MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Vol. 364: 87?95, 2008
doi: 10.3354/meps07504
Published July 29
INTRODUCTION
A rich literature describes the ecological and evolu-
tionary causes and consequences of herbivore diet
choice (Rosenthal & Berenbaum 1992). Rather than
being eaten alive, however, most plants senesce and
are processed via detrital food webs (Mann 1988,
Cebrian 1999). Detrital pathways infuse food webs
with plant production even in the relative absence of
herbivory, and influence nearly all ecosystems (Polis &
Strong 1996). Thus, the factors that drive feeding on
detritus rather than live tissues are pivotal to the flow
of energy and materials through food webs.
Salt marshes dominated by smooth cordgrass Spar-
tina alterniflora Loisel are among the world?s most
productive ecosystems and support ecologically and
economically important communities (Silliman & Bor-
tolus 2003). However, herbivores generally consume
much less than 10% of marsh plant production (Teal
1962, Silliman & Newell 2003), and most cordgrass
senesces and decays while still attached to the stem
(Newell 1993). The trophic importance of cordgrass
detritus has been debated for decades, with plant
detritus often considered too nutritionally inadequate
to be the sole food source supporting detritivore
production (Odum et al. 1979). However, benthic
macro-organisms such as shredder snails, amphipods,
and shrimps are particularly common on the marsh
surface and will eat dead cordgrass, at least in labo-
ratory settings (e.g. Rietsma et al. 1988, B?rlocher
& Newell 1994, Kneib et al. 1997, Pennings et al.
1998). Moreover, stable isotope evidence consistently
demonstrates the presence of cordgrass carbon in
nearshore food webs (Currin et al. 1995). Thus, cord-
? Inter-Research 2008 ? www.int-res.com*Email: parkerj@si.edu
A specialist detritivore links Spartina alterniflora
to salt marsh food webs
John D. Parker1, 2,*, Joseph P. Montoya1, Mark E. Hay1
1School of Biology, Georgia Institute of Technology, Atlanta, Georgia 30332, USA
2Present address: Smithsonian Environmental Research Center, 647 Contees Wharf Road, Edgewater, Maryland 21037, USA
ABSTRACT: Because most plant production is subject to senescence and is eventually consumed by
detritivores, the factors that drive detritivore diet choice are pivotal to the flow of energy and materi-
als through food webs. Here, we investigated the common salt marsh amphipod Gammarus palustris,
which is a habitat specialist that feeds specifically on the dead leaves of its living host plant, salt
marsh cordgrass Spartina alterniflora. Restricted use and consumption of dead S. alterniflora was
reinforced by superior amphipod performance (survival, size, and sexual development) on dead S.
alterniflora relative to other diets, and was driven at least in part by amphipods being physically able
to feed on soft, decaying plant tissues but not live, turgid tissues. Stable isotopes from field surveys
and laboratory assimilation assays suggest that amphipods also feed on S. alterniflora in the field, and
that the important marsh fish Fundulus hetroclitus feeds on amphipods. Thus, consumption of
G. palustris by F. heteroclitus may be an important trophic pathway linking cordgrass production to
nearshore food webs. Importantly, direct isotopic analyses of amphipods and their known food
sources demonstrated substantial deviation of observed fractionation factors from idealized stan-
dards. This suggests caution when using idealized trophic shifts to describe food web linkages, and a
renewed focus on assimilation assays to determine the realized fractionation of dietary isotopes.
KEY WORDS:  Spartina alterniflora ? Detritus ? Stable isotopes ? Food webs
Resale or republication not permitted without written consent of the publisher
Mar Ecol Prog Ser 364: 87?95, 2008
grass production is likely assimilated and passed up
the food chain. However, the pathways and mecha-
nisms linking cordgrass production to nearshore food
webs remain unclear, in part because few studies
explicitly link consumer diet choice to food web
structure (Winemiller et al. 2007).
In this study, we investigated the feeding prefer-
ences and trophic position of the salt marsh amphipod
Gammarus palustris, which is a common, although
visually cryptic, amphipod in salt marshes along the
eastern Atlantic coast of the United States, from New
Hampshire to Florida (Gable & Croker 1977). It is
found clustered in and around the interstices cre-
ated by the outer leaves, stems, and root masses of
Spartina alterniflora, with densities reported at up to
55 000 ind. m?2 (Van Dolah 1978). Previous studies
suggest that G. palustris is a generalized consumer of
macro- and epiphytic algae found on the marsh surface
(Gable & Croker 1977, Guarna & Borowsky 1993),
although direct observations of gut contents found a
?large amount of unidentifiable detritus? (Gable & Cro-
ker 1977, p. 129). Given its abundance, widespread
distribution, and close association with cordgrass, we
hypothesized that G. palustris could be a major link
between marsh primary production and higher trophic
levels if it feeds directly on decaying salt marsh cord-
grass. We asked the following questions: (1) What are
the dietary preferences of G. palustris? (2) What food
traits affect diet selection? (3) Does amphipod fitness
on different macrophyte diets reinforce patterns of
preference? (4) What do stable isotope analyses reveal
about the natural diet of G. palustris and of marsh food
web structure at our study site?
MATERIALS AND METHODS
Our study site was a salt marsh at the Georgia Insti-
tute of Technology?s marine lab on Skidaway Island,
Georgia, USA. Shorelines in this estuarine environ-
ment are dominated by smooth cordgrass Spartina
alterniflora interspersed with a fringing band of oyster
Crassostrea virginica reef in the middle intertidal, and
sparse cordgrass, mud, and sand sediments in the
lower intertidal. Tidal range varies from 2 to 4 m. Pre-
vious studies found the salt-marsh amphipod Gam-
marus palustris closely associated with S. alterniflora
(Gable & Croker 1977, van Dolah 1978); our findings at
this site were similar. During the 2 yr of these experi-
ments, repeated daytime and nighttime inspections
never revealed G. palustris in the upper intertidal, on
standing dead, heavily epiphytized cordgrass stems, or
in the water column. The amphipod was found only
within the leaf interstices of live S. alterniflora where
the older blades were senescing.
