120 J
. Grudemo, T. Bohlin J. Exp. Mar. Biol. Ecol. 253 2000 115 –127
medium competition from H . ventrosa one target snail 1 seven H. ventrosa competitor
22
snails: density 6160 m ; and level 5, high competition from H
. ventrosa one target
22
snail 1 23 H .ventrosa competitor snails: density 18 480 m
. These figures were chosen to represent low, intermediate and high population density and competition
Fenchel, 1975a; Morrisey, 1987. 2.4. Statistical analyses
Growth of the target snails was analysed with a four-factor analysis of variance ANOVA. In this, there were three orthogonal factors, Species SP, a fixed factor with
two levels H . ulvae and H. ventrosa; Competition C, a fixed factor with five levels
see above; and Sediment type ST, a fixed factor with two levels sandy or silty sediment, see above. The fourth factor was a random Bay B factor with eight levels,
nested under Sediment type ST. That is, eight sediments came from bays with coarse-grained sediment and eight from bays with fine-grained sediments.
This experimental design resulted in an ANOVA with the following linear model: x
5 m 1 SP 1 Cj 1 ST 1 BST
1 SP 3 C
ijklm i
k lk
ij
1 SP 3 ST 1 SP 3 BST 1 C 3 ST
ik ilk
jk
1 C 3 BST 1 SP 3 C 3 ST
jlk ijk
1 SP 3 C 3 BST 1 e
1
ijlk ijklm
where x is the growth of a specific target individual,
m is the growth averaged over
ijklm
all treatments, and e is the residual.
ijklm
We used three replications of each combination of treatments, resulting in a total number of 480 target snails. To obtain appropriate interactions Hurlbert and White,
1993, and to reduce the heterogeneity in variances, we used the logarithm of growth lngrowth11 in the ANOVA. Degrees of freedom and expected mean sum of squares
were calculated with the method in Underwood 1997, pp. 364–369. To obtain a balanced statistical analysis, we replaced dead snails with the average growth of the
survivors within the same treatment, and we accordingly reduced the degrees of freedom in the residual Underwood, 1997. Student–Newman–Keul’s SNK test was used as
post-hoc test to unveil differences among levels of the significant factors and interac- tions. If the relationship between growth and level of competition is similar between
species, then the interaction term SP3C will not be significant. If, on the other hand, intra- or interspecific competition is more important, the interaction effect will become
large.
3. Results
3.1. General pattern At the end of the experiment, 462 of the 480 marked snails 96 could be used for
growth measurement. Among the rest, four were lost during handling and 14 had either
J . Grudemo, T. Bohlin J. Exp. Mar. Biol. Ecol. 253 2000 115 –127
121
lost the paint or died. These 14 snails came from all competition treatments. The high survival at all competition intensities suggests that a difference in survival is no good
measure of differences in fitness within the timespan of 2 months. In contrast, there was large and significant variation in growth rate among snails from
different treatments. The ANOVA Table 1 showed that a major part of this variation could be attributed to the main factors Species and Competition. All other factors and
interactions were of minor magnitude, although some of them were significant.
3.2. Effect of sediment grain size on growth The two sediment-related main factors, Sediment type and Bay, were neither
significant nor important. Thus, if there is a difference in snail growth between these sediment types in nature, this experiment suggests that it is not caused by the sediment
type itself, but with other biotic or abiotic factors that may be correlated to sediment type.
The interaction Species3Sediment type was weak low SS and just significant P 50.050, and the SNK-test did not due to lower power discover any pairwise
differences. It appears that this interaction was caused by H . ventrosa growing slightly
better on coarse-grained sediments than on fine-grained, whereas H . ulvae grew similarly
well on both sediment types. Finally, the interactions C3ST and S3C3ST were non-significant, indicating that different competition intensities had similar impact on
growth on individuals of both species regardless of sediment type.
