishes (Monacanthidae) and trunkfishes (Ostraciontidae)

are among the most advanced forms of modern teleosts (Randall and

Hartman, 1968).

The low level of predation on sponges is particularly remarkable

when one considers their great abundance and wide distribution. Sponges

are a quantitatively important component of hard-substrate marine com-

munities. On coral reefs, the contribution of sponges to reef biomass

frequently exceeds that of hermatypic corals (Ruetzler, 1978). In the

spur and groove zones and on the outer fore reef at Carrie Bow Cay,

Belize, the standing crop of siliceous sponges may be as high as 2 kg

wet weight per m2 suitable habitat (Ruetzler and Macintyre, 1978).

Sponge biomass on the solid exposed reef of the fore-reef slope platform

at Discovery Bay, Jamaica, attains an estimated volume density of 3 1

per m2, and exceeds the coral-zooxanthellae tissue biomass (Reiswig,

1973). De Laubenfels (1950b) listed 115 species of shallow-water

sponges in the West Indian region, excluding Bermuda. If utilizable,

sponges clearly represent an extensive food resource.

The relative immunity of sponges to predation has been attributed

by many authors to the defensive protection provided by siliceous

spicules, tough organic fibers, and toxic or noxious chemicals (Hyman,

1940; Bakus, 1964, 1969, 1981; Randall, 1967; Randall and Hartman,

1968; Sara and Vacelet, 1973; Levi, 1973; Jackson, 1977; Bergquist,

1978; Vacelet, 1979; Bakus and Thun, 1979; and Hartman, 1981). Spicules

and fibers are considered to serve as mechanical deterrents to ingestion

and/or digestion, whereas chemical compounds, which may be emitted into

the surrounding sea water, presumably repel predators from a distance.

Not all authors agree on the utility of these mechanisms. The

defensive role of spicules and spongin is perhaps the most debated, some

authors (Bergquist, 1978) arguing that the functions of these elements

are strictly structural. Defensive utility is nevertheless suggested by

some field data. Pawlik (1983) reported that the sponge-feeding

polychaete Branchiosyllis oculata consumes only the soft parts of its

siliceous prey. Long, protruding spicules of Cinachyra antarctica may

serve to prevent nudibranch and asteroid predators from reaching the

sponge surface (Dayton et al., 1974). Other evidence of a defensive

utility of spicules is the presence of morphological adaptations in

predators, such as spicule-compacting organs in sponge-feeding dorid

nudibranchs (Forrest, 1953; Bloom, 1976, 1981); and by physiological

adaptations such as copious mucus production by the digestive tract of

spongivorous nudibranchs (Forrest, 1953; Fournier, 1969) and fish

(Randall, 1963).

The defensive utility of secondary metabolites in sponges is almost

universally accepted. Certain classes of compounds found in sponges,

particularly terpenoids, are widely recognized as predator deterrents in

other contexts in both marine and terrestrial systems (Harborne, 1977;

Norris and Fenical, 1982).

Despite considerable discussion of the role of chemical and

mechanical feeding deterrents in sponges in the literature, evidence

from field data is relatively limited. This can be attributed mainly to

the fact that few investigators other than sponge taxonomists have

undertaken field studies of sponges, because of the difficulty of iden-

tifying them. Randall and Hartman (1968) examined the diets of 11 West

Indian fish for which sponges constituted 6% or more of stomach con-

tents. In an effort to discern patterns, prey sponges were described in

terms of ash content, fiber content, color, and growth form. Dayton et

al. (1974) studied the effects of asteroid and nudibranch predators on

sponges at McMurdo Sound, Antarctica. Although the latter study was

primarily concerned with the ecologic effects of sponge predation,

useful descriptive information was obtained on the diets of the

asteroids and nudibranchs, and on the physical characteristics of prey

sponges. Green (1977), Bakus and Thun (1979), and Bakus (1981) investi-

gated the toxicity of marine sponges to fish.

My study of the feeding ecology of the hawksbill revealed both

heavy dependence on sponges and unexpected selectivity in the sponges

eaten. Because the literature so strongly implicated structural and

chemical deterrents in limiting spongivory, I decided to test whether

patterns in the hawksbill's diet could be explained on this basis. The

feeding deterrents that have been proposed for sponges are not uniformly

represented among the various taxa. Thus, my hypothesis was that

effective deterrents would be revealed by avoidance or limited

consumption in the dietary patterns, or by physiological or

morphological adaptation. My study of spongivory in the hawksbill had

two goals: 1) to try to explain how the species has been able to take


advantage of this rarely used, but potentially vast, feeding oppor-

tunity; and 2) to gain an understanding of spongivory as a feeding niche

in the coral reef community.

It seems probable that one of the reasons spongivory has received

little previous attention is the difficulty involved in identifying

sponges from the digestive tracts of the few known spongivores. Dorid

nudibranchs rasp their prey with radulae, and investigators are forced

to identify prey sponges from dissociated spicules found in fecal

pellets. Sponges are extremely difficult to identify even when whole,

and it is not surprising that quantitative analysis of the sponge diets

of nudibranchs has been limited. The small size of fragments is also a

problem in the case of sponge-feeding fish.

The hawksbill presents several advantages for a study of spong-

ivory. Bite-size is large, compared to that of other spongivores, and

food is not masticated. Like most turtles, hawksbills shear and gulp

their food, so relatively large, intact pieces of sponge are found in

the stomach and intestinal tract. The large amount of food in the

digestive tract provides a good sample for quantitative analysis. In

the present study, it was a further advantage that relatively few taxa

of sponges (22 genera) were represented, and the sponges were nearly all

siliceous species, generic identification of which is based solely on

spicule complement. For sponges in which spicule placement within the

tissue or overall morphology are necessary for diagnosis, study of

sponge-feeding patterns would be far more difficult.


Diet Analyses

Collection of Samples

Food samples were obtained from 68 hawksbills. The origins of the

turtles are given in Table 1 (see also Fig. 1). Sixty-one were captured

in Caribbean waters by subsistence fishermen using nets, spearguns or

harpoons, or were taken on nesting beaches. Three food samples (one

fecal pellet, two buccal cavity samples) were from live, wild turtles.

Four small turtles (14.0-21.3 cm straight carapace length) were obtained

through a government stranding network, after they had washed up dead or

moribund on Florida beaches. Data for these four are reported

separately because of the possibility that food items in the digestive

tract were not representative of the normal diet. Further justification

for considering these turtles separately is the likelihood that they

represent a distinct ontogenetic life history stage, with pelagic,

rather than benthic, feeding habits.

Samples included in quantitative analyses consisted of the

following: stomach and intestinal contents (37 turtles); stomach

contents only (17 turtles); stomach and partial intestinal contents (2

turtles); partial intestinal contents (4 turtles); and unknown site of

origin (1 turtle). Only one stomach was found to be empty; it was

included in calculations of percentage occurrence and average percentage



Table 1. Geographic origin of hawksbill turtles (Eretmochelys imbricata)
included in the feeding study.




Dominican Republic



Netherlands Antilles
(St. Martin)


Turks/Caicos Islands

United States


No. of Localities











No. of Hawksbills


















-H r4

O S-d
f> -

*: *







contribution. All intestines contained some food. Esophagus contents

were not included in the study because only rarely were they fully

recovered. No attempt was made to quantify the digestive tract contents

of the four stranded turtles because of the small amounts of food

present and the high percentage of unidentified material. The two buccal

cavity samples and the fecal pellet were likewise not quantified.

Because of regional differences in fishing techniques and customs,

a well-balanced size series was not obtained for each geographic area.

Large turtles are the principal target of the net fishery in Bocas del

Toro, Panama, the origin of the largest group of samples. Small turtles

captured there are usually released unharmed. In the West Indies, small

turtles are the usual quarry, traditional net fishing having been

replaced at most localities by fishing with spear guns.

Figure 2 shows the size distribution of the turtles included in the

study. Sizes are reported as straight carapace lengths. When only

curved carapace measurements were taken, they were converted to straight

lengths using a regression equation. Missing size data were calculated

for three turtles from a regression of head width against carapace

length, and for six turtles from a regression of intestinal tract length

against carapace length. Although no size measurements are available

for 18 turtles, all but two could be assigned to either adult or non-

adult age categories. The size at which hawksbills attain sexual

maturity is not firmly established. Nietschmann (1981) recorded an

adult female only 62.5 cm in carapace length from the Caribbean coast of

Nicaragua. At Tortuguero, Costa Rica, the smallest female that has been

observed on the nesting beach was 72.4 cm in carapace length (Carr,

unpubl. data). In the present study, turtles of both sexes over 70 cm


J 4.J

0 0


d 0



-- Q

^ '
J *

0) U
Qh -^

1-1 a a
tn *^







63lianI dO d8V3JlNn

S3ltnl rO 13an


carapace length were considered adults. A male 66.7 cm in length and a

female 68 cm were deleted from age-specific analyses.

Samples that have been quantitatively analyzed include 17 males and

22 females. The remainder are unsexed. Sex was determined by gross or

histological examination of the gonads. Tail dimorphism was not a

useful indicator of sex except in large turtles. The smallest male

turtle in which elongation and thickening of the tail was noted was 74

cm; a male 52 cm in carapace length could not be sexed by external

characters. Data on the reproductive condition of females were gathered

whenever possible.

Sample Treatment

Digestive tract contents were initially preserved in 1 part 37%

formaldehyde:19 parts sea water. Sponges and other invertebrates were

subsequently transferred to 70% ethanol. Stomach and intestinal

contents were kept separate (except in two cases). Prior to sorting,

digesta were placed in a strainer and flushed with water to separate

food items.

Food was initially sorted with the unaided eye. The degree of

sorting of stomach and intestinal samples was equal, except as regards.

sponges. Sponges in the stomach were identified as fully as possible,

with an effort being made to assign all fragments. Because of the gross

similarities of many sponges, especially within the family Stellettidae,

initial sorting had to be routinely verified by examination of spicule

preparations (see below). Sponges in the intestine were closely

examined, but, because of the progressive state of their digestion along

the intestine, quantification of species representation was not


attempted. All sponge material contained in the intestinal tract was

therefore categorized as unidentified sponge, in spite of the fact that

much was readily identifiable. Sponges contained in partial intestinal

samples, in combined stomach and intestinal samples, and in the sample

of unknown anatomical origin were treated similarly. Because sponges

constituted over 95% of the total dry weight of all samples, and because

siliceous spicules and spongin fibers are resistant to digestion, a high

percentage of intestinal contents could be identified to phylum.

Identification of sponges was made by comparison to a reference

series of specimens and spicule preparations that had been developed

with the assistance of a sponge specialist. In order to make permanent

spicule preparations, fragments of sponge were digested in 5.25% sodium

hypochlorite, and the spicules collected by centrifugation and careful

decanting. They were washed twice in water, and once each in 70%, 95%

and 100% alcohol solutions for dehydration. The spicules were collected

after each wash by centrifugation and decanting, and finally transferred

to microscope slides with a small amount of 100% alcohol. The alcohol

residue was removed by combustion. The spicules were permanently

mounted in Canada Balsam. Temporary spicule preparations to aid the

sorting of sponges were made by dissolving fragments of sponge in a few

drops of sodium hypochlorite directly on a microscope slide. Spicules

could then be examined immediately.

For a few species in which spicule placement or overall sponge

architecture was important for identification, whole mounts of sponge

tissue were prepared for microscopic examination. Thin sections were

hand-cut with a scalpel, then stained with 1% basic fuchsin dissolved in

95% ethyl alcohol. The sections were transferred with forceps through a


series of alcohol solutions (30%, 50%, 75%, 95% and 100%) for

dehydration (15 min. each). The sections were cleared in xylene, and

mounted on microscope slides with very viscous Permount.

Sponge classification follows Levi (1973) except where otherwise

indicated. Species names could not be assigned in many cases because of

lack of diagnostic characters in the material or problems in the taxon-

omy of the group. One of the most important families represented in the

samples, the Stellettidae, is badly in need of taxonomic revision.

