Journal of Experimental Marine Biology and Ecology 255 (2000) 153±174 www.elsevier.nl/locate/jembe

Evaluating the impact of predation by ®sh on the assemblage structure of ®shes associated with seagrass (Heterozostera tasmanica) (Martens ex Ascherson) den Hartog, and unvegetated sand habitats

Jeremy S. Hindella,b,* , Gregory P. Jenkins c , Michael J. Keough a aDepartment of Zoology, University of Melbourne, Parkville 3010, bQueenscliff Marine Station, Queenscliff 3225, Australia cMarine and Freshwater Resources Institute, Queenscliff 3225, Australia Received 22 November 1999; received in revised form 4 July 2000; accepted 1 September 2000

Abstract

The role of ®sh predation in structuring assemblages of ®sh over unvegetated sand and seagrass was examined using enclosure and exclusion cages to manipulate the abundance of predatory ®sh from November 1998 to January 1999. In our exclusion experiment, piscivorous ®sh were excluded from patches of unvegetated sand and seagrass to measure how they altered abundances of small ®shes, i.e., ®sh , 10 cm in length. Habitats from which piscivorous ®sh were excluded contained more small ®sh than those with partial cages, which in turn contained more ®sh than uncaged areas. These patterns were consistent between unvegetated sand and seagrass areas, although the relative differences between predator treatments varied with habitat. Overall, small ®sh were more abundant in unvegetated sand than seagrass. Atherinids and syngnathids were the numerically dominant families of small ®sh and varied in complex ways amongst habitats and cage treatments. The abundance of atherinids varied inconsistently between cage treatments through time. Only during the ®nal two sampling times did the abundance of atherinids vary signi®cantly across cage treatments. Syngnathids were strongly associated with seagrass and were signi®cantly more abundant in caged than uncaged habitats. In our enclosure experiment, ®ve individuals of a single species of transient piscivorous ®sh, Western Australian (Arripidae: truttacea Cuvier), were enclosed in cages to provide an estimate of the potential for this species to impact on small ®sh. The abundance of small ®sh varied signi®cantly between cage treatments. Small ®sh were more abundant in enclosure cages and exclusion cages than uncaged areas; however, there was no difference in the abundance of small ®sh in enclosure cages and partial cages, and no difference between exclusion cages and partial cages. These patterns were

*Corresponding author. Queenscliff Marine Station, PO Box 138, 3225, Australia. Tel.: 1 61-3-5258- 3686; fax: 1 61-3-5258-3632. E-mail address: [email protected] (J.S. Hindell).

0022-0981/00/$ ± see front matter  2000 Elsevier Science B.V. All rights reserved. PII: S0022-0981(00)00289-6 154 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 consistent amongst habitats. Atherinids and syngnathids were again the numerically dominant families of small ®sh; atherinids varied more with cage structure while syngnathids did not vary statistically between cages, blocks (locations within which a single replicate of each cage treatment was applied) or habitats. Dietary analysis of caged A. truttacea demonstrated the potential for this species to in¯uence the assemblage structure of small ®sh through predation ± atherinids were consumed more frequently in unvegetated sand than seagrass, and syngnathids were consumed only in seagrass, where they are most abundant. Observations of signi®cant cage or predation effects depended strongly on the time at which sampling was undertaken. In the case of the atherinids, no predation or cage effects were observed during the ®rst two sampling times, but cage effects and predation effects strongly in¯uenced abundances of ®sh during the third and fourth sampling times, respectively. Our study suggests that transient piscivorous ®sh may be important in structuring assemblages of small ®sh in seagrass and unvegetated sand, and seagrass beds may provide a refuge to ®shes. But the importance of habitat complexity and predation, in relation to the potentially confounding effects of cage structure, depends strongly on the time at which treatments are sampled, and the periodicity and multiplicity of sampling should be considered in future predation studies.  2000 Elsevier Science B.V. All rights reserved.

Keywords: ; Australia; Unvegetated sand; Caging experiment; Heterozostera tasmanica; Piscivory; Seagrass; Structural complexity; Temperate; Temporal variability