Amphipod feeding preferences. Previous studies
suggested that Gammarus palustris feeds primarily on
small epiphytes (Gable & Croker 1977) or macroalgae
(Guarna & Borowsky 1993), but not on cordgrass
despite its close association with this plant. To deter-
mine G. palustris feeding preferences among cord-
grass and other macrophytes common to our study site,
we conducted no-choice and multiple choice feeding
assays in the laboratory with 3 algal species (the green
alga Ulva sp. (henceforth Ulva), and 2 epiphytic red
algae common on cordgrass stems, Bostrychia radicans
[Montagne] Montagne and Caloglossa leprieurii
[Montagne] J. Agardh) and 4 types of cordgrass: the
tips of newly sprouted shoots, live green tissue from
the upper blades of older plants, yellowing, senescent
leaves, and dead, decomposing blades located near
the base of the stem.
For no-choice assays, we placed 2 amphipods into
each of 5 replicate 3.3 cm diameter cups with 50 ml of
filtered estuarine water and approximately 0.17 g of
plant tissue that had been pre-weighed after blotting.
Controls for changes in plant mass unrelated to amphi-
pod consumption consisted of identical portions from
the same plants placed into cups without amphipods.
Water was changed daily and all feces removed. After
92 h, each plant was blotted and re-weighed. We cal-
culated the mass of plant eaten by individual con-
sumers using: (Ti ? Cf/Ci) ? Tf, where Ti and Tf were ini-
tial and final wet masses of tissue exposed to
amphipods, and Ci and Cf were initial and final wet
masses of controls. Feeding on each diet was analyzed
with a paired t-test (treatment versus control); differ-
ences in consumption among diets were analyzed with
a 1-way analysis of variance (ANOVA) followed by
Tukey multiple comparison tests.
For multiple choice assays, 5 amphipods were
offered the 7 plant diets together in 13 replicate 8 cm
Petri dishes filled with 100 ml of filtered river water.
Controls were identical dishes without amphipods;
water was changed daily in all dishes. After approxi-
mately 92 h, individual plant portions from each dish
were blotted and re-weighed to estimate the mass of
plant eaten during the assay (calculated as above).
Results were analyzed via Friedman?s test followed by
multiple comparisons.
Nitrogen is often limiting for detritivores and herbi-
vores (Mattson 1980); thus we also tested whether
Gammarus palustris would ingest nitrogen-rich animal
matter if given the chance (e.g. via food falls on the
marsh surface). We conducted choice and no-choice
feeding assays according to the procedures outlined
above using lyophilized, powdered, and reconstituted
gel-like foods made from the following plant and ani-
mals common to salt marshes: dead cordgrass, body
tissue from the mud snail Ilyanassa obsoleta, the killi-
88
Parker et al.: Food web consequences of detrital specialization
fish (mummichog), Fundulus heteroclitus, and the
grass shrimp Palaemonetes vulgaris. Each gel-based
diet consisted of 0.5 g of powdered tissue poured
directly into a warm 4% agar solution, poured onto a
fiberglass window screen mesh (mesh size: 0.42 mm2),
and then pressed between 2 sheets of waxed paper.
This method allowed feeding rate to be quantified as
the number of window screen squares from which the
agar-based ?food? had been grazed over a given period
of time (Hay et al. 1998). No-choice assays were run
with 1 amphipod per 50 ml container; choice assays
were run with 3 amphipods per 350 ml container to
better capture significant feeding on the larger amount
of plant tissues presented in these assays. After 24 h
and no apparent feeding, we added 1 and 3 amphipods
to each no-choice and choice assay container, respec-
tively. Experiments were terminated after 66 h. Results
from no-choice and choice assays were analyzed with
a Kruskal-Wallis and a Friedman test, respectively,
followed by multiple comparisons.
Plant traits affecting feeding preferences. To assess
how variations in plant species toughness or texture
altered amphipod feeding preferences, we standard-
ized plant structure by freeze-drying each plant,
grinding it to a fine powder, incorporating this into a
gel-based food poured onto window screen, and offer-
ing these to amphipods in a no-choice assay. Assays
included Ulva, Caloglossa leprieurii, Bostrychia radi-
cans, and the 4 types of Spartina alterniflora (live,
senescent, dead, and shoots). Each replicate consisted
of 1 amphipod in a separate container. There were 10
independent replicates for each food type, and feeding
was measured after 92 h. If an amphipod died, that
replicate was excluded (thus, N = 8 to 10 for each treat-
ment). Before freeze-drying, plant toughness was
assessed using a penetrometer to determine the mass
required to pierce each plant. We also determined the
elemental composition (%N and %C) of plant or ani-
mal matter using a Carlo Erba NC 2500 elemental ana-
lyzer interfaced to a Micromass Optima mass spec-
trometer.
Amphipod performance on different diets. To
determine how amphipod performance on different
diets related to feeding preference, we measured sur-
vival, final size, and sexual maturity on sibling groups
of newborn amphipods reared on each of the 7 marsh
plant diets and on a starvation control. We obtained
offspring from 48 ovigerous females collected from
Priest Landing marsh on 15 and 17 June 2000. Each
mother was placed into an individual Petri dish with
filtered river water and 1 pellet of commercial fish
food. Water and food were changed every 2 to 3 d.
Newborn (<24 h old) amphipods were placed into
separate Petri dishes and started on 1 of the 8 treat-
ments.
Food preference in Gammarus palustris is under
genetic control (Guarna & Borowsky 1993). Thus, we
tested for familial differences in diet-related perfor-
mance by rearing the entire brood from each mother
on the same diet and by contrasting 3 to 6 families per
diet (resulting in 30 to 40 ind. per diet). Mothers were
preserved at death or after the molting process caused
the release of the last juvenile.