Algae and the shells of mollusks showed little evidence of

digestion along the tract and could be recognized and sorted from

all regions. Algae and seagrasses were identified with the assistance

of an expert phycologist. Mollusks, fish eggs, and bryozoans were

identified independently by appropriate specialists. Most other inver-

tebrates were identified by me.

Food items were sorted according to 165 categories: 32 for

sponges, 55 for algae, 43 for mollusks, 19 for other invertebrates, and

16 for miscellaneous items. Individual food items were dried to a

constant weight at 1050C, cooled in a desiccator and weighed to the

nearest 0.01 g. The presence of items weighing less than 0.01 g was

also recorded. The use of dry weights to quantify digestive tract

contents introduces a bias because of differences in the ash weights of

food items. Sponges with high levels of silica are overrepresented, for

example, whereas sponges with little or no silica, such as Chondrilla or

Chondrosia, are underrepresented. Biases exist across groups as well;

that is, algae are underrepresented as compared to sponges and mollusks,

and soft-bodied organisms such as coelenterates are more poorly repre-

sented than any other group. In spite of these problems, dry weight was


chosen as the measurement criterion because it was judged to be more

accurate than wet weight or volumetric measurements. In the case of

sponges these introduce unique problems (Ruetzler, 1978).

An inherent bias in diet studies based on digestive tract contents

is introduced by differential rates of digestion. Less digestible items

in the diet are overrepresented, particularly when intestinal contents

are included in analyses. This type of bias is difficult to correct

for, without detailed knowledge of the digestive physiology of the


A total of 12.4 kg (dry weight) of gut contents was examined from

61 turtles. More than 95% of the dry weight was made up of sponges,

which have an estimated dry:wet ratio of 1:5 (Ruetzler, 1978). An

approximation of the total wet weight of material examined is therefore

in excess of 62 kg.

Food samples obtained from the stomach averaged 13.4 g dry weight

(range 0-65.7 + 14.5,N = 54); intestinal samples weighed an average of

257.6g (range 0.1-1096.0 + 327.4, N = 35). One partial intestinal

sample exceeded this maximum value, weighing 1378.9 g. The entire

digestive tract contents of 37 turtles averaged 281.7 g (range 0.59-

1113.7 + 330.38).

Data Analysis

In order to make comparisons between food samples of different

amounts (i.e., from small vs. large animals, or empty vs. full digestive

tracts), dry weights of individual food items were converted to per-

centages for each turtle. The average percent dry weight of a

particular food item in all turtles was then calculated. The chief


advantage of mathematically weighting data in this way is that equal

weight is given to each individual in the sample (Swanson et al., 1974).

Analyses were also calculated on the basis of percentage of total dry

weight. The percent dry weight contribution of an individual food item

or category to the total dry weight of all food items consumed by all

turtles was calculated. Although the implications of this method are

perhaps more intuitively clear, this treatment has several disadvantages

(Swanson et al., 1974). A few individuals consuming large amounts of

rare food items can distort the data. Data can also be biased towards

large individuals because of their larger contribution to the total dry

weight of all food items.

Importance ranks were calculated as the product of the average

percentage contribution and the frequency of occurrence of the item in

all turtles. This ranking method was adapted from Hobsen (1974), with

dry weight percentages substituted for volumetric percentages.

Laboratory Analyses of Fresh Sponges

Collection and handling of sponges. Live sponges were collected in

the Florida Keys at Key Largo, Tavernier, and Big Pine Cay, and

transported on ice to the laboratory in Gainesville. Some were

then temporarily frozen for storage; others were processed immediately.

Sediment adhering to the surface of the sponge, or present in the

aquiferous system, was removed as thoroughly as possible with running

water and a soft brush. All visible epibionts were removed with

forceps. Large sponges were cut in blocks to facilitate drying. The

samples were dried to a constant weight at 60C in an oven with strong


circulation, and stored in plastic bags until used. For analyses of

nitrogen content, ash content, and energy content, dried sponges were

ground in a Wiley mill (#20 screen). Several fragments taken from

representative parts (mesohyl, pinacoderm) of each individual sponge

were pooled. Because of the small size of some of the specimens of

Chondrilla nucula, one of the samples is a composite of three

individuals. Maximum storage time of all samples was five months.

Spicule content. Several fragments taken from representative parts

of each individual sponge were pooled. The fragments were dried to a

constant weight (total 0.3-1.3 g) at 105C, and transferred to flasks

containing glass beads. Concentrated nitric acid was added, and the

flasks were gently boiled until no further reaction (foaming) occurred

and the solution became clear. Spicules were collected under vacuum on

Whatman glass fiber filters (934AH Reeve Angel) and thoroughly rinsed

with distilled water to remove acid solids. Spicules were flushed with

95% ethanol into dry, weighed aluminum pans, and dried to a constant

weight at 105*C. High (up to 10%) experimental error was observed using

this method and can be attributed to sampling difficulties imposed by

differential spicule distribution. This method has been used in order

to make results comparable to those of other workers.

Ash content. One-gram samples of ground sponge were dried to a

constant weight at 105C and ashed in a muffle furnace for 3 hr at 5000C

(Allen, 1974). Each analysis was carried out in replicate; values for

replicates were accepted within 2% error. Ash values were corrected for

water of hydration of the silica in the spicules, based on the findings

of Vinogradov (1953) and Paine (1964). The correction factor was

calculated from the weight loss observed upon ashing dry (105C) cleaned


spicules of Geodia neptuni for 3 hrs at 500 C. The spicules had been

isolated with boiling nitric acid according to the method described

above. The average weight loss observed for three samples was 3.95% (+

0.16; N = 3). Ash content was also determined for samples of intestinal

contents of three turtles. The digesta had been originally preserved in

formaldehyde, transferred to alcohol, and dried at 1050C. The same

procedure for ashing was followed as outlined above.

Scanning electron microscopy. Standard procedures were followed in

preparing sections of the intestinal epithelia for examination with the

electron microscope. The intestines had originally been fixed in

formaldehyde (1 part 37% formaldehyde:19 parts sea water) and then

transferred to 40% isopropyl. Digestive tracts were preserved and

transported with their contents in situ. Microscopic examination of the

intestinal epithelia had not been anticipated. The extent to which this

treatment affected the embedding of spicules in the epithelia is not

known. Given the delicate nature of the epithelia of the large

intestine and the abrasive characteristics of the digesta, I have little

doubt that embedding is a natural phenomenon. Nevertheless, handling

procedures may have caused additional embedding. Embedded spicules were

found in small numbers in the one intestine in which food was not

transported. The specimen was a reproductive female that had very

little food in its digestive tract when captured. The phenomenon of

spicule embeddment deserves additional study, using more appropriate

handling and preservation techniques.

Nitrogen determinations. Total nitrogen content was determined

using a semimicro version of the Kjeldahl method, with the salicylic

acid modification described by Nelson and Sommers (1972). The amount of


NH3 in 10 ml aliquots of the digests was determined by steam

distillation and hand titration. Replicates were accepted within 3%

error, except in the case of one specimen of Geodia neptuni (3.6%) and

one Spheciospongia vesparium (4.8%). Values were corrected for

percentage dry matter and percentage ash (corrected for water of

hydration) based on results of separate analyses using portions of the

same powdered sample. Dry matter replicates were accepted within 1%

error; ash replicates were within 2% error.

Energy content. Energy content of sponges was determined by

combustion of ground samples in a Parr oxygen bomb calorimeter

isothermall jacket). Procedure and calculations were carried out

according to the Parr manual (Parr Instrument Co., 1960). Corrections

for percentage dry matter and percentage ash were obtained by separate

analyses carried out on portions of the same samples. Replicate values

were within 3% error, except for Geodia neptuni (4.1%).


Composition of the Diet

Overall Composition

An overall summary of the diet is presented in Table 2. Several

broad categories of food items are ranked according to their percentage

contribution to the total dry weight of all food items examined. All

turtles are considered in the first analysis, including those for which

only partial digestive tract contents were available. Because of dif-

ferences in sample amounts and composition, gravid females have been

excluded from the second analysis. The sample is further restricted to

turtles for which the entire contents of both the stomach and intestine

were available, in order to remove any bias introduced by different

degrees of digestion of partial samples. The percentage composition is

very similar in both cases, and equivalent ranks result.

A second, perhaps more quantitatively accurate, approach to sum-

marizing the overall diet is presented in Table 3. This analysis, which

uses the restricted data set as specified above, reports the mean per-

centage of the dry weight contributed by each category. Categories are

then ranked by the product of this mean and the percentage occurrence of

items in the category in all turtles (Hobsen, 1974). This method of

summarizing the overall diet produces results almost equivalent to those

shown in Table 2. Sponges remain clearly dominant; the ranks of three

minor categories are rearranged.


Table 2. Overall composition of digestive tract contents of
hawksbill turtles (Eretmochelys imbricata). Values represent per-
cent dry weight contribution of items in each food category to total
dry weight of all food items consumed by all turtles.

Food Category


% Composition
12.4 kg (dry wt)
N = 61a

% Composition
10.3 kg (dry wt)
N = 28b



Substrate Material










aIncludes partial and complete contents.

bIncludes complete contents only; gravid females excluded.








Table 3. Overall composition of digestive tract contents of 28 hawksbills
(Eretmochelys imbricata). Sample consisted of 10.3 kg (dry weight) of
digesta. Gravid females are excluded from the analysis. Rank is calculated
as the product of the average percent dry weight contribution and the
percent occurrence in all turtles.

Food Category


R %
Dry Wt.


% Turtles
with Item



1 94.2 + 12.0





2.1 + 3.2

2.1 + 8.9

1.1 + 4.7

0.4 + 1.8

0.1 + 0.1 0-0.6

















Mollusks 6

53.6 0.03


Several categories of food items were usually represented in each

turtle, as indicated by the values for percent occurrence in Table 3.

However, sponges were clearly the dominant food category. The

cumulative contribution of all non-sponge food items in all analyses is

less than 6%. It should be pointed out that a sizable portion of this

6% was not ingested purposefully. Substrate material, algae, gastropod

mollusks, ophiuroids, hydroids, polychaetes, shrimp and scyphozoan

scyphistomae were found attached to, or inside of, sponges taken from

the digestive tracts.

Amounts of food present in the digestive tracts of 34 hawksbills

are plotted against carapace length in Figure 3. Tracts were sampled at

varying degrees of fullness, which explains the large variation in

values observed for large turtles. Female turtles that were gravid, as

evidenced by the presence of shelled eggs, or their being captured on a

nesting beach, had little or no food in their digestive tracts (stars in

Fig. 3). The average amount of food in all nine gravid females avail-

able for study was 15.4 g (+ 12.5, range 0.6-38.2) compared to an

average of 616.8 g (+ 275.6, range 230.4-1113.7) in 13 nongravid adult

females and adult males. The two samples spanned roughly equivalent

size ranges, as shown in Fig. 3. There was no overlap in values between

the two categories. The two nongravid adult females included in the

study contained large amounts of food (847.7 and 592.4 g).

The digestive tracts of gravid females showed conspicuous differ-

ences in appearance upon examination in the field. The tracts were

contracted, with small lumens, and contained appreciable amounts of

blackish-green fluid, presumably bile. In several of these females, the


U ,

0 44


-41 O


o n
,4 cc


cu 0)
0 c5
4-1 2
Q* a)
LC 41 -
(i e

60 M
t-1 ra
F 0*i-




* -
0 -


dO iNnOivV ivioi


C |


= I


F -






food present in the digestive tract was totally unrecognizable,

suggesting that it had remained there for a long time.

The maximum amount of food in a turtle of a given size appears to

be described by an exponential function (Figure 3). This is to be

expected, inasmuch as volume increases as the cube of a linear measure-

ment. Too few values are available for turtles between 50 and 70 cm

carapace length to allow plotting of the line. A maximum value of

1379 g was observed for a partial sample from a male hawksbill 72.9 cm

in carapace length.