3.3. Effects of intra- and interspecific competition The interaction Species3Competition and the main factors Species and Competition
were all significant. The interaction Species3Competition is of interest since it is directly related to the different hypotheses about the strength of intra- and interspecific
competition. Although the interaction was significant, it was rather weak. The SNK-test
Table 1 Result of the ANOVA based on data transformed with lnx 11
Factor SS
d.f. MS
F-ratio F
P versus
Species, SP 17.662
1 17.662
SP3BST 669.71
,0.0001 Competition, C
32.500 4
8.125 C3BST
206.48 ,0.0001
Sediment type, ST 0.010
1 0.010
BST 0.48
0.51 Bay, BST
0.289 14
0.021 Residual
0.77 0.69
SP3C 0.898
4 0.224
SP3C3BST 7.93
,0.0001 SP3ST
0.122 1
0.122 SP3BST
4.61 0.05
SP3BST 0.369
14 0.026
Residual 0.99
0.46 C3ST
0.183 4
0.046 C3BST
1.16 0.33
C3BST 2.204
56 0.039
Residual 1.48
0.02 SP3C3ST
0.083 4
0.021 SP3C3BST
0.73 0.58
SP3C3BST 1.584
56 0.028
Residual 1.06
0.37 Residual
8.051 302
0.027 Total
63.954 461
122 J
. Grudemo, T. Bohlin J. Exp. Mar. Biol. Ecol. 253 2000 115 –127
of the interaction showed that H . ulvae grew significantly faster than H. ventrosa at all
competition intensities level 1, 69; level 2, 68; level 3, 329; level 4, 74; level 5, 167. However, the SNK-test showed different grouping of growth of the two species
at different competition levels. In H . ulvae, competition levels grouped as 154.5.2.
3, indicating that growth was similar without competition and with medium competition from H
. ventrosa, and that high competition from H. ventrosa as well as medium and high competition from H
. ulvae reduced growth. Thus, for H. ulvae, intraspecific competition was stronger than interspecific. In contrast, for H
. ventrosa the levels grouped as 154.2.5.3, showing that H
. ventrosa was more influenced by competition from H
. ulvae than from its own species. The significant interaction in the ANOVA was caused by H
. ulvae being less affected by high population density of H. ventrosa, than H
. ventrosa itself Fig. 2. 3.4. Calculation of competition coefficients
Although not primarily designed for the purpose, we also used the results to estimate
Fig. 2. Interaction term Species3Competition. Average growth lngrowth11 of target snails, with standard error bars indicated, is shown for different treatments. j H
. ulvae target snails without competition and with competition from medium and high densities of H
. ulvae. h H. ulvae target snails with competition from H. ventrosa. d H
. ventrosa target snails without competition and with competition from H. ulvae. s H. ventrosa target snails with competition from H
. ventrosa.
J . Grudemo, T. Bohlin J. Exp. Mar. Biol. Ecol. 253 2000 115 –127
123
the quantitative effect of competition on growth by regression, assuming a linear relation Fig. 3 between population density and growth. In this analysis we used growth in mm
not on the logarithmic scale. For each species we first tested the slopes of the regression lines, with growth explained by the factor competition, the regressor
population density, and the interaction between them ANCOVA. The interactions were significant P ,0.0001 for both species, suggesting that the strength of intra- and
interspecific competition were different in both species. We therefore used linear regressions separately for intra- and interspecific competition in each species to test and
estimate the slopes Fig. 3.
The results were:
2
• H
. ulvae: intraspecific competition: n5135, r 50.695, P,0.0001, slope buu5
25 2
29.25310 ; interspecific competition: n 5140, r 50.066, P 50.0013, slope
25
buv5 21.89310 ;
2
• H
. ventrosa: intraspecific competition: n5143, r 50.403, P,0.0001, slope bvv5
Fig. 3. Linear regressions of growth of H . ulvae and H. ventrosa target snails at different population densities.
Filled symbols show growth with intraspecific competition and open symbols growth with interspecific competition. Each symbol is the average growth of 43–48 independent individuals, with standard error bars
2
indicated. The value at 770 m no competition is used in both regressions.
124 J
. Grudemo, T. Bohlin J. Exp. Mar. Biol. Ecol. 253 2000 115 –127
25 2
23.85310 ; interspecific competition: n 5138, r 50.727, P ,0.0001, slope
25
bvu5 27.10310 .
For H . ulvae, interspecific competition had thus less effect on growth than intra-
specific, whereas H . ventrosa suffered more from interspecific than intraspecific
competition. As the relations above between population density and growth were approximately
linear Fig. 3, we used the slopes to calculate Lotka-Volterra LV competition coefficients, for each species, obtained as the ratio slope under interspecific competi-
tion slope under intraspecific competition. The LV-coefficients were thus calculated to a 50.2 [5buv buu] from H. ventrosa on H. ulvae, and b 51.8 [ 5 bvu bvv] from
H
. ulvae on H. ventrosa.
4. Discussion