Petri dishes were checked daily to assess survivor-
ship. Water and dishes were changed every 3 to 5 d.
Food was replaced if there were signs of algal coloniza-
tion, when plant tissues senesced, or approximately
every 7 to 10 d, whichever came first. After 28 d, each
survivor was fixed in 5% formalin. Final sizes of all sur-
viving amphipods and mothers were determined by
photographing each amphipod under a microscope and
measuring the total length from the tip of the rostrum to
the tip of the telson using the segmented line feature in
Scion Image software. Surviving female amphipods
were considered sexually mature if we observed ripe
gonadal tissue through the body wall.
To assess whether amphipod survivorship differed
among diets, we performed a parametric survival analy-
sis of the number of days each amphipod persisted on
each diet, with data best fitting a lognormal distribution;
data were right censused to reflect that some animals did
not perish during the 28 d experiment. To assess which
diets led to higher or lower survivorship, we then per-
formed ?2 tests on survivorship at the end of the experi-
ment. The effect of diet on final size was analyzed with
1-way ANOVA followed by Tukey?s test. Maternal and
brood size effects on offspring final size were assessed
via linear regression. The proportion of females that
reached sexual maturity among diet treatments was an-
alyzed with a ?2 test. Familial differences in amphipod
survival were assessed with 1-way ANOVAs for each diet.
Stable isotope determination of marsh producers
and consumers. To determine the diet source of
amphipod production in situ and to get a general pic-
ture of the food web at our site, we determined the
mean values of ?15N and ?13C for sediments (mud and
sand), macrophyte producers (all diet types listed
above), and the following marsh consumers: the amphi-
pods Gammarus palustris and Uhlorchestia spartino-
phila (a sympatric cordgrass stem-dwelling amphipod
that specializes on fungal hyphae; Covi & Kneib 1995,
Kneib et al. 1997), the killifish Fundulus heteroclitus,
the snail Ilyanassa obsoleta, and the shrimp Palaemon-
etes pugio. Mud and sand samples provided a general
picture of the detrital and microalgal food sources
present on the marsh surface; the remaining samples
provided a picture of carbon flow from producers to
higher trophic levels. Initial samples were collected
from the field in year 2000, dried at 45?C, ground via
mortar and pestle, and acidified with dilute HCl to
89
Mar Ecol Prog Ser 364: 87?95, 2008
remove inorganic carbon. Subsamples (ca. 400 ?g)
were then weighed and packed into a tin capsule for
elemental and isotope analysis by continuous-flow
isotope ratio mass spectrometry (CF-IRMS) using a
Carlo Erba NC 2500 elemental analyzer interfaced to
a Micromass Optima mass spectrometer.
We also ran 2 experiments where amphipods were
fed a known diet from birth to examine trophic frac-
tionation factors used in food web studies (DeNiro &
Epstein 1978, 1981, Minagawa & Wada 1984, Vanderk-
lift & Ponsard 2003). First, we analyzed the isotope val-
ues of the 28 d old amphipods from the performance
experiment described earlier. However, these amphi-
pods had been initially fixed in formalin, and formalin
can alter ?13C values (Sarakinos et al. 2002). Thus, we
repeated the experiment at a smaller scale the follow-
ing year. For this experiment, we reared 8 or 9 amphi-
pods on each of 4 diets, including dead and senescent
Spartina alterniflora, mud, and sand for 29 to 36 d. At
the end of the experiment, each amphipod was pre-
served and processed as before but was never placed
in formalin. Surviving amphipods (N = 4 to 8) and their
diet sources were subjected to elemental and isotope
analyses as described above.
RESULTS
Amphipod feeding preferences
In no-choice feeding assays, amphipods could be
demonstrated to have fed on dead Spartina alterni-
flora, senescent S. alterniflora, and the epiphytic red
algae Bostrychia radicans and Caloglossa leprieurii
(Fig. 1), but not on the green alga Ulva or on live
S. alterniflora or its young shoots. When offered a
choice among salt marsh macrophytes, amphipods fed
on dead S. alterniflora in preference to all other plants
(Fig. 1). When offered animal versus plant tissues,
amphipods fed preferentially on dead S. alterniflora
relative to tissues from the marsh killifish Fundulus
heteroclitus, the grass shrimp Palaemonetes pugio, or
the mud snail Illyanassa obsoleta (Fig. 2).
Plant traits affecting feeding preferences
Dead cordgrass and Ulva were the softest macro-
phytes that could be tested with the penetrometer
(Table 1). The red alga Caloglossa leprieurii was softer
than the lowest mass our penetrometer could measure.
The filamentous red alga Bostrychia radicans could not
be tested because of its slender morphology, but it was
not overly ?tough? or rough to the touch. Live cordgrass
was the toughest tissue tested (Table 1). 
90
No choice
N = 5
p < 0.0001
0
2
4
6
8
*
*
*
*
C C
C
B
A,BA,B
A
Choice
N = 13
p = 0.0006
De
ad
 S
.a.
Se
n S
.a.
Liv
e S
.a.
Sh
oo
ts 
S.
a.
Bo
str
yc
hia Ul
va
Ca
log
los
sa
C
on
su
m
pt
io
n 
(m
g 
am
ph
ip
od
?1
 d
?1
)
0.0
0.5
1.0
1.5 A
B,C
B,C B,C
CC
B
Fig. 1. Gammarus palustris. No-choice and multiple choice
assays for amphipods feeding on 7 different macrophyte diets,
mean ? SE (S.a.: Spartina alterniflora; sen: senescent). Statis-
tics are from Kruskal-Wallis and Friedman tests, respectively.
Bars with the same letters were not significantly different in
post hoc multiple comparison tests across treatments within
each of the 2 separate assays. *Significant loss of plant mass
in treatments with amphipods (i.e. the amphipods fed on plant
material) vs. controls without amphipods (paired t-tests)
No choice
N = 7?9
p < 0.0001
N
o.
 s
qu
ar
es
 e
at
en
0
15
30
45
Choice
N = 10
p < 0.0001
0
15
30
45
A
B
BB
BB
B
A
De
ad
 S
.a.