Sponges were present in all but one of the 61 hawksbills included

in quantitative analyses and in 63 of the 68 available for study. Four

of the five without sponges belong to a size class that is believed to

occupy a pelagic habitat (see section on lost-year turtles). Two food

samples removed from the mouths of hawksbills encountered on reefs off

Palm Beach, Florida, and a fecal pellet from a 33.6 cm turtle caught off

Pine Cay, Caicos Islands, consisted entirely of sponge. For the purpose

of examining patterns in the percentage sponge composition associated

with size, sex, reproductive condition, and geographic origin, 37.

hawksbills for which entire digestive tract contents were available were

considered. In some cases, missing values for size, sex, and

reproductive condition dictated further restriction of sample sizes.

Gravid females showed considerable variation in the percentage of

sponges in the digestive tract (Figure 4) and as a group had a smaller

mean value (3 = 54.9% + 28.3, range 13.0-88.6, N = 9) than males and

nongravid females (x = 94.2% e 12.0, range 41.9-99.9, N = 28; Mann



c c


o1 a

0 >


4-( (U
^ < S






SI I I I I i i I

o0 0 0
0 E) NOdS D
3DNOdS d0

0 0 0 0 0



Lo E




0 Q-
* <
o c

0 0-



Whitney U Test, p = 0.0001). For seven of nine gravid females, however,

sponges were still the predominant food item. The wide variation in

percentage of sponges in gravid females can be attributed in part to the

small size of the food samples on which the percentages are based

(average sample = 15.4 g). When gravid females are excluded from the

data set, male and female turtles showed no significant difference in

the mean percentage of sponges in the digestive tract contents (females:

x = 95.2% + 7.5, range 78.3-99.2, N = 7; males: x = 96.4% + 7.7, range

72.3-99.9, N = 12; Mann Whitney U Test, p = 0.2067).

The percentage of sponges in the samples did not vary with size

(Figure 4). The mean value in samples from 28 males and females was

94.2% (+ 12.0, range 41.9-99.9). Gravid females were excluded from the

analysis. Other than these, there are only three outliers on the graph.

The most aberrant sample, with only 41.8% sponge, is from a 23 cm

hawksbill caught in the Dominican Republic. It is the smallest turtle

included in quantitative analyses. There is evidence that a major

ontogenetic change in habitat, and consequently diet, occurs at approx-

imately this size, and this would perhaps explain some of the unusual

aspects of the sample. The sample consisted of 47% invertebrates other

than sponges (largely goose barnacles and false corals). This was the

highest value observed for this food category for 61 turtles (see Table

3). It also contained vertebrae and fragments of the chondrocranium of

a fish. Fish remains were found in no other sample. The presence of

substrate material in the sample is an indication that the turtle was

feeding, at least in part, on the benthos.

Age classes (adult and nonadult) were also compared in order to

test for differences in percent sponge composition associated with size.


No significant difference was found between the means of these two

categories (adults: i = 96.2% + 7.6, range 72.3-99.9, N = 12; nonadults:

i = 92.4% + 14.9, range 41.9-99.6, N = 15; Mann Whitney U Test, p =


Geographic differences in the percentage of sponges in the samples

were also examined. Samples were grouped according to three regions of

geographic origin: Panama, the Dominican Republic, and the Lesser

Antilles (which includes the Leeward and Windward islands). Gravid

females were excluded from the analysis. No significant differences

were found in the mean values in samples from these three regions when

the aberrant sample from the 23 cm hawksbill from the Dominican Republic

(see above) was excluded from the analysis (Panama: x = 96.3% + 8.0,

range 72.3-99.9, N = 11; Dominican Republic: x = 95.8% + 2.2, range

93.4-97.9, N = 4; Lesser Antilles: i = 96.2% + 5.8, range 78.3-99.6, N =

12; Kruskal-Wallis Test, p = 0.1089).

A total of 584.0 g of sponges was examined from the stomach con-

tents of 54 turtles. Of this, 529.6 g (90.7%) could be identified. An

average of 91.1% (+ 15.62) of the sponges in individual samples was

identified. Stomachs contained an average of 10.8 g of sponges (+

13.64, range 0-65.2, N = 54). As many as 10 species were present in the

stomach of a single individual (i = 3.4).

Thirty-one species of sponges were identified, all belonging to the

Class Demospongiae (Table 4). No calcareous, sclerosponge or hexac-

tinellid sponges were found. Seven orders were represented in the

samples (Figure 5). The orders Astrophorida, Spirophorida and Hadro-

merida accounted for 98.8% of the total dry weight of all identified

sponges. These orders are members of the subclass Tetractinomorpha,


- ,
n) nn m

- co .
M > ca o ,- 4
o: o

O a r- cM CM4 r -It 00 0 r-



a) I a 4

o *
2 .u *** .(COCO ..

oLU 0. 03

o4 4 Ca4
S u1 oo

0 4

- r C
(U w 00 0

0 C:6 O S
.ur ao ar4r

< u -<" 4 C

"0 -o

C 4 1 CO

(f r- C: (f *< i- >- vi )-i C (0 r0 (O C8 (O C3 (OT )i-


Lnn n n *
^ *I |I |
-4o I

-- C,- C1 -44
.o n o "4
c X -N 'J'j- -i r- 0 0Q-4 (' 01 -r I w

.. ., ,0- -4

i ,-4 00 0f0 0 0 0 0
C4 O0 0 C



S+1+1+1+1+1+1+1+1 +1+1+ +I +Im +I o

4 -4 o o E
>,- o4 .8 8 CO
-, *

$4 ca

> w rnu r o a a
4 r o 1 ooo o < -

S o

O40 0 H 4

c c coa0n4 r-4 'o Q
C- >-4 0 10 3 0 P. 0

0 w S u M 4 0
W W W+ c M +1 C r- t) "1 0 t a cu m s 1
w bo0

CC ClO S. 4 o O UC
>I u o o i u r I a 0a u o -

4- 4 r. ajo c -

4 01 0 -4 $ 1o"C 1fl to 0
Ec- o o o
- a EO E' CC < Co CO pa0 O W >

Ufl 0 Co r$1

'J3 L >

w 0


4 2 .


0 0 cu

0u 'u 0

0 0.

0 0 -

0 4

41 0

-4 0 W 0
0 "-
0 0 0

u4 U 4
0 3 c

Z- 0

0 0 0.

e 0 o. 0
zo o2




VOIU3710SO'ldVH 0


VQIU1371SO71030d 03,



0 0


0 0
in IQI

0 0 0

d-0 3E)VIN30i3d



which is distinguished from the subclass Ceractinomorpha by oviparous,

rather than viviparous, reproduction.

Table 4 lists the average percent dry weight that each sponge

species represented in the stomach contents of all 54 turtles. The

sponges are ranked according to the product of this value and the per-

centage occurrence of the species in all 54 turtles. The ten species of

highest rank are listed in order in Table 5. Also listed in this table

are the 10 most important species as calculated by percentage contribu-

tion to the total dry weight of all sponges. The 10 species shown in

each of these two columns represent, respectively, 79.1% and 87.4% of

all identified sponge. All are either astrophorids or hadromerids.

Chondrosia and Chondrilla are considered to be incertae sedis in Levi's

(1973) classification, although he presents them in sequence with astro-

phorids and comments on their affinity with either this order or the

Hadromerida. The affinities of these two related genera and either the

Astrophorida or Hadromerida are widely recognized (Wiedenmayer, 1977;

Bergquist, 1978).

Rank indices based on the product of average percent dry weight

contribution and the frequency of occurrence (first method above) were

also calculated by genus. For this analysis, values within a genus

(i.e., for all Ancorina, all Myriastra and all Tethya) were combined.

The resulting rank indices are illustrated in Figure 6.

Other Elements of the Diet

Substrate material, defined as stones or gravel of calcareous

origin, was found in the digesta of all but seven of 61 turtles. Much

of it was attached to sponges and was probably ingested incidentally.


u(i4 (U ~ldU)4 -
U) UJ-c
4 41 z wt
CO 0) -4


u~ Lw
c 41 0 .
0 -4
C 1

c 3 0 -
.1 4 = 0.
(4 0

-0 CO)

co J
SC ,.q t
0 -,Q0

.46 0 )0

. C O r-
0 cac

4-j 41U

U) $0

4j Sw ca_^ *
-4 S <" a(


4-) 0-

0 1 0- U 0)
U) Uc U) U)

to U 4

E- = .0 0 0 (
00. U) (U

02 0 >.,4
gs^ (d
1-1m ou -
U -U U) C0
-4 (U c

.r(U p j

021 U)
(U 4-1 i a o

*0 I^ !* (d

4 n -i -
,0 30 0 U 0

2) U)( U "l*-
4-i Qc 0a 0
-W CJu)

.0 5-' O v

si U

dOU) U) 0C
.0U)U 0r~
rc~aU ( UU E -
P5-44:. 0



e M







-4- e0 n -.I

to (U


-- U

al 01 A

0 a 0 oj r
0.0 0. 0
U) 8C

0: B w )W -
(U u) 1-1
-4 (U U) -4 5i

o -a 0 U) c

0 a) U 3 4:
0 u n

i- 00 c


C* 0. -4


a -i o

(U 0

-4 cq CO IT tn r- 00 ON

ca c

co co
-4 0
30. ( --
U a) 4w4
E* 0 U
e c ( 0 U U
0. *( U U) 0 0
B3 U) 00 (U *
w0 ) ( ( '
r- U U) 0o a -i -4 U) U) U

a a- o 01 ( 0 e o
0 O O U U) O C 4-4
0 u OU 0 (U -' '-I O 0. 4-4
4: 0 u r-I 3 U 4 (U u
u 00 0 (X M t U < E-

o o

0 u



4 -4l

o to
u a

L0 c)
ci )



s.d >

U (d U)
0 '-I
* ) C
ca'S c
s I

o g
Ql >-a

*" UlS "
M30 *1
^ "~r q
(" C o



VAH 131



VtIN3N0103 0



TT ,
V~ ----------------{


0 0
0 0




0 0
0 0


The average percentage of this item in the gut was higher for gravid

females (R = 27.1% + 31.9, range 0-80.6, N = 9) than for all other

turtles (x = 2.1% + 3.2, range 0-16.6, N = 28). The digestive tracts of

two gravid females contained little other than substrate material (80.6%

and 77.3% of dry weight).

Over 50 species of algae were found in the digestive tracts of the

61 hawksbills included in the quantitative analyses. The 15 species

most frequently represented are listed in Table 6. Although algae were

present in most samples, they contributed an average of only 1.1% of the

dry weight in the 28 nongravid turtles for which entire digestive tract

contents were available (Table 3). In only six of these turtles did

algae contribute a larger percentage, the maximum being 25.1%. Several

species were found attached to sponges and were probably ingested inci-


Codium isthmocladum and Lobophora variegata were found in

sufficiently large pieces and quantities to suggest purposeful inges-

tion. One adult male hawksbill had eaten 158 g of Codium isthmocladum,

in addition to 457 g of sponges.

Seagrasses were present in very small quantities (maximum of 0.25

g) in 16 of 61 turtles. Thalassia testudinum, Syringodium filiforme and

Halodule wrightii were identified.

The 61 turtles included in quantitative analyses had remarkably

little man-made litter in their digestive tracts. In five individuals

small fragments of plastic, paper or string were found, the largest item

being a 0.13 g piece of plastic. Man-made litter was much more preva-

lent in the digesta of the four small hawksbills that stranded on

Florida beaches.


Table 6. Algae most frequently represented in the digestive
tracts of hawksbill turtles (Eretmochelys imbricata).
N = 61.

No. of % Turtles
Species Occurrences with Item

Dictyopteris delicatula 22 36.1

Dictyota sp. 19 31.1

Lobophora variegata 17 27.9

Microdictyon boergesenii 16 26.2

Halimeda sp. 15 24.6

Bryothamnion seaforthii 15 24.6

Codium isthmocladum 14 23.0

Kallymenia limminghii 13 21.3

Anadyomene stellata 13 21.3

Gelidiopsis planicaulis 12 19.7

Pterocladia bartlettii 11 18.0

Caulerpa microphysa 11 18.0

Galaxaura sp. 9 14.8

Caulerpa vickersiae 7 11.5

Gelidiella sanctarum 7 11.5


Lost-Year Turtles

There appear to be no data in the literature on the diet of wild

hawksbills of the size range represented by the four specimens that

stranded on Florida beaches (Witzell, 1983). Because of considerable

interest within the scientific community in marine turtles of this size

class--particularly as regards their habitat occupation--the results of

analyses of the digestive tract contents of these specimens are reported

in detail in Table 7.