1. Introduction

Predation can be an important process structuring post-settlement assemblages of ®sh (Choat, 1982), and one of the most abundant predators are other ®shes. Correlative studies often show that the abundance of piscivorous ®sh is negatively associated with the abundance of smaller (prey) ®sh (Hixon, 1991; Bailey, 1994; Connell and Kingsford, 1997, 1998). Dietary studies complement correlative analyses by demonstrating the importance of particular suites of small ®sh in the diets of predatory ®shes (Hall et al., 1990; Kingsford, 1992; Edgar and Shaw, 1995a; Connell and Kingsford, 1997; Connell, 1998). In concert, correlative studies and dietary analyses imply that predatory ®sh are important determinants of the structure of small ®sh assemblages. However, few experimental studies have unequivocally concluded that ®sh predation is an important contributor to variability in small ®sh assemblage structure (Hixon, 1991). Habitat structure is provided by both abiotic (consolidated and unconsolidated sediments and rock) and biotic (coral, wood, oyster reef, submerged and emergent vegetation) elements, but vegetation has received most attention due to its wide distribution and because abundances in vegetation are generally greater than alternative, usually unvegetated, areas nearby (Heck and Crowder, 1991). In marine environments, the positive association of with structurally complex, vegetated habitats is likely to be a re¯ection of interactions between habitat selection processes (Edgar and Robertson, 1992; Levin and Hay, 1996), hydrodynamics and larval supply (Jenkins and Black, 1994; Hamer and Jenkins, 1996) and survival as a function of refuge provision and habitat complexity (Choat, 1982; Orth et al., 1984; Orth, 1992). Predation J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 155 is an important source of mortality, and predation ef®ciency, as a function of detection, selection, pursuit and capture of prey (Mattila, 1995), decreases with increasing habitat complexity (Choat, 1982; Stoner, 1982; Gotceitas et al., 1997). Thus, patterns in the abundance of ®sh across habitats which differ markedly in structure may re¯ect differential predation pressure, i.e., the structure of the vegetation and the type of habitat complexity it generates determines the intensity and nature of predator±prey interac- tions, and thereby affects the structuring capacity of predation (Mattila, 1995). Seagrasses are a common form of biogenic habitat in marine and estuarine systems worldwide (Pollard, 1984; Bell and Pollard, 1989; Kemp, 1989), and compared with alternative, usually unvegetated habitat, generally contain higher numbers of predatory and other ®shes (Heck et al., 1989; Edgar and Shaw, 1995a,b; Jenkins et al., 1997b). The high, but temporally variable (Heck et al., 1989), association of small ®sh with seagrass habitat has led to a paradigm which promotes the importance of seagrass beds in the provision of nursery habitat for juvenile ®shes (Pollard, 1984; Bell and Pollard, 1989; Jenkins and Wheatley, 1998). Increased food availability (Bell and Pollard, 1989) and protection from environmental disturbance (Kemp, 1989; Edgar, 1990) are two popular theories why seagrass beds contain high abundances of small ®sh. But, it is also plausible that the structural complexity provided by aspects of the seagrass affects the ef®cacy or selectivity of predators (Levin et al., 1997), and provides juvenile ®sh with a refuge from predation (Orth et al., 1984; Orth, 1992). Previous studies suggest that broad-scale patterns in small ®sh assemblages may be in¯uenced by variable larval supply (Jenkins and Black, 1994; Jenkins et al., 1997a) and, micro-habitat selection, not predation, is the proximate cause for variability in the abundance of fauna within seagrass beds (Bell et al., 1987; Edgar and Robertson, 1992; Levin et al., 1997). However, many of these studies have manipulated predatory ®sh over relatively small (1 m2 ) spatial scales (Bell et al., 1987), which may re¯ect prey movements and behaviour more than predation effects (Englund and Olsson, 1996; Englund, 1997). Furthermore, while exclosure cages have been used extensively to manipulate abundances of predatory ®sh (Steele, 1998; Levin et al., 1997; Kennelly, 1991), few studies have used enclosure cages to assess the role of predatory ®sh in structuring assemblages of ®sh in structurally diverse habitats. A more thorough understanding of (a) the role of ®sh predation in structuring assemblages of small ®sh, and (b) the importance of seagrass beds in the provision of refuge from predation, will be gained by conducting carefully designed experiments which manipulate predator abundance using controlled enclosure and exclosure caging experiments over similar spatial and temporal scales in large plots of habitat (seagrass and unvegetated sand) that differ markedly in structural complexity. The primary aim of our study was to investigate whether predatory ®sh in¯uence small ®sh abundance in structured (seagrass) and unstructured (unvegetated sand) soft sediment habitats by manipulating the presence of piscivorous ®sh using exclusion and enclosure cages. In our enclosure experiment we evaluated the impact of a single species of predatory ®sh (Arripis truttacea) on assemblages of small ®sh over seagrass and unvegetated sand. Our results are discussed in relation to the importance of predatory ®sh, cage effects, habitat characteristics and temporal variability in altering the assemblage structure of ®shes. 156 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