Ki
llif
ish
Gr
as
s s
hr
im
p 
M
ud
 sn
ail
Fig. 2. Gammarus palustris. No-choice and multiple choice
assays for amphipods feeding on dead Spartina alterniflora
(S.a.) vs. animal tissue, mean ? SE. Statistics are from Krus-
kal-Wallis and Friedman tests, respectively. Bars with the
same letters were not significantly different in post hoc multi-
ple comparison tests across treatments within each assay
Parker et al.: Food web consequences of detrital specialization
Sand and mud sediments both contained relatively
little organic carbon relative to other food sources,
whereas animal tissues were the most nitrogen-rich
foods tested, containing roughly 4-fold more nitrogen
per dry mass than algal tissues, and 20-fold more nitro-
gen than sand, mud, and dead, senescent, and live
green Spartina alterniflora (Table 1). Cordgrass tissues
had C:N ratios ranging from 25 to 101, which is 47 to
594% higher than the maximum of 17 thought to sus-
tain animal nutrition (Russell-Hunter 1970).
Amphipods fed readily on all Spartina alterniflora
tissues when plant tissues were incorporated into
an agar matrix that destroyed their structural traits
(Fig. 3). However, feeding was still low on Bostrychia
radicans and Ulva.
Amphipod fitness on
different diets
Amphipod survival dif-
fered among diets (p <
0.0001, survival analysis),
and was highest on dead
Spartina alterniflora relative
to all other diets (p < 0.0001,
?2 tests; Fig. 4). Amphipods
reared on dead S. alterniflora
were also 19 to 39% larger
than amphipods reared on
other S. alterniflora diets (p <
0.001, Table 2), and there
was a trend toward more
females reaching sexual ma-
turity on dead S. alterniflora
versus alternative diets (p = 0.064, ?2 test; Table 2).
Maternal effects could not account for these differ-
ences. Although larger mothers had proportionately
larger broods (p < 0.0001, r2 = 0.354, linear regression),
there was no effect of maternal size (p = 0.936, r2 = 0.0,
linear regression) or brood size (p = 0.395, r2 = 0.26,
linear regression) on final offspring size. There were
significant familial differences in offspring lifespan
when amphipods were reared on live S. alterniflora,
the shoots of S. alterniflora, and on Ulva (Table 3), sug-
gesting heritable variation for amphipod performance
on these diets.
91
Table 1. Food quality parameters of potential diets. Data are mean ? SE (N); S.a.: Spartina
alterniflora and Sen: senescent. Different superscripts denote statistical differences among
means via 1-way ANOVAs followed by Tukey tests. Toughness was assessed using a pene-
trometer to determine the mass required to pierce each plant
Diet Toughness (g) % C % N C:N
Dead S.a. 4.81 ? 0.432 (5)a 26.6 ? 1.86 (14)c 0.428 ? 0.017 (14)a 62.15
Sen S.a. 48.3 ? 6.95 (5)c 31.2 ? 0.685 (15)c,d 0.309 ? 0.023 (15)a,d 101.0
Live S.a. 105.0 ? 6.77 (5)d 33.8 ? 0.583 (10)d 0.517 ? 0.030 (10)a 65.38
Shoots S.a. 34.3 ? 1.01 (5)b 34.5 ? 3.22 (7)d 1.37 ? 0.145 (7)a,c,f,g 25.18
Bostrychia Could not be tested 16.7 ? 1.12 (4)b,g 2.09 ? 0.095 (5)g 7.99
Ulva 5.25 ? 0.523 (5)a 24.5 ? 1.87 (9)c,e,f,g 2.83 ? 0.271 (9)b,f,g 8.66
Caloglossa <2.68 15.6 ? 0.092 (5)b 1.76 ? 0.078 (5)f,g 8.86
Fish Not tested 26.6 ? 1.04 (7)c,d 7.52 ? 0.433 (7)e 3.54
Shrimp Not tested 30.4 ? 0.496 (7)c,d,f 8.02 ? 0.171 (7)e 3.79
Snail Not tested 34.1 ? 1.08 (6)d,f,d 8.02 ? 0.694 (6)e 4.25
Sand Not tested 0.064 ? 0.0005 (2)a 0.011 ? 0.0006 (2)a 5.82
Mud Not tested 2.59 ? 0.174 (2)a 0.227 ? 0.002 (2)a 11.41
No choice
N = 8?10
p = 0.002
De
ad
 S
.a.
Se
n S
.a.
Liv
e S
.a.
Sh
oo
ts 
S.
a.
Bo
str
yc
hia Ul
va
Ca
log
los
sa
N
o.
 s
qu
ar
es
 e
at
en
0
10
20
30
40
50
60
A A A
A,B
C,D
B,C
A,B
Fig. 3. Gammarus palustris. No-choice feeding assay showing
the number of squares eaten by amphipods among 7 artificial
plant diets, mean ? SE. Statistics are from a Kruskal-Wallis
test; bars with the same letters were not significantly different
in post hoc multiple comparison tests. S.a.: Spartina alterni-
flora; sen: senescent
Dead S.a. (33) Caloglossa  (37)
Bostrychia  (30)
Sen S.a. (30) 
Shoots S.a. (32) 
Ulva  (32)
Live S.a. (40)  
Starvation (36)
Day
0 5 10 15 20 25 30
%
 S
ur
vi
va
l
0
20
40
60
80
100 p < 0.0001
Fig. 4. Gammarus palustris. Survival of amphipods on 7 differ-
ent macrophyte diets and a starvation control. Significance
and groupings (shown by vertical lines to the right of the plot)
are from ?2 analyses of survivorship on Day 28. Numbers in
parentheses are the number of individuals reared on each 
diet. S.a.: Spartina alterniflora; sen: senescent
Mar Ecol Prog Ser 364: 87?95, 2008
Stable isotope determination of marsh producers 
and consumers
Idealized trophic fractionation factors assume shifts
of 1? ?13C and 3.5? ?15N for consumers feeding on a
food source (DeNiro & Epstein 1978, 1981, Minagawa
& Wada 1984, Vanderklift & Ponsard 2003). Thus,
mean isotope values of field-collected samples sug-
gested that the organic contents of mud sediments (i.e.