Structural Characteristics of Prey Sponges

Inorganic Constituents

Table 8 presents data on the spicule content of astrophorid,

spirophorid, and hadromerid sponges that were identified in the stomach

contents of Eretmochelys imbricata or were represented in the samples at

the generic level. Sponges of these three orders accounted for 98.8% of

the total dry weight of all identified sponges. Because identification

to species was not possible for many of the sponges that had been eaten

by turtles, values in the literature for all Caribbean species of the

genera represented have been included. Data from Bergmann (1949) and.

Ruetzler and Macintyre (1978), used to supplement those obtained in the

present study, were derived by the same isolation technique.

Spicule content of the 31 sponge species found in the stomach

contents of hawksbills (Table 4) varies widely. Chondrosia, Halisarca,

and Verongia contain no spicules at all. Chondrilla nucula, the second

most frequently represented sponge in the samples, has very few, and all

are microscleres. Geodia, which was identified from 26 turtles, has one


"0 4-4

-) 0 4

0O 0
L m M

co z

th40 0 1

0. 0 4 0

MW 0 w 0
0 w 0 0 r-
4a) 4-J I.. 0

u.o. o 02 (U 1-i .

4 a 0 -a

0 -4 M > % M0 00 0 .4 co
TO ,4 u 0 ( M 0.. a 0
c C "0 M LW W a to -
(U ( 0 "- 41 ZC a 44 w0
2g cc U 14 0 w0 c 41
S 4- 1- 4
0 O > C 1 3 C co = 0 C
a u B O F i3 E-* 0 co


0 0

8 ca
A-i 8 4 33

co r 1-4

e U- w- 0 cu
t" -u C Q 1

w 0c a
W 0

0 com 1 M `
0 C: -4 ca 0. c -4

0 O d ^

O 0 0
a1 0 1 1. U M

0n a r
o co co 9 in W

CO CL T "-4

St -44 3 -4
3 4USI T ) 0 Q* 11.
(Uo i u-i .C (u ,4-4U 0
U00 a 3U ^| 0 J- (

0 a ; u 0 0 0 10J M 1.4 1-4.-4
..0. U W (rt .C 0 10 1 c *C *> ai 'o
4fl -O4- IO gC.-~ (UM*- 3 r O M 1


0 -

4-4 4 0

0 m Cw
41 m -cu wC

O w 0 0
0 Cc 0 o 0

W)O 0X 0 0

1-0 wO nI C
to *s ". 0 a

e -4 0 E..1 4- w
0 0 0ca 0 0 = C 0

o a C 0 0 3 o a-4 -i
U iC > J-40 k 0



S 4-1
0c ) C-i

*0 4-1
-40 4 0 c
c c0 CfO
ll *0 a4-1
4-1 00 0) CO
cc 1..4 -4 a M
2 CO U ~ CO L-
ca a J o] o
1 CO 4-1 )CO

a) cu 0 600O
2 J-9 2 4-i -4 0. E3 2 CO 1O -o
0) W 00 0) 0.* w--
4 4- C 4 U 4 ) *4-4 00 4t- 0 4-)
i-C c co c i -C $4 0) *Vcc -,4 U 4
U0 CO 0) -0 4-) CO U 4- ) *UO
ot (O 0s 3 CO -4 00 *4 -l C .0
0 ( CO Cw C '-o w M ^C -^ 3 1-. CO C
00 0 0 0 M *- w H

0) 0 3 0) ( 1 )C
CO)' Oni ag 4-)
0.w- CO 0.. >- C O

c W o,- C0 -
a 0-i >u 0 CO

00j 0 0 :1 0 N 3 1
"o "4 M C M i

EC M 00 X0 m 0 a
3 C O mB CO d 1 c i 4

0 rU o 1-1 0 Ci
4( s U o4-1 0 U4CO o-
W = -C4 -T .- u 0. c a C
U0 w C 10 a. r- n

M 00 O U -cTO V ) 0 S


Table 8. Spicule content of astrophorid, spirophorid and hadromerid
sponges. Included are genera or species (denoted with an asterisk)
represented in stomach contents of hawksbills turtles. N = 1-2 indi-
viduals for this study.

(% of dry wt.)



Geodia neptuni

Geodia gibberosa
*Myriastra kallitetilla
*Chondrilla nucula

Cinachyra kuekenthali
Cinachyra cavernosa

Tethya aurantia
Aaptos sp.
Suberites compact
*Spheciospongia vesparium





Spheciospongia sp.

This study
Ruetzler and
Macintyre, 1978
Bergmann, 1949
This study
This study
Bergmann, 1949

This study
Bergmann, 1949

Bergmann, 1949
Bergmann, 1949
Bergmann, 1949
This study
Ruetzler and
Macintyre, 1978
Bergmann, 1949


of the highest spicule contents of all siliceous demosponges; values are

given in the table for the two species that are common in the Caribbean.

High silica contents are characteristic of the Astrophorida and

Spirophorida, in general. Ancorina sp. 1, Ecionemia sp., and several of

the Myriastra species identified in the samples were very spiculate. I

find no data in the literature on the spicule contents of these sponges,

or of the hadromerid Placospongia, and my attempts to collect them in

the Florida Keys were unsuccessful.

The total amount of ash in sponges is also of relevance to preda-

tors. Ash content is a measure of total mineral content, and in the

case of sponges can be considered an indicator of mechanical strength or

fortification. It can be seen in Table 9 that there is considerable

variation in ash content among prey sponges. Comparison of Table 8 and

9 shows that for some species ash content greatly exceeds spicule con-

tent, e.g., for Cinachyra kuekenthali, Myriastra kallitetilla,

Spheciospongia vesparium, and Chondrilla nucula. It should be noted

that the same individual sponges were used in both analyses. In the

case of Chondrilla nucula, the difference between the two values is

largely due to adhering calcareous sediment. One habit of this species

is encrusting, and specimens frequently contain embedded sediment.

The highest ash content was found in Spheciospongia vesparium

(64.5%), the loggerhead sponge. This sponge species ranked sixth in

terms of contribution to the total dry weight of all identified sponges.

Geodia neptuni also has a notably high ash content.

The sponge Chondrosia (not analyzed in the present study) has one

of the lowest ash contents of the sponges represented in the diet. This

species lacks siliceous spicules and specimens are usually free of


Table 9. Ash content of a representative series of Caribbean
demosponges. Values are means (N = 1-3) + S.D. when N = 3. Species
identified in stomach contents of hawksbill turtles (Eretmochelys
imbricata) are denoted with an asterisk; "+" denotes genera that were
represented in the samples.

Sponge (% of dry wt.)

+Geodia neptuni 58.5
*Myriastra kallitetilla 36.6
*Chondrilla nucula 25.1 + 3.2

+Cinachyra kuekenthali 52.1 + 3.9

*Spheciospongia vesparium 64.5

*Iotrochota birotulata 41.6 + 4.3
+Agelas conifera 31.5

Haliclona compressa 39.1

Ircinia strobilina 37.2
Spongia tubulifera 31.0 + 2.8


adhering foreign calcareous material. Randall and Hartman (1968) deter-

mined a value of only 1.5% for Chondrosia collectrix, the most common

Caribbean species. Chondrosia was represented in 13 turtles in the

present study.

No ash content data are available for several sponge genera that

were important in stomach contents, e.g., Ancorina, Ecionemia,

Placospongia, and Suberites. Ash content is certain to be high for the

first three genera, because of their high silica content. It is notable

that the ash contents of Ircinia strobilina and Spongia tubulifera, both

of which lack siliceous spicules, are still of the order of 30-40%.

Ircinia is known to incorporate foreign calcareous particles within its

spongin skeleton, which may account for the high value. Spongia does

not incorporate particles but may contain iron in its spongin fibers.

Ash values of intestinal contents were determined for three

turtles. Samples that appeared to have high ash contents were purposely

selected, in order to establish a maximum value. Ash contents of 92.0%,

76.6%, and 74.3% were measured. Because of species composition, the ash

can be considered to be mostly silica. Figure 7 shows the glass-like

appearance of dried intestinal contents. The first sample was taken

randomly from 490 g of intestinal contents. Sediment (1.56 g), algae

(0.5 g), and gastropod mollusks (0.21 g) had been previously removed.

The latter two samples were taken from unsorted digesta contained in the

terminal part of the digestive tract, just anterior to the junction with

the cloaca.

Spicules in astrophorid, hadromerid, and spirophorid sponges are

not associated with spongin, and upon digestion are liberated in the gut

of the hawksbill. As a result, the large intestine contains


extraordinarily large numbers of sharp, free spicules. Scanning

electron micrographs of the intestinal epithelia revealed numerous

embedded spicules (Figure 8). The extent of penetration in the gut wall

was not histologically determined because of sectioning difficulties

caused by the large number of spicules.

The principle megascleres of astrophorids and spirophorids are

tetraxonid (4 axes) and are among the largest (up to 5.3 mm in one

species of Myriastra in the samples) siliceous spicules found in

shallow-water demosponges. Geodia, Myriastra, Cinachyra, Ancorina, and

Ecionemia contain trienes with sharp, and in some cases recurved, hooks.

Each clad is bifurcated in Ancorina sp. 1, so that one spicule actually

bears seven sharp points. The cladomes--bearing the hooks--are usually

directed outward, toward the surface of the sponges. Needle-like

monaxonid spicules of the hadromerid, Suberites, project from the

surface to form a hispid coat. The megascleres of Jaspis are robust,

double-pointed monaxons.

The principle megascleres of the orders of siliceous sponges that

are not consumed by hawksbills are simple (1-axis) oxeas. Megascleres

of non-prey sponge orders tend to be smaller than those of prey sponges.

Spicules are noticeably concentrated in the periphery of several

prey sponges. Millions of sterrasters are tightly packed to form a

thick (up to 4 mm), stony cortex in Geodia. It has been described as a

"sterraster armour" (de Laubenfels, 1950a). Placospongia also has a

stony cortex, formed by irregular polygonal plates of small sterrasters.

Cortices are not characteristic of the siliceous ceractinomorph orders,

Poecilosclerida, Haplosclerida, and Halichondrida.

Figure 7. Dried intestinal contents of a hawksbill turtle
(Eretmochelys imbricata). Glass-like needles are siliceous
spicules. Ash content ca. 92% of dry weight.

Figure 8.
thelia of

Scanning electron micrograph of intestinal epi-
a hawksbill turtle, showing embedded siliceous



Organic Constituents

There are two distinct components of the organic skeleton of demo-

sponges: spongin and collagen fibrils. Both are made of the fibrous

protein, collagen. The sponges identified from the stomach contents of

hawksbills show distinct properties with respect to both of these con-


The sponges that were predominant in the samples apparently contain

no spongin in the form of fibers speculatedd spongin fibers or horny

fibers), and little, if any, spongin in other forms. As Table 10 indi-

cates, the Astrophorida, Spirophorida, and Hadromerida are three of six

orders that lack spongin fibers. With the exception of the small and

primitive group Homosclerophorida, these are the only orders of sponges

that lack spongin fibers and are possible food sources, the Desmophorida

and Tabulospongida being unsuitable because of their stony consistency.

The types of sponges that were identified in the stomach contents

of hawksbills are rich in collagen fibrils. Sponges of the subclass

Tetractinomorpha tend to have a higher density of collagen fibrils in

the intercellular matrix than do those of the subclass Ceractinomorpha

(Garrone, 1978). By contrast, loose-textured sponges are characterized

by extracellular spaces poor in fibrillar components. The tetractinel-

lid tetractinomorphs (which include Astrophorida and Spirophorida) are

particularly rich in collagen fibrils (Levi, 1973).

