J .P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
217
minimum and maximum prey size with changes in predator size using least absolute values regression with bootstrapped estimates of coefficient standard errors. We used the
rule n . 10 q
to determine quantiles q reflecting trends in minimum and maximum prey size given sample size n Scharf et al. 1998. Total length mm was the size parameter used in all
regressions. Total lengths of flounder consumed by searobins in the field were estimated from measured standard lengths using the regression equation published in Able and
Fahay 1998
TL 5 1.213 SL 2 0.447 2.3.2. Winter flounder size selection
The nonlinear relationship between prey size and mortality prevented the use of a linear logistic model to test for the effects of prey size and substratum type on prey
selectivity. Therefore we used logistic generalized additive models GAMs with spline smoothers S-plus 4.5, 1997 to nonparametrically model the effects of prey size on the
mortality of flounder in the presence and absence of sand. Logistic GAMs use the ratio of response frequencies logits and scatterplot smoothers to fit data-defined and
unspecified functions to the relationship between response and predictor variables Hastie, 1993. Since strong predator–prey body size relationships were not evident see
Results, replicates in which prey were consumed were pooled within treatments. Individual fish were scored as live or dead and these response frequencies were used in
GAMs constructed independently for the two substrata with prey size as the predictor variable. The strength of the relationship between prey size and mortality were assessed
2
with approximate x tests Hastie and Tibshirani, 1990; Hastie, 1993.
2.3.3. Day–night prey selection and prey switching A two-way ANOVA was used to test for the effects of prey species flounder shrimp
and experimental period day night on searobin prey consumption in the day–night selection experiments. Only replicates in which searobins consumed prey were used in
this analysis. Chessons a with food depletion; Chesson, 1983 was calculated for
i
replicates in which two or more prey were consumed, and used as the prey selectivity index in the switching experiment. The null hypothesis of no prey selection
a 5 a 5
i j
0.5 was tested at each ratio using a t-test and Bonferroni probability adjustment P ,0.05 350.017 to guard against multiple testing errors.
3. Results
3.1. Field studies 3.1.1. Patterns of searobin abundance and diet composition
Striped searobins n 578; 121–367 mm TL were collected in shallow water gillnets from June through October, 1998, with most fish captured before the end of August.
218 J
.P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
Two size classes of fish 180–240 mm, 270–370 mm TL were present in June, while fish collected in July and August were primarily from a single class 190–280 mm TL
Fig. 2. Searobins were collected at four stations in Sandy Hook Bay, but were absent from the Navesink River Fig. 1.
Sand shrimp was the dominant prey of searobins occurring in 81 of stomachs Table 2. Mysids and YOY winter flounder were also common in diets in which amphipods
primarily Ampelisca sp. and Gammarus sp. and lady crabs Ovalipes ocellatus also occurred. Other prey, including blue crabs
Callinectes sapidus, xanthid crabs, and blue mussels Mytilus edulis, were present infrequently three stomachs.
Searobin diets were substantially different in June than in July and August Table 2. In June, sand shrimp was the dominant prey and as many as 18 shrimp were consumed
by individual predators. Winter flounder were ranked second in importance in June and occurred in 69 of stomachs. Searobins with flounder in stomachs were collected at
three stations Fig. 1a and, in a sample of 26 predators, 85 of the searobins had consumed an average of 360.6 flounder maximum511. With one exception, flounder
Fig. 2. Length frequency histograms for striped searobins collected in gillnets in 1998.
J .P
. Manderson
et al
. J
. Exp
. Mar
. Biol
. Ecol
. 242
1999 211
– 231
219 Table 2
Frequency of prey occurrence and contribution to the total prey weight in the diets of striped searobins collected in shallow water gillnets in NSHES in 1998 Prey
June n 536 July-August n 539
September-October n 53 Total n 578
species Mean
Mean Mean
Mean by wt.
Frequency by wt.
Frequency by wt.
Frequency by wt.