small particulate detrital matter and benthic microal-
gae on the marsh surface) were likely the primary diet
source fueling Gammarus palustris, Uhlorchestia
spartinophila, Palaemonetes pugio, and Ilyanassa
obsoleta (Fig. 5). The endpoint for this production, par-
ticularly for production of G. palustris, appeared to be
marsh killifish Fundulus heteroclitus. Using the ideal-
ized trophic fractionation factors, one would conclude
that consumers were not subsisting on diets of cord-
grass, macroalgae, or the organic contents of sand sed-
iments.
However, the isotope values of amphipods reared on
known diet sources since birth did not reflect idealized
trophic shifts. In particular, ?13C values for amphipods
reared on cordgrass diets were 0.47 to 5.08? more 13C
depleted than their food source (Fig. 5), in contrast to
the expected 1? 13C enrichment. This occurred in
both years regardless of whether samples had been
preserved in formalin. Amphipods reared on dead
cordgrass and fixed in formalin were 15N-enriched
(1.0?) and 13C-depleted (1.2?) relative to amphipods
that were reared on dead cordgrass but never fixed.
These values are consistent with published effects of
formalin preservation (Sarakinos et al. 2002). Unex-
pected ?13C and ?15N shifts were also found for
amphipods feeding on algal and sand diets (Fig. 5);
only the marsh mud diet resembled idealized fraction-
ation factors.
DISCUSSION
One hypothesis for why the ?world is green? is that
live plant tissues are often chemically and structurally
defended from herbivores (Rosenthal & Berenbaum
1992), and only upon leaf death does most plant mate-
rial get incorporated into food webs (Mann 1988,
Cebrian 1999). Salt marshes are one of the world?s
?greenest? ecosystems, with annual productivity rival-
ing that of coral reefs and tropical rain forests (Silliman
& Bortolus 2003). Our study suggests that salt marshes
are ?green? because the living plants are physically and
chemically resistant to the invertebrate consumers that
dominate contemporary marshes.
92
Table 2. Gammarus palustris. Mean ? SE (N) total length of all
amphipods surviving to Day 28, and % of female amphipods
that reached sexually maturity when reared on different
macrophyte diets. Different letters denote statistical differ-
ences via 1-way ANOVA followed by Tukey?s test. S.a.: Spar-
tina alterniflora, Sen: senescent
Diet Survivor size (mm) % Sexually mature f
Dead S.a. 4.25 ? 0.197 (26)a 72.2 (18)
Sen S.a. 3.06 ? 0.184 (10)b 25.0 (8)
Live S.a. 3.19 ? 0.133 (14)b 0 (11)
Shoots S.a. 3.36 ? 0.139 (8)b 14.3 (7)
Bostrychia 3.27 ? 0.211 (7)b 0 (6)
Ulva 3.58 ? 0.208 (7)a,b 20.0 (5)
Caloglossa 3.42 ? 0.178 (7)a,b 25.0 (4)
Table 3. Gammarus palustris. Separate 1-way ANOVAs test-
ing for familial differences in the lifespan of offspring reared
on different diets. Bold values were significant at p < 0.05. 
S.a.: Spartina alterniflora, Sen: senescent
Diet df MS F p
Dead S.a. 2,29 1.4 0.02 0.975
Sen S.a. 5,22 41.6 0.47 0.796
Live S.a. 4,35 577.7 7.35 <0.0001
Shoots S.a. 4,26 282.3 3.42 0.022
Bostrychia 5,23 27.2 0.44 0.813
Ulva 5,25 262.3 3.61 0.014
Caloglossa 3,33 26.0 0.37 0.773
Fig. 5. Dual isotope plot depicting mean (? SE) ?15N and ?13C
values for sediments, primary producers, and consumers, ex-
pressed as parts per mil (?). Icons represent the amphipod
Gammarus palustris, the killifish Fundulus heteroclitus, the
snail Illyanassa obsoleta, and the shrimp Palaemonetes pugio.
U.s.: salt-marsh amphipod Uhlorchestia spartinophila. Open
and filled symbols: samples collected in 2000 and 2001, re-
spectively. Arrows attached to symbols: observed trophic
shifts for G. palustris reared on known food sources since
birth. Idealized trophic shift is 1? ?13C and 3.5? ?15N for a
consumer feeding on any given food source. Sen: senescent
Parker et al.: Food web consequences of detrital specialization
The common salt marsh amphipod Gammarus palus-
tris strongly preferred dead over live cordgrass Spar-
tina alterniflora in laboratory feeding assays (Fig. 1),
and live cordgrass was 7- to 20-fold tougher than dead
tissues (Table 1). However, when the toughness of live
plant tissues was destroyed by grinding and incorpora-
tion into a soft agar matrix, amphipods consumed
equal amounts of live, senescent, and dead cordgrass,
(Fig. 3). Similarly, the omnivorous marsh crab Armases
cinereum also consumes live cordgrass following
destruction of its physical toughness (Pennings et al.
1998); thus, salt marsh plants appear to be too tough for
consumption by common marsh invertebrates. Chemi-
cal defenses may also be important, as phenolic acids
and other water-soluble extracts from senescent and
live S. alterniflora deter feeding by invertebrate and
vertebrate consumers (Buchsbaum et al. 1984, Valiela
& Rietsma 1984, B?rlocher & Newell 1994, Pennings et
al. 1998). The solubility of these types of defenses may
have allowed them to leach from our artificial foods,
thus underestimating their possible effects.