There is considerable documentation in the literature of a high

collagen fibril content in several genera that are consumed by

hawksbills. Tethya and Chondrosia are singled out by Garrone (1978) as

examples of dense-textured sponges. In the latter, fibrils constitute

the only skeletal framework of the sponge (Garrone et al., 1975). A



II 1 1 I




0u w






)0 0


a -4

a u
L .


c. 4l

02 0

>CO '

c a

-4 a

en >
00 0

C ^
0. u



.0 e


0 00

s S "
i- 5 j
.0 w 02
CO a M
ca 0.CO



SI I I +


0 4 4 4 'T cO

' 0 0 0 -i -.-4
. ,C w r-4
u 0. 0. 0. 2E -4
0 000 0
41J *I4 04 V -.
w) w 0. 1) CO X

co "o oa a v






I I 1++ -



+ 1

+ C+ II


5-i s-i 'O 'C -l

0 U 0 0) -i

4- -4 =1 0 0
U r4 t 0 >,
CO U -1 -4 4U
Ui 04 0.
0 C CO
CL a a Q


w 4t

5- 0




+ 1

ca *-4 '0
.4 4

0 -i 0
0. 0 0.
0 00
0 0 0






high collagen fibril content has been observed in Jaspis stellifera

(Wilkinson, 1979) and in Stelletta grubii (Simpson, pers. comm.). The

latter is a member of the Stellettidae, which includes the prey genera

Myriastra, Ancorina, and Ecionemia. Fibrillar bundles, formed by the

association of several hundred collagen fibrils, have been observed in

Chondrosia, Tethya, and Suberites (Garrone, 1978).

The amount of collagen fibrils present in the digestive tract

contents is high, not only because of the particular species of sponges

present, but also because large amounts of fibril-rich ectosome or

cortex had been eaten. Densely packed collagen fibrils form the cortex

of Chondrosia, Chondrilla, and Tethya and the thickened ectosome of

Jaspis stellifera and Suberites massa (Garrone, 1978; Wilkinson, 1979).

Collagen fibril content is also high in the external asexual buds that

occur in some sponges, such as Tethya lyncurium (Connes, 1967). A large

number of buds of Tethya cf. actinia were present in the digesta.

Toxicity and Antibiotic Activity of Prey Sponges

A considerable body of data on the secondary metabolites of sponges

is accumulating as a result of natural products chemistry research. In

only a few instances has the relevance of specific chemical constituents

been developed in the context of predator-prey interactions. Data on

the toxicity and antibiotic activity of these chemical constituents are

far more available. Toxicity is usually tested by immersing fish in

water containing sponge extracts. Evidence from the literature bearing

on the toxicity to fish of sponges eaten by Eretmochelys is presented in

Table 11. All data available for genera that were represented in the

stomach contents of turtles are included. As is evident in the table,


O CO O-4

4-4C E ,- 3 C
00 0

co a)nr 3
w 4 )
M -c 00 0 er c o
Cd c

Oa 44 to cc 0a

CC ~ ~ o- C CCC H M C
0 u uu Q u W Q

-4 -, c -H v,

W~LJ $.i
0 000 0 0w

AD 1 1 4 1

00 0 00000 r-00 c

.2 00'-40 00o000 -4"14000 0.
4-J CD S
z5- z = i< z x X:zz -



,w WU 41 414

sw W CC C ,-4C :

5~ '"
ca V c ca -4 s4 "

r 4 >1 C-14co C 0) raca.0

C4-4W 3 c u4- 0 3 -
w 0 w I 0 a 4-) (0 M
Q) c "a C 0Cu W (B-4 .-40C.-4I r.-i 3
u3 0 c C0) u C uO -4 0 0 to
0 c co-i 3u 1-4 3 Cu cn o u CO O u 'I
0. 0> w 0 C C C G 0 -4
Q) =^ Q wo io (o i w Q o Q) V 'V t4 "

41 .C0 w-4 43 .0 E!i 0 0 0l 4 1 ca co 0,
e) w zO C1 w 0) = l w w = w 0 a)

C E E- 1= I= 1 0' U c E-00 0u

CU .* U fi 41( 0 ( i- 0.IC

4 00


boC 4
0 c 0a 4
0 (00 0
0. u 0 0 i-4
tk to 0U 31 iQ 0 < -1> ) xU ^ t l

(: C: "c c; W c

(0 w C4 co
0C 0-
U u "o C: o

0tj co 0u Li:0
0 Cu3 0u3 0)0 01

U-4 J co twJ CJnLW
-4 C .6CO 4 L
?n j 0n r. i- S4l vU = <
>, Q C:II co no cc u1

-1 0) cn 0 U > *
-. j w W4

U )
C trQ Oi- -

* I -i 0% ^t- <

) CO

COO b4
a) 0 0
E1 4 Cu -c
C- -4Cu 0


0 U
00 U U U
X X X -4 X
U O0 0 0 0 0
*) au au o u
4W1 1 4- 4 l i- *- l

1U 0 0 OJ 0 0 4,J
0 -4 0 00 .,- 0
4 O 4 C0H 4 J0 0 44 40

zsszZ>z zz


0 0 C
*f 4-4 20

0 C 0

C% -4d 0 ta = o
En ci U .14I rniw

C0a 0 *w U = C,
l >H i-lrl 4- ( *
0 Cd 0 0 0 0c) -4
*H O a i ca C ,n *m
C e 3 i a *i E o &

0 w- OOU -0
r eS c' U 0c 2 0 n

004 0c14-J 2 T 0

Q) ON -4i-
C -4


0 (O

U 0' )O

- C) 00 cO

-4 0 0) 0

F-ii t ( a
*- OJ Uc S-i !


intrageneric and intraspecific differences in toxicity of sponges have

been observed. The bioassay techniques used by different authors vary

in detail, but are generally similar. Test fish used in the bioassays

are as follows: Green (1977), goldfish (Carassius auratus); Bakus and

Abbott (1980), mosquito fish (Gambusia affinis); Bakus and Thun (1979),

sergeant majors (Abudefduf saxatilis); Bakus (1981), goldfish. Criteria

for toxicity ratings also vary from study to study, but in all cases are

based on fish responses, i.e., loss of equilibrium, convulsions, paraly-

sis, and death.

Several sponge genera and species that were determined to be toxic

to fish in these tests were important components of the stomach contents

of Eretmochelys, including Geodia, Chondrilla nucula, Tethya actinia,

and Spheciospongia vesparium. The toxicity of different species of

Geodia appears to vary, ranging from nontoxic for G. neptuni to mildly

toxic for G. gibberosa. Both are Caribbean species. Chondrilla nucula,

one of the most common sponges in the stomach contents, was found in all

tests to be toxic to some degree. Wrasses (Halichoeres bivittatus)

force-fed Chondrilla nucula from Caribbean Mexico showed "paralysis-like

signs" within 7 min and "convulsive-like signs" within 8 min (Green,

1977). Goldfish placed in water containing extracts of this species

died in only 34 min (Green, 1977). Specimens of Chondrilla nucula

collected in Puerto Rico have been reported to cause contact dermatitis

in humans (M.B. Mathews, pers. comm.). This malady is commonly asso-

ciated with the sponges Tedania ignis and Neofibularia nolitangere; I am

unaware of any previous reports attributed to Chondrilla nucula.

Tethya actinia obtained from Veracruz, Mexico, was rated as

moderately toxic by Green (1977). Tethya was a particularly common


genus in the stomach contents. One digestive tract was completely

filled with a sponge that was very similar, if not identical, to this

species. Another sponge that was considered mildly toxic in the above

tests, Spheciospongia vesparium, has been shown to be toxic when in-

jected intraperitoneally in mice (Halstead, 1965). Suberites ficus was

found to be nontoxic to fish. Extracts of another species of this

genus, Suberites domunculus, found in European waters, caused hemor-

rhaging and death in a wide variety of lab animals (Richet, 1906a,b).

Representatives of three sponge genera of minor importance in

stomach contents were also determined to be toxic in these bioassays.

Both Lissodendoryx aff. kyma and Hymeniacidon ? amphilecta were highly

toxic to fish; lotrochota birotulata, present in small amounts in 6

turtles, was found to be nontoxic in tests by Green (1977) and mildly

toxic in those of Bakus and Thun (1979). Green (1977) reported that

fish avoid the colored, strong-smelling exudate of this species.

Another area of sponge chemistry of possible relevance to predator-

prey interactions is that of antibiosis. The current, broad interpreta-

tion of this term, elucidated by Burkholder (1973), is that of "a

phenomenon in which special products of certain organisms severely limit

the life activities of other organisms" (p. 118). Marine demosponges

exhibit a high incidence of antibiotic activity. The usual test

organisms used in screening for this activity are bacteria and yeast,

although tumors and viruses are also tested. Bergquist (1979) points

out that "antibiotic activity demonstrated in the laboratory is a mani-

festation of something which in nature could also be toxic, bad tasting

or active in quite another way" (p. 390). Antibiotic activity is often

used to screen potential sources of secondary metabolites. According to


the literature, several sponges consumed by the hawksbill turtle have

been demonstrated to exhibit antibiotic activity (Table 12).

Nutritional Characteristics of Prey Sponges

Little has been written about the nutritional characteristics of

sponges. These animals are of no importance as a food source to people

and figure only slightly in the diets of most other animals. A thorough

study of the nutritional characteristics of sponges is obviously beyond

the scope of the present study. I have instead gathered data on a few

basic nutritional parameters for those sponges eaten by hawksbills and

for a few representatives of major non-prey orders. Although the diges-

tive physiology of the hawksbill remains unstudied, nutritional data on

its food are useful background in a discussion of feeding patterns.

Organic matter, energy, and nitrogen content of several sponge

species and genera represented in stomach contents of hawksbills are

given in Table 13, along with data for common reef-dwelling representa-

tives of major non-food orders. Sponges eaten by hawksbills vary widely

with respect to all of these parameters. The highest percentage of

organic matter was observed for Chondrilla nucula, a species that was

well represented in stomach contents. Geodia neptuni, Cinachyra

kuekenthali, and Spheciospongia vesparium are low in organic matter, and

this is reflected in their total dry weight energy and nitrogen con-

tents. This pattern can also be expected to hold true for the other

heavily silicified astrophorids in the diet, e.g., Ancorina, Myriastra,

and Ecionemia, and for the hadromerid Placospongia. Total dry weight

values, which include ash content, are perhaps of greatest relevance

from the standpoint of predators. When high-ash food items are


Table 12. Antibiotic activity of sponge species (denoted with an
asterisk) or genera that were represented in stomach contents of
Eretmochelys imbricata.

Sponge activity


Cinachyra cavernosa
*Spheciospongia vesparium

Geodia cydonium
*Chondrilla nucula

Tethya aurantium

Suberites domuncula

Placospongia decorticans

Cinachyra cavernosa
*Spheciospongia vesparium
Hymeniacidon sp.

Ancorina alata
Cinachyra n. sp.
Tethya aurantia
Hymeniacidon perleve

Chondrosia collectrix


No activity
No activity
No activity




Burkholder, 1968

Burkholder and Ruetzler,

Burkholder, 1973

Bergquist and Bedford,

Stierle and Faulkner,


44 r
0 "-4


V) '


r .c

* 2

co I u
40 .

J-4 04
0 )

&-4 -4
0^ 0



-4 "-

) '1-

"W (

a S

ifl 1 4















s- Ju

C l










1-4 .



M ^0




coi i




* *lr

co CO

ca 4 4-4

C co 0 0

-4 0 a 0 3ca -4 4)
c c w 0 c c c 'I

w 0 m 1- u o C2
Q O CH U ] U *0 C0
a) 0 w -H-4 U -4 ".4
m u u o 0) wf 0 -, .0
r-4 41 .0 i w i A ol

0 (fl *' I
a0 c Q-
o)W In1

M oo oo
S00 0\



0 .4 0 c

S1Co-4a a
0 C < O C $O
C -H z w 01
30' C O ..C 0I 60
0.0 4- a 0 0

C0 w >C 0 CO
41 K *H + D
u 4 w
< (f 33



consumed, larger amounts must be eaten to obtain the same nutritional


Energy and nitrogen contents of the various sponges are more

homogeneous when put on an ash-free basis. Ash-free energy values of

prey species and genera are relatively low when compared to other animal

tissues (Paine, 1964). A very approximate estimation of crude protein

content of the sponges can be obtained by multiplying nitrogen content

by 6.25.