2. Materials and methods

2.1. Study site

Our study was carried out at Grand Scenic, which is situated on the southern shore of the Geelong Arm of Port Phillip Bay (Fig. 1). The currents in this region of Port Phillip Bay are weak, generally much less than 17 cm s21 (Rosenberg et al., 1992), and the area is protected from the prevailing southerly winds. These largely sheltered conditions facilitate sedimentary processes that generate a substrate composed primarily of silty sand (Anon., 1973), rich in detritus (J. Hindell, pers. obs.). Tides in this area are semidiurnal and the range is less than 1 m (Jenkins et al., 1998). Grand Scenic contains large contiguous subtidal beds of Heterozostera tasmanica, whose distribution is broken by patches of unvegetated sand. A variety of algae and small patches of rocky reef occur sporadically throughout the beds of seagrass. One other species of seagrass, Zostera muelleri, Irmisch ex Ascherson, also occurs in this region of Port Phillip Bay, but its distribution is largely con®ned to the intertidal.

2.2. Exclusion experiment

In our exclusion experiment, predatory ®sh were excluded from 16-m2 plots of unvegetated sand and seagrass to test whether predatory ®sh in¯uence the assemblage structure of small ®sh, and whether any observed predator effects are independent of habitat complexity and time of sampling. A variety of predatory ®shes, including pike-headed hardyheads, Kestratherina esox Klunzinger (Atherinidae), Arripis truttacea, yank ¯athead (Platycephalus speculator Klunzinger), and rock ¯athead (Platycephalus laevigatus Cuvier), were potentially excluded using cages, but only K. esox and A. truttacea have been found to consume the small, early post-settlement stages of ®sh sampled in the cage treatments at this location (Hindell et al., 2000). Exclusion cages were constructed from 2.1 m long galvanised steel starpickets, hammered into the substrate at each corner of a 4 3 4-m square plot. Black, 20-mm polypropylene netting of 1.5 m height, was placed around the perimeter of the plot, enclosing an area of 16 m2 (Fig. 2a). This mesh size was chosen because it is small enough to prevent the passage of predatory ®sh. The mesh was tightened at the top and bottom of each cage with 5 mm nylon cord, and the bottom line was weighted to prevent ®sh from swimming between the bottom of the cage walls and the surface of the substrate. The top of the cage walls exceeded the water level throughout the tidal range, however, the substrate within each cage treatment was always submerged. Partial cages were constructed from the same materials and in the same dimensions as exclusion cages, except only half of each wall of the cage was attached (Fig. 2b). Each partial cage provided the structure of a cage but did not change the abundance of predatory ®shes inside partial enclosures relative to uncaged areas (J. Hindell unpublished data). These cages were used to assess the role of cage structure per se in altering ®sh assemblages. Uncaged areas were unmanipulated 4 3 4-m plots of habitat. Four locations (blocks) were selected randomly at Grand Scenic, i.e., by dividing this location into blocks of 50 3 50 m and using a random number generator to select each J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 157

Fig. 1. Location of study site, Grand Scenic, in the Geelong arm of Port Phillip Bay. Insets: location of Port Phillip Bay in Australia and location of study region in Port Phillip Bay.

block. Within each location, the caging treatments (exclosure cage, partial cage±cage control, and uncaged), were each applied to haphazardly chosen patches of unvegetated sand and seagrass within randomly chosen locations (blocks) along the shore. Our 158 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

Fig. 2. Design of (a) exclusion/enclosure cages and (b) cage controls used to manipulate the abundance of predatory ®sh. experiment was arranged in a completely randomised block design with ¯ 10 m between cage/habitat treatments within blocks, and ¯ 50 m between blocks. Following construction, cages were left for 1 week before sampling. Each cage was cleaned weekly to reduce the build up of drift algae which interferes with water movement and may attenuate light (Virnstein, 1978). Cages remained in the ®eld for 1 month. Little algae grew on cage walls and the seagrass did not appear to be in¯uenced detrimentally by excessive sedimentation or overgrowth from epiphytes (J. Hindell, pers. obs.). Recent studies have shown that there is no difference in the particle size distribution, amounts of combustible organic matter, or composition of meiofauna between exclusion and partial cages (J. Hindell, pers. obs.).