Detrital feeding often requires assistance from mi-
crobial processing. Fungi and bacteria soften plant tis-
sues, neutralize consumer-deterrent phenolic acids in
Spartina alterniflora (Newell 1993), and can increase
the nutritional value of such plant material (Findlay &
Tenore 1982, B?rlocher & Newell 1994). In marked
contrast to intense animal?microbe competition for
more energy-rich resources (Burkepile et al. 2006), this
?microbial conditioning? of leaf litter appears to be a
general phenomenon enhancing the palatability of leaf
litter to detritivores (B?rlocher 1980). As a possible
example, this conditioning may explain why amphi-
pods ignored senescent cordgrass in the presence of
dead cordgrass (Fig. 1), performed poorly on senescent
relative to dead cordgrass (Fig. 4, Table 2), yet were
willing to feed on senescent cordgrass in the absence
of alternative diets (Fig. 1).
The plant traits that precluded feeding on other
macrophytes were less clear but suggestive of chemi-
cal defenses. The green alga Ulva was physically soft
and nutritious relative to cordgrass (Table 3), but was
not eaten as fresh tissues and was consumed in only
limited amounts when incorporated into agar (Figs. 1 &
3, see also Borowsky & Borowsky 1987), suggesting a
chemical defense. Gammarus palustris is inhibited by
chemical extracts from Ulva (Borowsky & Borowsky
1990), and Ulva spp. produce dimethylsulfoniopropi-
onate (DMSP), a precursor compound that is activated
into more toxic compounds (acrylic acid and dimethyl
sulfide) when algal cells are ruptured by herbivores
(Van Alstyne & Puglisi 2007). In contrast, the filamen-
tous red alga Bostrychia radicans was eaten at modest
levels when presented as fresh tissues but was rejected
when incorporated into agar (Figs. 1 & 3). Reasons for
this difference in palatability are unclear, but it is pos-
sible that this alga has compounds that interact with
the agar matrix to reduce its utility as a food, or that it
has defenses that are activated by the process of grind-
ing and incorporation into agar (Cetrulo & Hay 2000).
Differential predation risk may also explain res-
tricted host-use by Gammarus palustris. G. palustris
resides almost exclusively within the leaf interstices of
Spartina alterniflora and only sparingly on other marsh
substrates (Gable & Croker 1977, van Dolah 1978, this
study). This pattern is not explained by competition for
space; instead, intense predation by marsh killifish
Fundulus heteroclitus limits G. palustris to the protec-
tion of cordgrass stems (Van Dolah 1978). Moreover,
stable isotope values suggest that killifish derive a
large portion of their diet from G. palustris (Fig. 5).
Thus, predation risk likely restricts G. palustris to cord-
grass stems, whereas food cues like plant toughness
and secondary chemistry restrict feeding to decaying
cordgrass versus other commonly available macro-
phytes.
The evolution of host-plant specialization has been
widely studied among herbivorous insects and some
marine amphipods (Bernays & Graham 1988, Poore et
al. 2008), but less so among detritivores. Gammarus
palustris is an apparent specialist on dead Spartina
alterniflora but possesses the evolutionary ?raw mater-
ial? necessary for a diet breadth expansion, particularly
onto the green alga Ulva. Gammarus palustris exhib-
ited heritable variation for survival on the shoots and
live tissues of S. alterniflora and on Ulva (Table 3), with
some families surviving the duration of the experiment
on these tissues and others rapidly perishing. Diet
choice is under genetic control in G. palustris, with
particular genotypes possessing distinct salivary
enzymes leading to preferential consumption of Ulva
over other macrophytes (Borowsky et al. 1985, Guarna
& Borowsky 1993). Amphipods in our study did not
tend to consume Ulva (Fig. 1), but the presence of her-
itable variation for feeding and survival on alternate
hosts suggests that G. palustris could use hosts other
than S. alterniflora. This genetic potential may be
stunted by ecological factors like predation risk in the
field (e.g. Duffy & Hay 1994, Poore & Steinberg 2001),
or by the temporal and spatial variability of hosts like
Ulva compared to more predictable hosts like S.
alterniflora.
Stable isotopes are widely used for rapid and inte-
grated pictures of consumer diet, trophic position, and
the general flow of energy and materials through food
webs. The basic assumptions are that primary produc-
ers possess distinct isotope ratios, and most impor-
tantly, that producer isotopes are passed on to con-
sumers in predictable shifts of ??13C = 1? and ??15N =
3.5? for each trophic transfer (DeNiro & Epstein 1978,
93
Mar Ecol Prog Ser 364: 87?95, 2008
1981, Minagawa & Wada 1984, Vanderklift & Ponsard
2003). Our initial analysis of stable isotope ratios sug-
gested that field populations of Gammarus palustris
incorporated most of their carbon from detrital and
microalgal food sources associated with marsh mud,
not from dead Spartina alterniflora (Fig. 5). However, a
direct test of this hypothesis showed that the isotope
values of laboratory amphipods reared on marsh mud
were enriched in 13C and thus dissimilar from field-col-
lected G. palustris (Fig. 5), suggesting marsh mud con-
tents as an unlikely diet source for G. palustris.
The isotope values of amphipods reared in the labo-
ratory on dead Spartina alterniflora, however, closely
resembled field-collected amphipods (Fig. 5), and
Gammarus palustris clearly inhabits, preferentially
feeds on, and performs best on diets of dead S. alterni-
flora over all other hosts tested (Figs. 1?3, Table 2). All
findings suggest that field amphipods derive a large
portion of their diet from dead cordgrass. We could not
have reached this conclusion without using controlled
assimilation assays. Amphipods reared on dead
S. alterniflora were depleted by over 3? in 13C relative
to their food source, not enriched by the expected 1?
13C. Moreover, formalin effects could not explain the
general isotopic resemblance of laboratory amphipods
reared on cordgrass to field-collected amphipods
(Fig. 5); both fixed and unfixed amphipods reared on
live, senescent, or dead cordgrass exhibited similar
directions in trophic shift, i.e. 13C-depleted rather than
enriched.