Frequency SE
n SE
n SE
n SE
n Pseudo-
17.7 68.6
1.5 2.56
0.0 0.00
8.8 32.5
pleuronectes 3.0
24 1.5
1 1.8
25 americanus
Crangon 61.1
94.3 33.5
74.4 0.0
0.00 44.7
80.5 septemspinosa
4.8 33
5.8 29
4.1 62
Mysids 8.6
57.1 52.4
71.8 0.0
0.00 30.5
57.1 2.4
20 6.8
28 4.4
48 Amphipods
0.6 31.4
6.2 25.64
0.0 0.00
3.4 27.3
0.4 11
3.1 10
1.6 21
Ovalipes 0.0
0.0 3.9
10.26 33.3
33.33 3.3
6.5 ocellatus
2.0 4
33.3 1
1.6 5
Other 9.1
45.7 2.6
7.7 33.3
33.3 6.7
,3.90 3.8
16 1.8
3 33.3
1 2.3
,3
220 J
.P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
were absent from searobin diets in July and August when mysids were the dominant prey Fig. 1b.
3.1.2. Predator–prey body size Searobins 188 –350 mm TL consumed winter flounder ranging in size from 15 to 57
mm TL n 582; 12–48 mm SL, 5.4–24.3 of predator TL; Fig. 3a. Although slopes of regressions estimating minimum and median prey sizes 15th and 50th quantiles were
significant, the coefficients were small, suggesting that the predator–prey size relation- ships were weak Table 3a. Although too few searobins were collected to permit a
rigorous analysis n 535, the contribution of flounder to searobin diets did not appear to vary with predator size.
Sand shrimp found in searobin stomachs 181–354 mm ranged in size from 10–49 mm in total length n 5380; 4.4–17.1 of predator TL; Fig. 3b. Predator–prey body
size relationships were significantly positive at all quantiles tested Table 3a but the
Fig. 3. Predator–prey body size relationships between striped searobins collected in gillnets and a winter flounder and b sand shrimp found in stomachs. Lines are quantile regressions estimating maximum, median,
and minimum prey sizes see Table 3a.
J .P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
221 Table 3
Estimates of slopes and intercepts from quantile regressions performed to determine body size relationships reflecting maximum, median, and minimum prey size for searobins a collected in the field and b used in
laboratory experiments examining size selection
a
Quantile Slope estimate 6SE
Intercept
b
a Field collections Winter flounder n 582
ns
85th 0.02
60.04 31.71 610.36
50th 0.04 6 0.02
18.19 66.23
ns
15th 0.05 60.02
9.57 65.21
Sand shrimp n 5380
ns
95th 0.13 60.02
0.12 65.68
ns
50th 0.05 6 0.01
4.80 62.86
ns
5th 0.03 60.01
1.86 61.72
b Laboratory experiments with winter flounder prey Sand absent n 569
ns ns
85th 20.07
6 0.20 66.22
6 41.16
ns ns
50th 20.08
6 0.08 51.85
6 20.05
ns
15th 0.08 60.03
10.00 67.46
Sand present n 565
ns ns
85th 0.04
60.22 49.40
650.49
ns ns
50th 0.10
60.08 23.50
619.50
ns ns
15th 0.01
60.05 29.88
611.90
a
SE , boot strapped estimate of standard error. See methods for formula used to select quantiles
b
representing maximum and minimum prey size given sample sizes. P ,0.05; P ,0.01; P ,0.001; ns, not significant.
slope estimating maximum prey size 95th quantile was significantly higher than the slope for minimum prey size 5th quantile; F 535.53, df51,758; P ,0.001. This
pattern suggests that although larger searobins consumed increasingly large shrimp, they continued to feed on small prey.
3.2. Laboratory experiments 3.2.1. Winter flounder size selection
Total consumption of flounder by searobins averaged 4.5 prey per experiment SE53.2, range50–10 prey and was not influenced by the presence of sand t-test:
t 5 20.177, df526, P 50.86. Searobins fed in 79 of replicates. Prey as large as 110 mm TL were eaten 40 of predator body length, but the predators failed to consume
the largest prey in any replicate Fig. 4.
Searobin size did not influence the size of prey consumed in the laboratory Table 3b. Slopes of quantile regressions were not significant except at the 15th quantile in the
absence of sand, where the coefficient was small. Estimates of maximum prey size for flounder 85th quantile were 70.4 and 62.8 mm TL in the presence and absence of sand
25 and 24 of predator length.
Maximum prey size was constrained by predator esophageal width Fig. 4. Searobin
222 J
.P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
Fig. 4. Body size relationships between striped searobins and winter flounder consumed in laboratory size selection experiments conducted a with and b without sand substratum. Closed circles indicate prey
consumed and open circles prey which survived. Lines indicate estimated maximum prey size mm TL for winter flounder if mouth width M or esophageal width E imposed a morphological constraint on prey size
see Table 5.
mouth widths 33–56 mm were larger than esophageal widths 24–30 mm and increased more dramatically with increases in predator total length Table 4. Based on
predator length gape and prey length body depth relationships, predictions of maximum flounder size ranged from 120 to 173 mm TL 33–56 mm BD if mouth width
determined the maxima and from 73 to 88 mm TL 24–30 mm BD if esophageal width determined the maxima. The largest flounders consumed were at least 39 mm TL smaller
than predicted by mouth width.