2.3. Sampling small ®sh

In our study, small ®sh refers to ®sh that are generally less than 5 cm in length. Most of the small ®sh were juveniles. Only individuals from the family Syngnathidae, which were included in the small ®sh category, were commonly adults and their length often exceeded 10 cm. The assemblages of small ®sh in each treatment were sampled on the J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 159 same day during low tide using a modi®ed beach seine net (4 m wide 3 1.5 m high 3 1.5 m deep 3 1-mm mesh). In this study, a pole was attached to each side of the net to aid in hauling and foam ¯oats and lead weights were attached to the top and bottom of the net, respectively. The net was drawn between two steel posts inside and at one end of the cage, and hauled through to the opposite side of the cage by two people, one person holding each pole. Because pilot studies showed that roughly 90% of all small ®sh were captured in the ®rst haul of this type of net, regardless of habitat, only a single haul of the net was conducted in each experimental arena. Captured ®sh were anaesthetised in benzocaine and preserved in 70% ethanol. This sampling procedure was repeated once weekly for four consecutive weeks. In the laboratory, ®sh were counted and identi®ed to species (Gomon et al., 1994).

2.4. Enclosure experiment

In our second experiment, enclosure cages, each containing juvenile Arripis truttacea were added to the experimental design used in our exclosure experiment and the experiment was re-conducted using reconstructed cage treatments. Juvenile A. truttacea, whose total length at the study location ranges between 10 and 15 cm (J. Hindell, pers. obs.), are transient and gregarious, and commonly occur in shallow water over mosaics of seagrass and rocky reef interspersed with patches of unvegetated sand. A. truttacea are perennial in this type of habitat, and at the time of year our study was conducted, they feed voraciously on early post-settlement ®shes, particularly atherinids (Hindell et al., 2000). During winter and early spring, when small ®shes are generally less abundant, juvenile A. truttacea consume a range of pelagic invertebrates, the most common of which are crustaceans of the order Mysidacea (J. Hindell, pers. obs.). A. truttacea was chosen to enclose in cages because they are robust to handling stress, easy to catch and maintain, and previous research has shown that their abundances are negatively related to local abundances of juvenile atherinids and sillaginids (Hindell et al., 2000). The structure and dimensions of enclosure cages were identical to the exclusion cages. In our enclosure experiment, each cage type (exclusion cage, enclosure cage, partial cage and uncaged) was applied to haphazardly chosen unvegetated sand and seagrass plots within each of four blocks. The block component of the enclosure experiment incorporated both temporal (amongst weeks) and spatial (position along the shore) variability, and each block of the experiment (orthogonal combination of cage and habitat) was conducted independently and successively. After the construction of each block, ®ve juvenile Arripis truttacea (each , 15 cm total length), approximately the ambient ®eld density of this species at this location (J. Hindell, pers. obs), were added to each enclosure cage. A. truttacea were captured 2 days before being placed in enclosure cages and were maintained in 300-l ¯owing- seawater aquaria at the Queenscliff Marine Station. After 2 days of con®nement, the A. truttacea, as well as the small ®sh assemblage, in each combination of habitat and cage, within a block, were sampled using the net and methods described earlier. Short periods of enclosure of predatory ®sh were chosen to mimic the temporal patchiness of A. truttacea in the ®eld. All ®sh were anaesthetised and preserved in 70% ethanol. Small ®sh were counted and identi®ed to family using Gomon et al. (1994). 160 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 nvegetated b b bb bb bbbb b b bb Exclosure experiment a ±2±±±± ±± ± ± 4±625±±±±± 2±±±±±±±± ± ± 1±4±±±4±±±± ± ± ±±2±±±2±±±± 1 ± 1 ±± 1 4 2±±±±±±± ± ±± ± ± 04 0417±± ± ± ± ± 40142117±± 1 16 94014± 2 ±±22±±±±±1 ±±±±±±±±± ± ± ±± ±± ± 1 ± ±±±±± 2 3±17± ±36±±3±±±±±±±± ± ±± ±±±±±±±±17± 1 SeagrassC CC PC UC Bare sand C CC PC UC Seagrass C CC UC Bare sand C CC UC p12±±±±±±±± ± 6± ± ± sp.1 Leptatherina presbyteroides Atherinid recruitsClinidae Heteroclinus perspicilatus Gobiidae 4Monacanthidae ±Brachaluteres jacksonianus ±Odacidae Platycephalidae 25Neoplatycephalus aurimaculatus 4 ± ± ± 93 86 15 13 12 3 Table 1 Percent abundance of smallsand), ®shes in pooled each across regime sampling of cage timesSpecies (C, and exclusion blocks, cage; for CC, the cage control; enclosure PC, and enclosure exclusion cage; experiments UC, uncaged) and habitat (seagrass, u Atherinidae Atherinasoma microstoma Ketstratherina esox Enclosure experiment Favonigobius lateralis Nesogobius Mugilidae Aldrichetta forsteri Acanthaluteres spilomelanurus Meuschenia freycinetti Neoodax balteatus J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 161 b b b bb b b 3548746 10108121213 3±31±1 1 1 1 2 23±±23±17± 2±±±2±5±± ± ±± ±4 ± 10 ± ± 4 10 10 17 1 1 15 12 24 30 7±57±±± ±± ±±±625±±±±1224±±± ± ±±±±±±±±± ± ±± ±±±±±±±±±± ± ± 3±±±±±±±±±± 1 ±177±25177±±± 1 ±± ± ± 25±±±±±±±±± 58 ± ±± ± ± 61 ± 25 57 63 17 3 8 33 65 48 52 0.5. , Percent abundance All data are rounded to whole numbers. a b Scorpaenidae Sillaginidae Syngnathidae Lissocampus caudalis Lissocampus runa Stigmatopora argus Urocampus carinirostris Vanacampus phillipi Total number of ®shMean number per cageNumber of species 56 17 16 8 5 18 6 4 1 52 17 29 10 19 6 6 2 1543 96 720 45 71 549 4 366 34 273 23 17 Pleuronectidae Rhombosolea tapirina Gymnapistes marmoratus Sillaginodes punctata Stigmatopora nigra Vanacampus margaritifer 162 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