Fractionation by crustaceans in particular is highly
variable, often showing unpredictable levels of 15N
enrichment and 13C-depletion rather than enrichment
(Vanderklift & Ponsard 2003, Crawley et al. 2007).
Potential explanations include differential assimilation of
13C-depleted compounds like plant lipids or bacterially-
degraded lignin (Benner et al. 1987, van Dongen et
al. 2002), fasting (Vander Zanden & Rasmussen 2001),
and the consumption of N-poor foods (Vanderklift &
Ponsard 2003). In our study, a likely explanation is that
amphipods in the field augment diets of dead cordgrass
with epiphytic algae, consistent with gut content ob-
servations of algal remains (Gable & Croker 1977), and
with our observations that amphipods consumed the
2 ephiphytic red algae Bostrychia radicans and Calo-
glossa leprieurii in the laboratory (Fig. 1). Although the
isotope composition of amphipods reared on epiphytic
algae did not resemble values for field amphipods
(Fig. 5), both diets were 15N-enriched and 13C-depleted
relative to dead cordgrass. Thus, opportunistic algal
consumption could result in 15N-enrichment and 13C-de-
pletion relative to a diet of dead cordgrass alone (Fig. 5),
further demonstrating the value of combining dietary
and stable isotope data to infer food web structure
(see also Winemiller et al. 2007).
The isotope content of marsh killifishes Fundulus
heteroclitus suggests predation on Gammarus palus-
tris (Fig. 5), a tentative conclusion given that we did
not verify this trophic fractionation with assimilation
assays. Nevertheless, it is consistent with field and lab-
oratory experiments showing that killifish are often the
major source of mortality for G. palustris and other
marsh invertebrates (Vince et al. 1976, van Dolah 1978,
Kneib 1982), and with other isotope analyses suggest-
ing killifish are major predators in marsh food webs
(Currin et al. 1995, Wainright et al. 2000). Thus, con-
sumption of G. palustris and similar marsh inverte-
brates by F. heteroclitus may be an important trophic
pathway linking cordgrass production to nearshore
food webs (Covi & Kneib 1995, Kneib et al. 1997), a
pathway made possible because microbes liberate
cordgrass production that in life is protected by struc-
tural and chemical defenses.
Acknowledgements. We thank E. Kinard, J. Long, D. Burke-
pile, A. Hollebone, C. Kicklighter, D. Hasan, R. Kneib, and S.
Pennings for assistance with amphipod collections, rearing,
and identification, C. Payne for stable isotope analyses, and
3 anonymous reviewers for constructive comments on the
manuscript. Support was provided by an NSF IGERT grant
DGE 0114400 and the Teasley Endowment to Georgia Tech.
LITERATURE CITED
B?rlocher F (1980) Leaf-eating invertebrates as competitors of
aquatic hyphomycetes. Oecologia 47:303?306
B?rlocher F, Newell SY (1994) Phenolics and proteins affect-
ing palatability of Spartina leaves to the gastropod Litto-
raria irrorata. PSZN I: Mar Ecol 15:65?75
Benner R, Fogel ML, Sprague EK, Hodson RE (1987) Deple-
tion of 13?C in lignin and its implications for stable carbon
isotope studies. Nature 329:708?710
Bernays E, Graham M (1988) On the evolution of host speci-
ficity in phytophagous arthropods. Ecology 69:886?892
Borowsky B, Borowsky R (1987) Feeding by the salt-marsh
amphipod, Gammarus palustris on Ulva and Entero-
morpha. Am Zool 27:55A (abstract)
Borowsky R, Borowsky B (1990) Feeding inhibition of the salt-
marsh amphipod Gammarus palustris Bousfield, 1969, by
heat-labile substances in Ulva lactuca L. Crustaceana 59:
299?301
Borowsky R, Borowsky B, Milani H, Greenberg P (1985) Amy-
lase variation in the salt-marsh amphipod, Gammarus
palustris. Genetics 111:311?323
Buchsbaum R, Valiela I, Swain T (1984) The role of phenolic
compounds and other plant constituents in feeding by
Canada geese in a coastal marsh. Oecologia 63:343?349
Burkepile DE, Parker JD, Woodson CB, Mills HJ, Kubanek J,
Sobecky PA, Hay ME (2006) Chemically mediated compe-
tition between microbes and animals: microbes as con-
sumers in food webs. Ecology 87:2821?2831
Cebrian J (1999) Patterns in the fate of production in plant
communities. Am Nat 154:449?468
Cetrulo GL, Hay ME (2000) Activated chemical defenses in
tropical versus temperate seaweeds. Mar Ecol Prog Ser
207:243?253
Covi M, Kneib R (1995) Intertidal distribution, population
94
Parker et al.: Food web consequences of detrital specialization
dynamics and production of the amphipod Uhlorchestia
spartinophila in a Georgia, USA, salt marsh. Mar Biol 121:
447?455
Crawley KR, Hyndes GA, Vanderklift MA (2007) Variation
among diets in discrimination of 13?C and 15?N in the
amphipod Allorchestes compressa. J Exp Mar Biol Ecol
349:370?377
Currin C, Newell S, Paerl HW (1995) The role of standing
dead Spartina alterniflora and benthic microalgae in salt-
marsh food webs ? considerations based on multiple sta-
ble-isotope analysis. Mar Ecol Prog Ser 121:99?116
DeNiro MJ, Epstein S (1978) Influence of diet on the distribu-
tion of carbon isotopes in animals. Geochim Cosmochim
Acta 42:495?506
DeNiro MJ, Epstein S (1981) Influence of diet on the distribu-
tion of nitrogen isotopes in animals. Geochim Cosmochim
Acta 45:341?351
Duffy J, Hay M (1994) Herbivore resistance to seaweed chem-
ical defense ? the roles of mobility and predation risk.