Composition of the Diet


In the digestive tract samples from hawksbills over 23 cm in

carapace length sponges were clearly the dominant food item. No dif-

ferences in the percentage of sponges were found for turtles of dif-

ferent sexes (except gravid females), sizes (over 23 cm), or geo-

graphic origins. Sponges were also the dominant food item in samples

from gravid females, although they contributed a smaller percentage to

the total digestive tract contents. The difference was made up

largely by substrate material.

The high percentage of sponges in the diet and the high degree of

homogeneity among samples from turtles of different sizes, sexes and

origins provide strong evidence that the hawksbill is a strict spongi-

vore. No other food category contributed significantly to the

samples; much of the non-sponge material was apparently ingested acci-

dentally along with the sponges. The only vertebrates known to have

comparable diets in terms of percent sponge are the gray angelfish

(Holacanthus arcuatus, 98.3% sponge, N = 6, Hobson, 1974), the queen

angelfish (Holacanthus ciliaris, 96.8% sponge, N = 24, Randall and

Hartman, 1968) and the rocky beauty (Holacanthus tricolor, 97.1%

sponge, N = 24, Randall and Hartman, 1968).



The fact that sponges were dominant in samples of such wide geo-

graphic origin (7 countries, 19 localities) suggests that spongivory

in hawksbills is not a parochial tendency but a widespread feeding

habit. Spongivory is such a peculiarly specialized feeding habit that

it seems unlikely that it would occur in only a portion of any given


Table 14 lists all records of sponge-feeding by Eretmochelys that

have been reported in the literature, received by me through personal

communications or compiled in the present study. The table documents

the fact that sponges are eaten by hawksbills, at least to some

degree, throughout the range of the species. Without more quantita-

tive data, one cannot say that the hawksbill feeds primarily on

sponges throughout its range. This will probably prove to be the case,

however, when adequate samples are available.

Other Elements of the Diet

The presence of substrate material in the samples can in most

cases be attributed to incidental ingestion. The percentage of this

item in the samples varied little (standard deviation 3.2) except in

gravid females, and this is consistent with the hypothesis that sub-

strate material enters the diet incidentally.

The high levels of substrate material observed in several of the

gravid females that had not been feeding are more difficult to

explain. They might be a consequence of retention in inactive

digestive tracts or of purposeful ingestion. Several other reptiles,

including other turtles, crocodiles, and lizards, are known to ingest

sediment purposely (Sckol, 1971). The purported adaptive aims of























4 0







*M CO 0

C4'1 CO

r4 a0

e-40 )
4 0>



4-W 40
0 0L


0C 0


0 "



C o



0 4










*0 CO C
\o ce

C L|
c u M4-
4- 0.

W0 4

CO 4-4
0 Co

0. 0
W (



C e

00 0










3 C

u c
(O 0.

ao o




















-4 c
-4 cc



o 0
Q. o




.-4 04



















e U










0 U



0 5
o )





" u
10 -4


> 0


41 0. r
C 54.J

wo o
O0 0

0 oo o







>0 0


-4 /\

i- 0
m ,D


u .
0 C-







e co

4- 0


U, -


-** C
r-4 -





* *-
C -O



lithophagy or geophagy, as it is called, are varied. In crocodiles,

stones in the digestive tract have been proposed to have gastrolithic

and hydrostatic functions (Cott, 1961; Webb et al., 1982). In the

present case a gastrolithic function would not appear to be applicable

because the animals that had the largest amount of substrate material

in their digestive tracts were not actively feeding.

Marlow and Tollestrup (1982) reported that female desert

tortoises (Gopherus agassizi) actively seek out and eat calcium-rich

deposits of earth during the nesting season and suggested that this

activity served to replenish calcium reserves depleted by egg shell

production. Calcium requirements are undoubtedly high in

Eretmochelys. According to Pritchard (1979a) Atlantic hawksbills have

the largest average clutch size of any turtle (about 150 eggs). The

fact that the amount of sediment in two reproductively active females

was comparable to or higher than the maximum amount found in other

turtles, including those with full digestive tracts, offers support to

the hypothesis of purposeful ingestion. Perhaps more convincing

evidence are the observations of a number of turtle fishermen and

commercial divers in the Leeward Islands and Panama who reported to me

that hawksbills feed on coral rubble, gravel and even Millepora coral.

Algae were a minor component of the samples and, in most cases,

can be considered to have been ingested accidentally. Codium

isthmocladum and Lobophora variegata were the only species that

appeared to have been ingested purposefully. Codium was mentioned as

one of the two genera of algae found in the stomachs of two hawksbills

captured in the Central Visayas, Philippines (Alcala, 1980). Codium

is a major dietary component of Hawaiian green turtles (Balazs, 1980).


Although algae were of little importance in the samples examined

in the present study, their possible role in the diet of some

hawksbill populations cannot be dismissed. Algae are mentioned as

having been found in digestive tract contents of hawksbills at locali-

ties in the Atlantic (Carr and Stancyk, 1975; Den Hartog, 1980;

Bjorndal, in press), Pacific (Swinhoe, 1863; Pritchard, 1977, 1979b;

Limpus, 1979; Alcala, 1980) and Indian oceans (Fryer, 1911; Hornell,

1927; Deraniyagala, 1939; Hirth and Carr, 1970).

Few of the above authors reported the amount or relative impor-

tance of this food item in their samples. Hirth and Carr (1970) and

Den Hartog (1980) found only small amounts of algae in specimens they

examined. Swinhoe (1863), Hornell (1927), Carr (1952) and

Deraniyagala (1939) stated or implied that algae were important com-

ponents in samples examined by them. Hornell (1927) provided the most

detailed information, stating that the stomachs of adult hawksbills in

Seychelles waters were repeatedly found to be full of masses of

sargasso weed (Sargassum) in various stages of digestion.

Deraniyagala (1939) reported that the hawksbill frequently subsists on

an entirely vegetarian diet, although he cited data on only one

specimen. Swinhoe (1863), too, had examined only a single specimen.

The question whether hawksbills can digest algae has been raised

by Den Hartog (1980). He noted that algae found in the digestive

tract of a specimen examined by him seemed poorly digested. Observa-

tions made during the present study are consistent with those of Den

Hartog. Algae appeared relatively unaltered by digestive processes

all along the tract. In Hornell's observations in the Seychelles, he

mentions finding Sargassum in varying states of digestion. It is


significant that Sargassum is the genus that was found in the

digestive tracts of very small turtles of lost-year sizes.

The small amount of seagrasses in the samples, together with the

minor importance of algae, are clear evidence of different food

requirements of the hawksbill and green turtle at the various study

sites. These two species are very commonly found in close association

in coastal waters in the study area and elsewhere in the Caribbean.

Throughout the world the green turtle is known to be a rather strict

herbivore (Mortimer, 1982). In the Caribbean green turtles feed

primarily on the seagrass Thalassia testudinum (Bjorndal, 1980;

Mortimer, 1981). Their feeding habits in the Lesser Antilles have yet

to be studied, but the herbivorous feeding preference of the species

is widely established. Limited evidence gathered during the present

study suggests that immature green turtles at some localities in the

Lesser Antilles consume appreciable quantities of algae, as well as

seagrass. In neither case, however, do they appear to be in competi-

tion with hawksbills for food.

There are few records in the literature of hawksbills feeding on

seagrasses. Alcala (1980) mentions the presence of seagrass in the

stomachs of two specimens from the Central Visayas, Philippines.

Their abundance in the samples is not reported. Seagrasses were also

reported in the diet of hawksbills in the Eastern Caroline Islands,

Micronesia (Pritchard, 1977).

Lost-Year Turtles

A significant gap exists in knowledge of the life history of all

sea turtles from the time newly emerged hatchlings leave the nesting


beach to the time they appear in the foraging habitats characteristic

of subadults and adults. Marine turtles of all species are rarely

sighted during this period, and this has led biologists to call this

stage of the life history the lost year (Carr, 1967). The length of

the lost-year interlude and the sizes at which turtles of various

species enter coastal habitats have yet to be established. In the

Lesser Antilles, where much of the present study was carried out,

hawksbills less than 23 or 24 cm carapace length are rarely sighted.

Interviews with turtle fishermen and commercial divers during the

course of field work yielded information on only one or two specimens

of this size range.

There is considerable evidence that small turtles of at least

some species spend the lost year in the open sea (Carr, 1967; Carr and

Meylan, 1980). In the Atlantic Ocean, green turtles and loggerheads

have repeatedly been found drifting in association with rafts formed

by the floating alga Sargassum (Carr and Meylan, 1980; Carr, 1983).

There is little evidence, however, linking post-hatchling hawksbills

to this habitat. Only a few notes in the literature refer specifi-

cally to lost-year hawksbills. Hornell (1927) reported an observation

made by L. E. Lanier of hawksbills drifting in association with masses

of seaweed many miles from land. Vaughan (1981) reported that

hatchling-sized and slightly larger turtles are frequently found in

the deep sea associated with long skeins of rubbish and seaweed

downcurrent from a major hawksbill nesting beach in the Solomon

Islands. Whether these were hawksbills could not be verified,

although this seems likely.


Data collected by Kajihara and Uchida (1974) on the carapace

lengths of 146 hawksbills caught for the taxidermy trade in Southeast

Asia offer some of the most convincing evidence ever presented for the

existence and length of the lost-year period for hawksbills. In spite

of intensive economic incentive for fishermen to supply the taxidermy

trade, no turtles under 15 cm carapace length and only a few in the

15-20 cm range were found in the factory. The authors suggested that

a change in habitat occupation takes place at approximately 16-18 cm

carapace length.

An alternative solution to the lost-year puzzle for hawksbills is

offered by Witzell and Banner (1980), who reported that at least some

post-hatchling hawksbills (> 4 cm) inhabit coral reefs in Western


The contents of the digestive tracts of four hawksbills reported

here provide corroboration of the theory that the lost year is spent

associated with Sargassum rafts, although caution must be taken in

interpreting data from stranded specimens. The possibility exists

that atypical foods were consumed subsequent to the injury or onset of

disease that resulted in death. The food sample from UF 54846 can

probably be considered free of this bias because death was almost

certainly due to asphyxiation by tar. Food present in the digestive

tract was therefore consumed beforehand, and can be assumed to be

characteristic of the normal diet.

Sargassum was present in all four specimens, although in only two

cases was the material identified as one of the pelagic species of the

genus that is known to form large floating mats. Fish eggs of the

suborder Exocoetoidei were attached to Sargassum in UF 54846. This


suborder includes flying fish, half-beaks and needlefish; most of the

species within it are known to be pelagic. The presence of these eggs

in the digestive tract is evidence of surface feeding, in any case, as

is that of bouyant styrofoam particles and plastic beads.

The relative importance of plant and animal matter is difficult

to assess with the limited sample. Both were well represented.

Sargassum was present in sufficient quantity to suggest purposeful

ingestion. Norris and Fenical (1982) discuss the apparent avoidance

of Sargassum by many herbivores in the Caribbean and suggest that the

presence of tannin-like polyphenolic substances within members of the

family Sargassaceae may be responsible. In a wide survey of the

feeding habits of West Indian fish, Randall (1967) found that rela-

tively few fish feed on drifting Sargassum, sea chubs and the trigger-

fish Melichthys being notable exceptions.