The stomach of each re-captured Arripis truttacea was excised and the gut contents were categorised and counted. The dietary composition of A. truttacea was described using percent frequency of occurrence (F ), percent mass (M), and percent abundance (N) (Hyslop, 1980).

2.5. Statistical analysis

The exclusion experiment was analysed as a repeated measures (time), three factor (block, habitat and cage) randomised blocks design. Habitat and cage were treated as ®xed factors. Block was treated as a random factor and time was the repeated factor. Raw data were log(x 1 1) transformed where the assumptions of homogeneity of variances and normality were not met. The assumption of sphericity was checked by the Greenhouse-Geisser (G-G) epsilon value (e). The potential for sphericity to in¯uence our results was controlled by using the G-G adjusted probability (P) values, however, where the adjusted P value did not alter the signi®cance of the un-adjusted P value, the un-adjusted P value was used. A priori tests were used to determine how the levels of the cage effect varied. Where the number of a priori tests exceeded the degrees of freedom (df) for the effect being tested, the signi®cance level (a) was adjusted to control for the experimentwise Type I error rate by dividing the signi®cance level for that test (0.05) by the number of comparisons in excess of the df for the effect being tested. This gave a critical level of a/(no. planned comparisons 2 df). Where interactions were found between a main effect and time, separate one factor ANOVAs and a priori tests were conducted for each time to determine where the levels of the interacting main effect varied. The enclosure experiment was analysed as a three-factor (block, habitat, cage) randomised blocks design. Tests for assumptions and comparisons of main effects were carried out as described for our exclosure experiment. All analyses were carried out  using SYSTAT statistical software (Wilkinson et al., 1992).

3. Results

3.1. Exclusion experiment

A diverse assemblage of small ®sh from 10 families and 22 species was sampled throughout our study (Table 1). Small ®sh were in¯uenced both by manipulating predatory ®sh using cages, and by the habitat within which this manipulation was conducted, but habitat and caging acted independently (Table 2). There was signi®cant variability in the abundance of small ®sh amongst cage types (Table 2 and Fig. 3). The abundance of ®sh in uncaged areas was signi®cantly lower than either exclusion cages (P 5 0.001) or cage controls (P 5 0.012). Cage controls contained signi®cantly fewer ®sh than exclusion cages (P 5 0.034). The abundance of small ®sh also varied signi®cantly amongst habitats and times (Table 2 and Fig. 3). More ®sh were captured over unvegetated sand than seagrass, and the abundance of small ®sh, regardless of cage and habitat, varied signi®cantly across sampling times J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 163

Table 2 Three factor repeated measures analysis of variance comparing the numbers of small ®sh (total), atherinids and syngnathids amongst blocks, habitats (seagrass versus unvegetated sand) and cage types (exclusion cage, cage control and uncaged) (n 5 24)a Source Num df Den df Small ®sh Atherinids Syngnathids Between subjects Block (B) 3 6 0.141 0.806 0.015 Habitat (H) 1 3 0.048 0.378 , 0.001 Cage (C) 2 6 0.002 , 0.001 0.039 B 3 H 3 6 0.688 0.157 0.877 B 3 C 6 6 0.911 0.781 0.725 H 3 C 2 6 0.299 0.116 0.764 Residual MS 6 1.966 1.108 1.017 Within Subjects Time (T) 3 9 , 0.001 0.004 0.119 T 3 B 9 18 0.999 0.608 0.133 T 3 H 3 9 0.061 0.115 0.120 T 3 C 6 18 0.057 0.010 0.489 T 3 B 3 H 9 18 0.051 0.523 0.534 T 3 B 3 C 18 18 0.262 0.911 0.204 T 3 H 3 C 6 18 0.076 0.411 0.333 Residual MS 18 0.646 1.634 0.444 a The data table shows, for each group of small ®sh analysed, the probabilities associated with each of the terms in the model (Source) and the Residual MS. This information allows full reconstruction of the original ANOVA table. Data were log(x 1 1) transformed prior to analysis.