Ecology 75:1304?1319
Findlay S, Tenore K (1982) Nitrogen source for a detritivore:
detritus substrate versus associated microbes. Science
218:371?373
Gable M, Croker R (1977) The salt marsh amphipod, Gam-
marus palustris Bousfield, 1969 at the northern limit of its
distribution. I. Ecology and life cycle. Estuar Coast Mar Sci
5:123?134
Guarna MM, Borowsky RL (1993) Genetically controlled food
preference: biochemical mechanisms. Proc Natl Acad Sci
USA 90:5257?5261
Hay M, Stachowicz J, Cruz-Rivera E, Bullard S, Deal M,
Lindquist N (1998) Bioassays with marine and freshwater
macroorganisms. In: Haynes K, Millar J (eds) Methods
in chemical ecology, Vol 2. Chapman & Hall, New York,
p 39?141
Kneib R (1982) Habitat preference, predation, and the inter-
tidal distribution of gammaridean amphipods in a North
Carolina salt marsh. J Exp Mar Biol Ecol 59:219?230
Kneib RT, Newell SY, Hermeno ET (1997) Survival, growth
and reproduction of the salt-marsh amphipod Uhlorches-
tia spartinophila reared on natural diets of senescent and
dead Spartina alterniflora leaves. Mar Biol 128:423?431
Mann KH (1988) Production and use of detritus in various
freshwater, estuarine, and coastal marine ecosystems.
Limnol Oceanogr 33:910?930
Mattson W (1980) Herbivory in relation to plant nitrogen con-
tent. Annu Rev Ecol Syst 11:119?161
Minagawa M, Wada E (1984) Stepwise enrichment of 15N
along food chains: further evidence and the relation
between ?15N and animal age. Geochim Cosmochim Acta
48:1135?1140
Newell SY (1993) Decomposition of shoots of a salt-marsh
grass. Adv Microb Ecol 13:301?326
Odum WE, Kirk PW, Zieman JC (1979) Non-protein nitrogen
compounds associated with particles of vascular plant
detritus. Oikos 32:363?367
Pennings SC, Carefoot TH, Siska EL, Chase ME, Page TA
(1998) Feeding preferences of a generalist salt-marsh
crab: relative importance of multiple plant traits. Ecology
79:1968?1979
Polis GA, Strong DR (1996) Food web complexity and commu-
nity dynamics. Am Nat 147:813?846
Poore AGB, Steinberg PD (2001) Host-plant adaptation in an
herbivorous marine amphipod: genetic potential not real-
ized in field populations. Evolution 55:68?80
Poore AGB, Hill NA, Sotka EE (2008) Phylogenetic and geo-
graphic variation in host breadth and composition by her-
bivorous amphipods in the family Ampithoidae. Evolution
62:21?38
Rietsma CS, Valiela I, Buchsbaum R (1988) Detrital chemistry,
growth, and food choice in the salt-marsh snail (Melampus
bidentatus). Ecology 69:261?266
Rosenthal GA, Berenbaum MR (eds) (1992) Herbivores: their
interactions with secondary metabolites: evolutionary and
ecological processes. Academic Press, San Diego, CA
Russell-Hunter WD (1970) Aquatic productivity: an introduc-
tion to some basic aspects of biological oceanography and
limnology.Collier MacMillan, London
Sarakinos HC, Johnson ML, Vander Zanden MJ (2002) A syn-
thesis of tissue-preservation effects on carbon and nitro-
gen stable isotope signatures. Can J Zool 80:381?387
Silliman BR, Bortolus A (2003) Underestimation of Spartina
productivity in western Atlantic marshes: marsh inverte-
brates eat more than just detritus. Oikos 101:549?554
Silliman BR, Newell SY (2003) Fungal farming in a snail. Proc
Natl Acad Sci USA 100:15643?15648
Teal JM (1962) Energy flow in the salt marsh ecosystem of
Georgia. Ecology 43:614?624
Valiela I, Rietsma CS (1984) Nitrogen, phenolic acids, and
other feeding cues for salt marsh detritivores. Oecologia
63:350?356
Van Alstyne KL, Puglisi MP (2007) DMSP in marine macro-
algae and macroinvertebrates: distribution, function, and
ecological impacts. Aquat Sci 69:394?402
Van Dolah R (1978) Factors regulating the distribution and
population dynamics of the amphipod Gammarus palus-
tris in an intertidal salt marsh community. Ecol Monogr
48:191?217
Van Dongen BE, Schouten S, Sinninghe Damst? JS (2002)
Carbon isotope variability in monosaccharides and lipids
of aquatic algae and terrestrial plants. Mar Ecol Prog Ser
232:83?92
Vander Zanden MJ, Rasmussen JB (2001) Variation in ?15N
and ?13C trophic fractionation: implications for aquatic
food web studies. Limnol Oceanogr 46:2061?2066
Vanderklift MA, Ponsard S (2003) Sources of variation in con-
sumer-diet 15?N enrichment: a meta-analysis. Oecologia
136:169?182
Vince S, Valiela I, Backus N, Teal J (1976) Predation by
the salt marsh killifish Fundulus heteroclitus (L.) in rela-
tion to prey size and habitat structure: consequences for
prey distribution and abundance. J Exp Mar Biol Ecol
23:255?266
Wainright SC, Weinstein MP, Able KW, Currin CA (2000) Rel-
ative importance of benthic microalgae, phytoplankton
and the detritus of smooth cordgrass Spartina alterniflora
and the common reed Phragmites australis to brackish-
marsh food webs. Mar Ecol Prog Ser 200:77?91
Winemiller KO, Akin S, Zeug SC (2007) Production sources
and food web structure of a temperate tidal estuary: inte-
gration of dietary and stable isotope data. Mar Ecol Prog
Ser 343:63?76
95
Editorial responsibility: Robert Feller, 
Columbia, South Carolina, USA
Submitted: February 4, 2008; Accepted: March 31, 2008
Proofs received from author(s): July 7, 2008