The abundance and diversity of man-made debris in the digestive

tract contents reveal the vulnerability of marine turtles-at least at

this life history stage--to oceanic pollution. All four specimens

examined had plastic refuse in the digestive tract; some had several

different types. Of the many oceanic pollutants, petroleum products

undoubtedly represent the greatest threat to survival. Death of at

least one, and probably two, of the specimens can be attributed with

some confidence to this cause. Two were fouled externally, and three

had tar present in the digestive tract. The esophagus of UF 50027 was

heavily coated, and tar aggregates were present throughout the

digestive tract.

The presence of oceanic pollutants in the digestive tracts of the

turtles may be a result of their association with the Sargassum raft


community. Pollutants such as oil, styrofoam and other plastics are

well known components of the rafts. Their presence there has been

identified by Carr (1983) as a potential threat to marine turtles of

lost-year size.

Feeding Selectivity

The sponge diet of Eretmochelys, as indicated by the samples, is

restricted to a relatively few taxa. Sponges belonging to the orders

Astrophorida and Hadromerida represented 97.6% of the dry weight of

all identified sponges. The order Spirophorida, which represented an

additional 1.15%, is considered by Wiedenmayer (1977) to be a suborder

within the Astrophorida. These represent three of the five orders of

the subclass Tetractinomorpha; the two not represented in the samples

are the Desmophorida, a group with a stony composition, and the

Axinellida, which includes several reef-dwelling sponges. The

remaining sponges, all ceractinomorphs, represented 1.25% of the

sponges identified.

That the Astrophorida, Spirophorida, and Hadromerida make up a

relatively small part of the Caribbean sponge fauna is evidence of

strong selectivity by foraging hawksbills. Figure 9 shows the com-

position of the sponge faunas at four localities in the Caribbean.

The number of species within each order present at each locality is

indicated. Slightly different classification schemes are employed by

the various authors. The order Choristida, used in the figure, is

synonymous with Astrophorida in the classification system of Levi

(1973), which has been employed in the present study. For comparison,

orders that include astrophorid and hadromerid genera (as defined by

c u

0 ri





w a
c ^

L e


o- C:



0 0

(3) -4


C -

*^ 0

C -*


Co 0

a a

0 0

0 C

L iL


0S C

*0- 4-4(

C:1 t-e 0-

(0 r-.

0 '

I 4-

4.J Mc

w 41

co 0^c

to c



01 a^t..4

m~J 0
41 95 !
U (f
(U 01"
A > 01~
U -l0
ta *>*
S o
0 ;

4J i o




VOISVyOdid3 a |


S VOia137SO1O3Od



co 0 a v S I t

VllSieOdids NI

valIdi3aiOV a o



l QIta3Ios O1330d

vaIO 1soidVH

I VsoIVa3i






T-j I
& z




vaYOisIOMO r

VaIsvItodid3 i


0 vo iu31osoo03od










Levi, 1973) that were represented in the stomach contents of

hawksbills are marked with stippling. An average of 22% (range 13-27)

of the total number of species represented at each locality is

included in the stippled columns. This is an overestimate of the

percentage of the fauna represented in the digestive tract samples of

hawksbills, because not all genera or species within prey orders were

consumed. No comparable data have been published on the composition

of sponge faunas in the western Caribbean.

Another measure of feeding selectivity can be obtained by com-

paring Figure 9 to Figure 5. The latter shows the ordinal composition

of the sponges found in stomach contents. Hadromerids and

astrophorids represent less than a quarter of the fauna, and yet they

constitute 97.6% of the total dry weight of all sponges identified in

stomach contents.

All sponges included in Figure 9 are considered shallow-water

sponges. In the studies of Hechtel (1965), Wiedenmayer (1977) and

Cambiaso (1981), sponges were collected by diving with snorkel or

SCUBA gear. De Laubenfels' (1936) survey additionally included

dredged specimens, but only those collected in water less than 50 m

deep have been used in the figure. Considering the diving capacity of

the hawksbill (individuals have been sighted at 80 m, Frazier, 1971),

nearly all of these sponges would potentially be available as food.

The order Keratosa (= Dictyoceratida plus Dendroceratida, Levi,

1973) was not represented in stomach contents except for a few small

fragments of the dendroceratid Halisarca. A small number of fragments

of keratose sponges were also seen in intestinal contents. This is

a large group, and as shown in Figure 9, one that is very well


represented in Caribbean sponge faunas. Van Soest (1978) listed 52

well established species (21 genera) of Keratosa in the West Indies.

Of the 33 species described in his study, 18 (10 genera) preferred

reef habitats.

The order Haplosclerida is another large group that was nearly

absent from the samples. Van Soest (1980) listed 62 West Indian

haplosclerids. Sixteen species (7 genera) of the 36 included in his

study were described as preferring reef habitats. Fragments of

Callyspongia and Cribochalina (see Table 4) were the only material

representing this large order.

The order Poecilosclerida, which also includes reef-dwelling

species, constituted only 0.63% of all identified sponges in the

stomach contents. No axinellids were represented. In the survey of

De Laubenfels (1936), the order Axinellida is treated as part of the

Halichondrida (see Figure 9).

The sponge diet of the hawksbill as reflected by the samples is

also restricted in terms of the number of genera and species repre-

sented. Only 22 genera (31 species) were identified in the stomach

contents of all turtles from all localities. Ten species accounted

for 87.4% of the total dry weight of all identified sponges. The

cumulative total of shallow-water demosponges present at the col-

lecting localities is unknown, but is certain to be well over one

hundred. De Laubenfels (1950b) listed 115 species from the West

Indies (excluding Bermuda). Over a hundred species of sponges occur

on the fore reef slope at one locality in Jamaica (Reiswig, 1973).

Feeding selectivity is also indicated by the high degree of

similarity in the sponge composition of digestive tract samples from


the widely separate geographic localities (Table 15). Many genera

were represented in all regions by the same species. Myriastra,

however, was represented by different species (a total of 6) at each

of three localities: Panama (1); Carriacou (2); and the Leeward

Islands (3). Both of the buccal cavity samples from live hawksbills

at Palm Beach, Florida, were Geodia. The fecal pellet from the

juvenile hawksbill captured in the Caicos Islands consisted entirely

of Chondrilla nucula.

In assessing the actual biomass represented by prey species, both

frequency of occurrence and size must be considered. A few prey

sponges--e.g., Spheciospongia vesparium, Chondrilla nucula, and

Geodia--are considered common. Ruetzler and Macintyre (1978) listed

S. vesparium and G. neptuni among the ten most common siliceous

sponges at Carrie Bow Cay, Belize. Spheciospongia is also abundantly

represented on Jamaican reefs (Reiswig, 1973). Both of these genera

are also very large. S. vesparium was reported by De Laubenfels

(1936) to be the largest representative of the phylum Porifera,

although data by Dayton et al. (1974) suggest that this species may be

rivaled in size by some species of Antarctic hexactinellids.

Specimens of G. neptuni a meter in diameter have been observed

(Wiedenmayer, 1977).

Other genera in the samples--e.g., Ancorina, Ecionemia,

Myriastra, and Placospongia--are poorly represented in faunal lists of

Caribbean sponges (De Laubenfels, 1936, 1950a; Hechtel, 1965;

Wiedenmayer, 1977; Cambiaso, 1981) and are considered relatively

uncommon by some sponge biologists working in the Caribbean and on the

Florida reef tract (S. Pomponi, pers. comm.; G. Schmahl, pers. comm.).



0 0
S 41






3 &



c5 0

cc 11
E-i Q ,

c a0
0 0 w m
0 *3 *
0 u
3~l m 1^ n i

C: (0 0. ". 0 0 C1

c n 0 ( 0 ( 0
-4 0 o 40 4 aI P4
C O W 0 0) $.' O C 4) u

U <- OU 0 m W( 0 U 0

X$ X > X

u m

i u







The reason for their poor representation, or at least low apparency,

in known faunas is not clear. The sponges may simply have been

overlooked in these surveys because of sparse distribution or incon-

spicuousness, in which case their abundance in the digestive tract

samples is due solely to the feeding selectivity of the hawksbill.

Another possible explanation is that there may be greater between-site

variability in sponge faunas than is currently recognized. If this is

the case, these sponges may prove to be more common when additional

faunas are studied. Their poor representation may also be due to

other sampling artifacts; they may occur in microhabitats that are

difficult to sample, such as caves and ledges, or at depths beyond

those normally sampled. Some prey genera are definitely known to

occur in deep water. Placospongia has been dredged from a depth of 70

m in the Florida Keys (De Laubenfels, 1936). Based on West Indian

records De Laubenfels (1950b) listed Placospongia and three species of

other prey genera (Ancorina, Myriastra, and Cinachyra) as deep water

(> 50 m) species. The distribution and abundance of prey sponges

clearly deserve further study.

Narrowness and specificity of the diet of the hawksbill are sup-

ported by data from other investigators. A mature male hawksbill

examined by Carr et al. (1966) at Tortuguero, Costa Rica, contained

only large amounts of Geodia gibberosa. A second specimen also con-

tained this sponge, as well as other invertebrates. In a later study

at this same locality Carr and Stancyk (1975) reported that Geodia

gibberosa was one of the two most important components in stomach

contents of 20 hawksbills. The tunicate Styela was the other. G.

gibberosa was present in 90% of the turtles they examined. The only


other sponges represented in more than 5% of the turtles were uniden-

tified choristids (= astrophorids), which were present in 25% of the

animals. Chondrilla nucula was one of the species they identified.

When the data of Carr and Stancyk (1975, p. 164) are considered

according to Levi's (1973) classification, all prey sponges are

hadromerids or astrophorids, except for one poecilosclerid identified

from a single turtle. In reexamining the material on which Carr and

Stancyk's (1975) paper is based, I found large pieces of the same

species of Suberites as identified in the present study, as well as

fragments of Placospongia.

Additional data on the species of sponges eaten by hawksbills are

available for several of the reports listed in Table 14. The juvenile

captured at St. Thomas, U.S. Virgin Islands, had been feeding on

Chondrilla nucula (W. Rainey, pers. comm.). The digestive tract of

the 61 cm individual captured at Andros Island in the Bahamas was

filled with Chondrilla nucula, Geodia neptuni, and Polymastia sp. (W.

Rainey, pers. comm.). Polymastia is a hadromerid. Chondrilla nucula

was also identified from the juvenile captured at La Parguera, Puerto

Rico (Erdman, unpub. ms.). Hawksbills have been reported to feed on

clionid sponges at Carriacou, Grenada (M. Goodwin, pers. comm.).

Clionids are hadromerids. It is notable that so many reports have

identified the same orders, and in some cases the same species, as

those found in the present study. The reports encompass a wide

geographic range in the Caribbean--Costa Rica, the U.S. Virgin

Islands, the Bahamas, Puerto Rico, and Carriacou, Grenada.

Two accounts in the literature report feeding on sponges other

than hadromerids and astrophorids. Hawksbills at Ascension Island


were reported to eat the keratose sponge Ircinia. This identification

was apparently based on a description, rather than examination of

specimens, and deserves further study (A. Carr, pers. comm.). Two

species of sponges found in the digestive tract of a hawksbill

captured in the eastern Atlantic were not identified, but based on the

descriptions given (Den Hartog, 1980) are clearly not hadromerids or

astrophorids. One was a keratose sponge. Species identifications are

not available for any of the other reports listed in Table 14.

Role of Feeding Deterrents

The selective feeding habits of the hawksbill indicate that not

all sponges are acceptable as food. One aspect of the present study

was to investigate whether patterns in the diet were correlated with

the presence or absence in prey sponges of feeding deterrents such as

siliceous spicules, tough organic fibers, or secondary metabolites.

Inorganic Constituents

The large amount of silica present in important prey sponges and

the wide variation in silica content among the various prey species

suggest that siliceous spicules do not influence feeding patterns of

hawksbills. This conclusion is supported by data on the geometry and

placement of spicules in prey sponges. The large size and hook-like

shapes of spicules, and their concentration in thick, stony cortices,

are characteristics that should confer maximum deterrent effects. The

fact that sponges with spicules having these attributes are major com-

ponents in the digestive tract samples is evidence that spicules are

ineffective in deterring predation by hawksbills. Non-prey orders of


demosponges, by contrast, tend to have lower spicule contents, and

smaller and geometrically more simple spicules. In non-prey orders

there is no equivalent of the stony cortices of Geodia and

Placospongia. Ash content, which is a measure of total mineral con-

tent, closely parallels silica content in prey sponges. Because it,

too, is a measure of mechanical strength in sponges, wide variation

and high values for this parameter support the same conclusions.