(Table 2 and Fig. 3). Although the main effects (block, habitat and cage) acted independently, there were near signi®cant interactions between time and habitat, time and cage, and time and the habitat 3 block interaction (Table 2 and Fig. 3). Atherinidae and Syngnathidae were the numerically dominant families of small ®sh in this experiment (Table 1); therefore, these two families of ®sh were analysed separately. Despite a trend suggesting otherwise, the abundance of atherinids did not vary between habitats (Table 2 and Fig. 3). Additionally, the abundance of atherinids varied inconsistently between cage treatments through time (Table 2 and Fig. 3). During the ®rst and second sampling times, the abundance of atherinids did not vary between cage treatments (df2,18 , MS 5 0.761, P 5 0.427) and (df 2,18 , MS 5 0.876, P 5 0.378), respec- tively. During the third sampling time, the abundance of atherinid recruits at Grand Scenic increased substantially (J. Hindell, pers. obs.). This event corresponded with signi®cant variability in the abundance of atherinids, now predominantly juvenile ®sh, across cage treatments for the third and fourth sampling times (df2,18 , MS 5 14.319, P , 0.001) and (df2,18 , MS 5 13.194, P , 0.001), respectively. In the third sampling time the abundance of atherinids did not vary signi®cantly between exclusion cages and cage controls (P 5 0.748), however, uncaged areas contained signi®cantly fewer atherinids than exclusion cages (P , 0.001) and cage controls (P , 0.001). In the ®nal sampling time exclusion cages contained signi®cantly more atherinids than cage controls (P 5 0.001) and uncaged areas (P , 0.001). Cage controls and uncaged areas contained similar numbers of atherinids (P 5 0.180). 164 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

Fig. 3. Mean abundance of small ®sh, atherinids and syngnathids in unvegetated and seagrass habitats during each sampling time in each caging treatment (exclusion cage, cage control and uncaged).

Syngnathids showed a strong association with patches of seagrass; signi®cantly more syngnathids occurred in seagrass than unvegetated sand (Table 2 and Fig. 3). Syngnathids also varied signi®cantly between cage treatments (Table 2), but this result was driven primarily by the much higher abundance of syngnathids in exclusion cages compared to uncaged areas (P 5 0.015), particularly in seagrass (Fig. 3). The abundance of syngnathids did not vary signi®cantly between partial cages and either exclusion J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 165 cages (P 5 0.294) or uncaged areas (P 5 0.069), nor across times (Table 2, Fig. 3). While the abundance of syngnathids varied with the position along the shore (block) (Table 2), we were more interested in the variance component and the subsequent reduction in residual variation associated with the block effect, than speci®c differences between groups. Therefore, no further analysis was conducted for this effect.

3.2. Enclosure experiment

The abundance of small ®sh did not vary between habitat types or blocks (Table 3 and Fig. 4a). However, the abundance of small ®sh did vary across cage treatments (Table 3); uncaged areas contained signi®cantly fewer ®sh than either exclusion cages (P 5 0.001), cage controls (P 5 0.005) and predator enclosures (P 5 0.012) (Fig. 4a). Small ®sh did not vary between cage controls and predator enclosures (P 5 0.464); however, there were signi®cantly more small ®sh sampled from exclusion cages than predator enclosures (P 5 0.021). The abundance of small ®sh almost varied signi®cantly between cage controls and exclusion cages (P 5 0.058). When the level of signi®cance was adjusted to control for the experimentwise Type I error rate, only the difference in the abundance of small ®sh between exclusion cages and cages containing predators became non-signi®cant. As in the exclusion experiment, syngnathids and atherinids were the numerically dominant families of small ®sh in our enclosure experiment (Table 1), and separate analyses were conducted for each of these families. The atherinids showed similar patterns to the total ®sh in that the abundance of atherinids did not vary with block or habitat (Table 3 and Fig. 4b), but varied signi®cantly across cage treatments (Fig. 4b). Exclusion cages contained signi®cantly more atherinids than either cage controls (P 5 0.029), predator enclosures (P 5 0.018) or uncaged areas (P 5 0.007). There was no signi®cant difference in the abundance of atherinids between uncaged areas and predator enclosures (P 5 0.465) or cage controls (P 5 0.292), nor between cage controls and predator enclosures (P 5 0.722). After the signi®cance level was adjusted to control for Type I errors (P 5 0.016), only a

