Kingsford et al.: The influence of elemental chemistry on the widths of otolith increments in Pomacentrus coelestis 
137 
from each tank were randomly selected and frozen 
(-20°C) for microstructural analysis of otoliths. 
The SL of the P. coelestis used in the experiment 
(mean=15.2 mm ±0.07 SE) did not differ between 
treatment groups for any year of the experiment 
(ANOVA: 2000: P 4> 16 =3.31, P>0.05; 2001: F 1 16 =1.43, 
P>0.05; 2002: F x 16 = 2.4, P>0.05) or among years 
(ANOVA: F 2 51 =0.24, P>0.05). The otoliths (the left 
or right sagittae was chosen randomly), were rinsed 
three times in Milli-Q water (Millipore, Billerica, 
MA). and were allowed to dry. Otoliths were polished 
as described above (see Wild fish section. Following 
elemental analysis of the otoliths by means of la- 
ser ablation inductively coupled mass spectrometry 
(for a complete description of the elemental analysis 
see Patterson et al., 2004a), the daily increments 
were counted and measured in the same manner 
as described earlier. Note that all increments were 
counted and measured, not just those pertaining 
to the experiment. Settlement marks, based on the 
criteria defined by Wilson and McCormick (1999), 
were also noted. 
Crystallography 
The orientation of crystals was examined with SEM 
near the sulcus of sectioned sagittae. The surface of 
sectioned otoliths was lightly etched (EDTA 0.1 to 
0.01M for 3 minutes) to observe the orientation of 
the crystal-lattice. Although etching altered the edge of 
crystals, orientation of the long-axis of the crystal-lattice 
was still easily seen (e.g., Linkowski et al., 1993). The 
crystal-lattice was observed for four fish from each treat- 
ment (ocean vs. lagoon) for the year 2000 experiment. 
The crystal-lattice of experimental fish was compared 
with that of 18 control fish. These fish were collected in 
the same light traps as fish used in the experiments, but 
were not subjected to any experimental conditions. 
Statistical analysis 
Data were tested for homogeneity of variances with a 
Cochran’s C-test and were found to be homogeneous. 
Wild fish otolith increments were analyzed with a nested 
ANOVA design with the factors habitat (ocean vs. lagoon) 
and site nested within habitat. However, because site 
nested within habitat was found to be nonsignificant at 
the P = 0.25 level, sites were pooled with the residual 
to increase degrees of freedom and the ANOVA was 
undertaken again with only habitat as a factor (Under- 
wood, 1997). 
To analyze otolith increments corresponding to both 
the pre-experimental (nine days before the start of the 
experiment) and the experimental phase (nine days 
of the experiment), fully nested ANOVAs were used 
for each year of the experiment. This design used two 
factors (treatment and tank nested within treatment). 
Treatment was a fixed factor and tank was a random 
factor. One-factor ANOVAs and Tukey’s HSD post hoc 
tests were used to examine differences among years 
A 
Sites 
B 
Figure 1 
Mean ( + 1 standard error) increment width (pm) in oto- 
liths of nonexperimental neon damselfish ( Pomacentrus 
coelestis) (n = 7 per site) collected from two sites on the 
reef slope in ocean water (Ocean 1 and 2) and within the 
lagoon (Lagoon 1 and 2) at One Tree Island, Great Bar- 
rier Reef, Australia. Tukey’s HSD contrast groupings are 
indicated as letters (i.e. , A, B) where overall differences 
were significant (P<0.05). 
within treatment groups. Paired /-tests were also used 
to examine the relationships of increment widths before 
and during the experiment for each experimental treat- 
ment. Based on the findings of Patterson et al. (2004a), 
correlation analyses were used to examine the relation- 
ships between incremental widths and elemental ratios 
(i.e., Ba/Ca, Sr/Ca). 
Results 
Neon damselfish in natural habitats displayed signifi- 
cant variation in increment widths; increments were nar- 
rower in otoliths of fish collected in the lagoon (ANOVA: 
P 4 26 =14.76, P<0.001; Fig. 1). For experimental fish, 
which all originated in ocean water, no difference in 
incremental spacing was found before the start of the 
experiment by treatment or by tank nested within treat- 
ment for any year of the experiment (ANOVA: 2000: P 4 
4 = 1.34, P>0. 05; 2001: P 4 4 = 0.17, P>0.05; 2002: F x 4 = 
0.04, P>0.05; Fig. 2). In addition, no difference in Ba/Ca 
ratios was found before the experiment (Fig. 3; Patterson 
et al., 2004a). Fish in ocean water in 2000 and 2002 had 
wider increments than those in the lagoon treatment 
groups (ANOVA: 2000: P 4 4 =28.53, P <0.01; 2001: P, 4 = 
1.62, P>0.05; 2002: P, 4 =26.42, P<0.01), as well as a 
significant difference in Ba/Ca ratios (Figs. 2 and 3; 
Patterson et al., 2004a). In addition, there was a signifi- 
cant difference in increment widths among tanks within 
treatments for 2002. A Tukey’s HSD test (P<0.05) indi- 
cated this was due to a tank of lagoon water where fish 