Despite widespread acceptance of a defensive role for spicules in

sponges, previous studies have also revealed little evidence that high

spicule content or ash content in sponges deters predators. Randall

and Hartmann (1968) noted that two of the sponges most frequently

consumed by West Indian fish had a low spicule content, but they found

no correlation between spicule content and frequency of occurrence in

the diet among the next 20 most common species. Nine species of

astrophorids, including Geodia gibberosa, were among the 70 sponges

they identified. A high ratio of ash to organic matter is character-

istic of hexactinellid sponges, which are a regular dietary component

of asteroid and nudibranch predators at McMurdo Sound, Antarctica

(Dayton et al., 1974).

Ash contents of intestinal samples from hawksbills provide a

crude estimate of the percentage of silica in the digesta. Micro-

scopic examination of the material before ashing confirmed its

siliceous composition (see Figure 7). Ash constituted 92.0%, 76.6%,

and 74.3% of the dry weight of three samples. Using 50% as a con-

servative estimate of the percentage of silica in digesta throughout

the digestive tract, it can be calculated that as much as 557 g of

silica are present at one time in an actively feeding adult turtle.


With the exception of other strict spongivores, few animals have

a comparable diet in terms of silica. Silica is a prominent

structural component in a few groups of algae (notably diatoms),

protozoans (sarcodines, radiolarians), and plants (grasses and

cereals--Poaceae, sedges--Cyperaceae, and scouring rushes--Equi-

setaceae). In few, if any, of these groups, however, is silica

content comparable to that in sponges. In scouring rushes and rice,

which are considered to be among the most heavily silicified plants,

silica accounts for only 20% of dry weight (Kaufman et al., 1981).

Silica in grasses is often contained in projecting hairs or trichomes.

It is considered to act as a feeding deterrent to herbivorous range

animals. Diatoms are notably high in silica. Silica content of

frustules of some species is as high as 72% of dry weight (Volcani,

1981). The percentage of silica on a whole weight basis was not

given. It would be interesting to determine the silica content of

digesta of fish or microcarnivores that feed on diatoms.

The abrasive quality of the digesta of hawksbills deserves

discussion. Gut contents could not be handled without gloves and

tools. Spicules easily pierce human skin and cause painful reactions.

It is not clear how material of this abrasive nature is passed through

the tract without causing mechanical damage to the intestinal

epithelia. Scanning electron micrographs reveal that the tips of

spicules do become embedded in the tissues (Figure 8).

The extent to which spicules cause mechanical damage in

spongivores has never been investigated. Forrest (1950) reported that

spicules often pierce the stomach wall of the nudibranch Archidoris

pseudoargus. Bloom (1976, 1981) correlated the presence of spicule-

compacting organs in some species of spongivorous nudibranchs with the


consumption of "non-reticulate" sponges, i.e., sponges in which the

spicules are not bound by spongin. The sponges consumed by hawksbills

are of this type. Nudibranchs that feed on reticulate sponge prey

were found to lack spicule-compacting organs, but showed other

morphological adaptations, such as large radular teeth and muscular

intestines (Bloom, 1976, 1981). These characteristics were judged to

facilitate the handling of sponges containing spongin. I found no

evidence in hawksbills of gross morphological adaptations for handling

spicules. Large numbers of spicules were free throughout the large


Copious mucus production by nudibranchs has been proposed as a

physiological mechanism for handling abrasive sponges in the diet

(Forrest, 1953). The sponge food of some dorid nudibranchs is

liberally coated with mucus produced by glands of the digestive tract

(Forrest, 1953; Fournier, 1969). Randall (1963) observed a thick coat

of mucus on sponges in the stomachs of angelfishes and proposed a

similar function. Mucus production by hawksbills was not addressed in

this study. Mucus present in the digestive tracts would have been

likely to have been destroyed by preservatives before the digesta were

examined. In the digestive tracts of the few turtles that I examined

immediately after they had been killed by fishermen mucus was not

conspicuous. The turtles were all gravid females, however, and may

not be representative because of low feeding rates.

Organic Constituents

Spongin (the spongin B of Gross et al., 1956) is a type of

collagen unique to sponges. It forms the macroscopic organic skeleton


of many species and is a component of a number of specialized

structures. Spongin is the organic constituent of sponges most often

implicated as a feeding deterrent. It can constitute a large per-

centage of the volume and dry weight of a sponge (e.g., 48.2% of the

dry weight of Mycale acerata, Dayton et al., 1974). The spongin

content of some keratose sponges may be even higher.

One of the highest correlations found between patterns in the

diet of hawksbills and assumed feeding deterrents involved spongin.

With the exception of the small group Homosclerophorida, the orders

Astrophorida, Spirophorida, and Hadromerida are the only edible (non-

stony) demosponges that lack this skeletal constituent. Spongin

fibers are present in all other sponges, and in many, form extensive

skeletons, either alone or in combination with inorganic elements. In

the skeletons of axinellids, poecilosclerids, haploscierids, and

halichondrids, spongin is usually associated with silica. Sponges of

the Dictyoceratida and Dendroceratida, the keratose sponges, contain

no spicules, but instead have highly developed fiber skeletons. The

fibers in these two orders are either homogeneous, cored with a

medullary substance, or impregnated with foreign bodies such as sand

grains, exochthonous sponge spicules, or even radiolarian and

foraminiferan skeletons. These fibers, as well as the spongin

filaments of Ircinia, also contain iron deposits in the form of

lepidocrocite (Towe and Ruetzler, 1968). Iron can constitute as much

as 5.5% of the dry weight of the fiber (Junqua et al., 1974). The

functional significance of this mineralization is unknown, but it can

be speculated that iron adds structural rigidity to the fibers and

thus enhances their defensive utility.


Spongin, with its various reinforcements, provides strength and

elasticity to a sponge (Levi, 1973), but the apparent avoidance of it

by hawksbills is difficult to explain on the basis of mechanical

deterrence. Hawksbills have very powerful jaws, as evidenced by their

ability to feed on heavily silicifed sponges such as Geodia and

Plascospongia, and on very rubbery, cartilagenous species like

Chondrosia. The jaws of hawksbills are certainly more powerful than

those of the various angelfishes known to feed on fibrous sponges,

such as Callyspongia (Randall and Hartman, 1968). In any case, one

would expect some predation on sponges with weak spongin fiber

development, but this is not the case.

One of the unusual properties of spongin that may be relevant to

the present discussion is its resistance to enzymatic hydrolysis

(Gross et al., 1956; Junqua et al., 1974). Spongin fibers have been

found to be resistant to diverse bacterial collagenases and other

proteolytic enzymes, and to mild acid or alkaline hydrolysis (Garrone,

1978). The fact that spongin is affected by cuprammonium hydroxide-a

reagent that dissolves cellulose--has led to speculation that there

are molecular interactions in spongin that are comparable to those

binding polysaccharide chains in cellulose (Garrone, 1978). Whether

or not spongin is digestible by hawksbills is not known. Even if one

assumes that it is not, this would not satisfactorily explain why it

is not eaten. Several of the sponges consumed by hawksbills contain

high levels of silica, which is totally indigestible. There is

circumstantial evidence for the avoidance of spongin by other sponge

predators. Both asteroid echinoderms (Dayton et al., 1974) and dorid


nudibranch mollusks (Garrone, pers. comm.) have been observed to eat

around the spongin fibers.

Feeding patterns of hawksbills also show correlation with the

collagen fibril content of sponges. The types of sponges that were

found in the digestive tract contents are rich in collagen fibrils.

Collagen fibrils (the spongin A of Gross et al., 1956) are a struc-

tural form of collagen visible only with the electron microscope. The

fibrils are similar, if not identical, to those found in connective

tissue throughout the animal kingdom (Bairati, 1972). Although

universally present in the phylum Porifera, the fibrils vary in

density in the interstitial stroma of various species (Garrone, 1978;

Wilkinson, 1979).

A high collagen fibril content imparts a dense, rubbery con-

sistency to a sponge. This is particularly apparent in species that

contain little or no silica, such as Chondrilla or Chondrosia. This

consistency could conceivably serve as a mechanical feeding deterrent

to some predators, but does not appear to discourage predation by


A high collagen fibril content in sponges may represent a posi-

tive attribute from a predator's standpoint because of the nutritional

value they impart. The fibrils have been found to be among the most

highly glycosylated in the animal kingdom (Garrone, 1978). Carbo-

hydrates were found to constitute 15% of the ash-free dry weight of

collagen fibrils of Spongia graminea (Gross et al., 1958) and 10% of

the weight of fibrils of Ircinia variabilis (Junqua et al., 1974).

Data on the amino acid composition, nitrogen content, and carbohydrate

content of fibrils of various sponge species are given by Gross et al.


(1956), Gross et al. (1958), Piez and Gross (1959), Junqua et al.

(1974), and Garrone et al. (1975).

The nutritional value of sponge fibrils is dependent, however, on

their being digestible. Although they are structurally and bio-

chemically indistinguishable from those found in the rest of the

animal kingdom, they have the unique property of being resistant to

enzymatic hydrolysis (Garrone, 1978). The fibrils are unaffected by

collagenases of various origins and other proteolytic enzymes

(Garrone, 1978). Whether hawksbills are capable of digesting this

form of collagen is not known.

Carbohydrate-rich compounds glycoproteinss and acid mucopolysac-

charides) that are associated with the fibrils (Thiney and Garrone,

1970) may represent a more substantial and accessible source of

nutrition than the fibrils themselves. Various studies of the inter-

cellular matrix have revealed the presence of uronic acid,

hexosamines, acid polysaccharides, glycoproteins, and several sugars,

such as glucose, galactose, mannose, xylose, fucose, and arabinose

(Garrone, 1978). Although these compounds have been isolated from

sponges of diverse taxonomic groups--not all of which can be con-

sidered rich in collagen fibrils--some are known to be intimately

linked to the fibrils, and thus would impart additional nutritional

value to fibril-rich sponges.

Lack of knowledge of the nutritional requirements and digestive

capabilities of hawksbills makes it difficult to speculate further on

the significance of the patterns observed in the collagen composition

of sponges in the diet.


Chemical Constituents

Sponges have long been known to produce irritating and odorous

chemicals. As a result of recent interest in marine natural products

chemistry, there has been a concerted effort to isolate and charac-

terize these compounds. Because sponges proved to be a rich source of

novel compounds--particularly ones with antibiotic activity--they have

become one of the best studied marine invertebrate phyla (for reviews

see Minale et al., 1976; Minale, 1978).

Several functions have been proposed for secondary metabolites in

sponges, including predator deterrence (Bakus and Green, 1974;

Bergquist, 1978; Fenical, 1981; Thompson et al., 1983); facilitation

of feeding by the sponge (Bergquist, 1978); inhibition of nonsymbiotic

bacteria (Thompson et al., 1983); and participation in allelochemical

interactions with other sedentary reef organisms (Jackson and Buss,

1975). Secondary metabolites are present in large amounts in sponges

(up to 13% of dry weight in Verongia aerophoba, De Rosa et al.,

1973a), and are known to be released into the surrounding sea water by

some species (Thompson et al., 1983). These two observations are

consistent with the hypothesis that metabolites serve to deter preda-

tion, although other functions are likewise supported.

There is abundant evidence that sponges have inhibitory, noxious

and sometimes lethal effects on other organisms. Sponge extracts in-

jected into laboratory rabbits, dogs, mice and fish cause hemor-

rhaging, hypertension, paralysis and death (Richet, 1906a,b; Halstead,

1965; Baslow, 1969). Brominated metabolites isolated from the sponge

Aplysina fistularis have been shown to inhibit feeding by fish

(Thompson et al., 1983). Fish that are force-fed sponges have been