Table 3 Three factor analysis of variance comparing the numbers of small ®sh (total), atherinids and syngnathids amongst blocks, habitats (seagrass, unvegetated sand) and cage types (exclusion cage, cage control, enclosure cage and uncaged) (n 5 24)a Source Num df Den df Small ®sh Atherinids Syngnathids Block (B) 2 6 0.587 0.084 0.121 Habitat (H) 1 2 0.212 0.783 0.121 Cage (C) 3 6 0.003 0.029 0.096 B 3 H 2 6 0.696 0.796 0.553 B 3 C 6 6 0.818 0.798 0.359 H 3 C 3 6 0.932 0.989 0.953 Residual MS 6 0.762 0.943 0.436 a The data table shows, for each group of small ®sh analysed, the probabilities associated with each of the terms in the model (Source) and the Residual MS. This information allows full reconstruction of the original ANOVA table. Data were log(x 1 1) transformed prior to analysis. 166 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

Fig. 4. Mean abundance of (a) small ®sh, (b) atherinids and (c) syngnathids in the enclosure experiment for each cage treatment (exclusion cage, uncaged, cage control, predator enclosure) within unvegetated sand and seagrass habitat (pooled across blocks). J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 167

Table 4 The percent abundance (N), percent mass (M), and percent frequency of occurrence (F ) of dietary categories from stomach contents of Arripis truttacea (n 5 15) enclosed in cages within unvegetated sand and seagrass habitats Prey items Bare sand Seagrass NMFNMF Fish Atherinidae 5 12 13 3 2 7 Syngnathidae ±±±547 Unknown ®sh 3 ±a 13 ± ± ± Other Crustaceans 92 87 60 89 93 40 327±±± a Value , 0.5. signi®cant difference in the abundance of small ®sh between exclusion cages and predator cages existed, despite trends which suggest otherwise (Fig. 4b). The syngnathids in the enclosure experiment displayed similar patterns to those seen for syngnathids in the exclosure experiment (Figs. 3 and 4c). Despite the trends (Fig. 4c), the abundance of syngnathids did not vary statistically across cages, blocks or habitats (Table 3 and Fig. 4c).

3.3. Dietary analysis of enclosed Arripis truttacea

In seagrass habitats, 40% of Arripis truttacea contained no food compared with 53% of A. truttacea enclosed over unvegetated sand. Of the A. truttacea with stomachs containing prey, crustaceans were the most common dietary item, and represented 92 and 87% abundance (N) and percent mass (M), respectively. Teleost prey, which included atherinids, syngnathids and unknown ®sh remains (Table 4), appeared to be a more important component in the diets of A. truttacea enclosed over unvegetated sand than seagrass (Table 4). While atherinids were consumed in both unvegetated sand and seagrass, syngnathids were consumed only in seagrass, the habitat within which they occur most commonly.