Flounder mortality was strongly influenced by their size particularly in the presence of sand Table 5, Fig. 5. Mortality probabilities estimated using logistic GAMs followed
trends in mean proportions of prey consumed and showed maxima of 72 SE54 for
J .P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
223 Table 4
Results of linear regressions performed to determine the relationships between searobin total length n 513, 227–320 mm TL and mouth and esophageal widths mm, and between winter flounder total length TL mm
and body depth n 5346, 30–110 mm TL
2
Dependent variable Effect
Coefficient t
P r
SE Searobin mouth width
Intercept 210.587
21.130 0.282
0.795 mm
9.367 Searobin length
0.218 6.651
,0.001 TL mm
0.033 Searobin esophageal width
Intercept 9.714
1.726 0.112
0.411 mm
5.629 Searobin length
0.059 2.948
0.013 TL mm
0.020 Flounder length
Intercept 6.792
9.407 ,0.001
0.930 TL mm
0.722 Flounder body depth
2.856 67.556
,0.001 mm
0.042
flounder in the 40–50 mm size class in the absence of sand and 65 SE53.5 for the 50–60 mm size class when sand substratum was present. Although mortality declined
for larger sized flounders in both treatments, substratum type appeared to influence the likelihood that searobins consumed fish in the size classes. The mortality curve for
flounder on sand substratum was shifted to the right of the curve for fish without sand, suggesting that larger fish 50–70 mm were more at risk on sand, while small fish ,50
mm were more vulnerable when sand was absent.
3.2.2. Day–night prey selection More searobins fed day: n 517; night: n 515 and slightly more flounder were
consumed than sand shrimp during the daytime treatment Fig. 6a. However, the effects of prey type and diurnal period were not significant at the traditional P ,0.05 level in
the ANOVA Table 6.
Table 5 Results of binomial generalized additive models GAM used to predict the likelihood of mortality for size
classes of winter flounder confronted with searobin predators in the presence and absence of sand in the laboratory
2
Treatment df
x P
Sand absent Prey length
2.8 7.17
0.058 Residual deviance 149
149 Sand present
Prey length 2.8
12.34 0.005
Residual deviance 195 163
224 J
.P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
Fig. 5. a Proportion of total prey consumed and b GAM predictions of mortality for winter flounder size classes in laboratory experiments of searobin size selectivity with and without sand substrata. Error bars
indicate 2 standard errors in probability plot.
3.2.3. Prey switching Searobins consumed shrimp and flounder in proportions which were not different
from initial relative prey densities Fig. 6b; Table 6b. t-Tests on Chessons a showed no
prey selectivity at all three ratios when a Bonferroni correction was applied critical P 50.017. Predators consumed 10 or more prey offered in 9 replicates for each
prey ratio, and as many as 13 winter flounder and nine sand shrimp were consumed in 15:5 and 5:15 treatments.
3.2.4. Predator behavior Searobins fed in half of the 2 h trials 5 of 10 and individual predators consumed as
¯ many as ten flounders x 55.4. Flounder were ingested head or tail first and consumed
whole. An average of 65 627 of attacks total n 542 resulted in captures, and most attacks 85614.5 and observed captures 14 of 20 occurred on the bottom. The
J .P. Manderson et al. J. Exp. Mar. Biol. Ecol. 242 1999 211 –231
225
Fig. 6. Results of a day–night prey selection and b daytime prey switching experiments examining striped searobin selection for winter flounder and sand shrimp in the laboratory.
number of flounder consumed was significantly correlated Pearson’s r 50.85, P 5 0.002 with the time predators spent ‘walking’ on or probing the substratum with
modified pectoral fin rays and thus potentially searching the substratum for prey. Prey were flushed from the sediment as a result of this behavior in 55 of attacks. Three
searobins exhibited ‘digging’ behavior in which they anchored themselves with posterior pectoral fin rays and moved the sediment with anterior free fin rays. Buried flounder
were flushed from sand substratum as a result of this behavior and attacked.
4. Discussion