4. Discussion

Fish assemblages vary over a range of spatial and temporal scales (Kingsford, 1998). While the importance of choosing an appropriate spatial scale to study ®sh is well documented (Chesson, 1998; Sale, 1998), very little attention is paid to the role of small-scale temporal variability in experimental studies. Speci®cally, ®sh assemblages vary over relatively short time scales, i.e., tidal and diel cycles (Robertson, 1980; Sogard and Able, 1994; Gibson et al., 1996), and therefore, the associated biological interac- tions, such as predation, are also likely to vary over shorter time scales (Laprise and Blaber, 1992). But many studies make only a single recording of ®sh assemblage 168 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 structure at the end of an experiment, even though selecting the appropriate duration of an experiment is crucial to the nature of impacts observed (Minello, 1993), and therefore, may `lose' information about how effects may change through time. In our study, the time at which our experiment was sampled had a pronounced effect on not only whether differences between cage treatments consistent with predation effects were observed, but whether these differences were confounded by cage effects. For example, during the ®rst two sampling times in the exclusion experiment, the mean numbers of atherinids were similar between cage treatments, and therefore ®sh were responding neither to the structure of the cage treatments nor predation. At the third sampling time, uncaged areas contained signi®cantly fewer ®sh than exclusion cages or cage controls, and this suggested that atherinids were responding to the cage structure rather than predation. At the ®nal sampling time, exclusion cages contained more ®sh than either cage controls or uncaged areas, which contained similar numbers of ®sh, and this clearly demonstrated that predation by ®sh, not some artefact associated with the structure of the cage, was important in determining the abundances of atherinids. Clearly, sampling the same experimental units at short time intervals (several days) strongly in¯uences the relative variability in the numbers of ®sh between caging treatments. However, it is unclear whether these results represent temporal variability in predation and cage effects, or whether they simply re¯ect an initial prey response to structure, with a later-emerging predation effect. Replicated experiments that evaluate the consistency of results from caging studies over variable lengths of time, and over the same temporal scales at different times, are needed to separate these alternatives. Irrespective of how the numbers of ®sh varied between cage treatments in our study, this work clearly demonstrates that a strong effect of piscivorous ®shes can be measured after 4 weeks, and this effect depends on the habitat within which it was measured. Investigations into the impacts of predatory ®sh on their teleost prey have traditionally used exclusion cages to manipulate predator abundance (Doherty and Sale, 1985; Kennelly, 1991; Steele, 1998). But interpreting the importance of predation relative to cage effects can be problematic because of the dif®culty in creating a suitable control, i.e., a structure that has all the physical effects of a cage but does not keep out predators (Virnstein, 1978). Preliminary analysis of our caging experiment suggested that predation was decreasing abundances of ®sh outside of areas enclosed by a cage, i.e., small ®sh abundances were greater inside exclusion cages than uncaged areas. However, the intermediate numbers of ®sh associated with cage controls implied a cage effect, i.e., differential attraction of ®sh to structure provided by exclusion and partial cages and/or intermediate physical effects of the cage controls Ð partial interference with predatory ®shes. The use of enclosure cages, in association with exclusion and partial cages, enabled us to further evaluate predator effects without confounding results with variable levels of cage structure, i.e., the controls and treatments have the same amounts of arti®cial structure Ð although experiments in which predatory ®sh are enclosed have been criticised because, even though relatively large enclosures allow more normal behaviour by predators (Virnstein, 1978), the contrived conditions under which predatory ®sh are enclosed may alter their behaviour and generate `unreal' impacts. In our enclosure/exclusion experiment, signi®cantly more ®sh were sampled from within J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 169 partial and enclosure cages, both of which contained similar numbers of small ®sh, than uncaged areas. And there was a trend for more ®sh to be associated with exclusion cages than either enclosure cages or cage controls. These data suggest that either predation pressure is similar inside enclosure and partial cages, or the various effects of these two treatments balance. Recent experiments using video recorders to quantify local abundances of predatory ®sh have shown that the design of the cage controls used in this study neither in¯uences the ambient densities or movement of Arripis truttacea, nor attracts juvenile ®shes (J. Hindell, unpublished data). Therefore, the predation pressure inside partial cages is likely to be similar to that exerted in areas without any form of cage structure, and the structure per se of cage treatments does not appear to be important in attracting juvenile ®shes. Furthermore, enclosed A. truttacea were observed to swim and feed uninhibited Ð they consumed similar categories of prey to those published elsewhere (Robertson, 1982; Hoedt and Dimmlich, 1994). Given that predation impact is similar in enclosures and partial cages, we contend that partial cages may be a form of intermediate protection for small ®sh, and differences between partial and exclusion cages are more likely to re¯ect predation and not simply a linear increase in ®sh attracted to additional arti®cial structure per se. Because partial cages may not provide an unambiguous test for cage effects, the results from caging studies which exclude predators should be augmented with additional data which elucidates the effects of cage structure (Connell, 1997). Tradition- al measures of cage effects, such as organic content, particle size distribution and meiofaunal abundances in sediments were unaffected by the design of the cages and controls used in this study (J. Hindell, unpublished data); however, a more ecologically meaningful alternative to assessing cage artefacts, especially in relation to the attraction of juvenile ®shes, may be through measuring the variability in abundances of ®sh between cage treatments that have a strong af®nity with habitat structure. Sygnathids are strongly associated with structurally complex habitats (Gomon et al., 1994), but can be found in unstructured habitats, such as unvegetated sand, where drift algae occurs. Additionally, the general length of syngnathids measured in our study are rarely consumed by Arripis truttacea, therefore, differences between cage structures are more likely to re¯ect habitat selection rather than predation, and these ®sh may be useful in evaluating the role of cage effects related to the facilitation of structure. In our experiment, unvegetated sand habitats contained larger numbers of syngnathids where cages and cage controls provided structure, but there was no difference in the numbers of syngnathids between cage treatments. This implies that there may be some threshold level of structure (see Gotceitas and Colgan, 1989), as re¯ected in our cage controls, above which, additional structure is less important in facilitating increases in syngnathids. Our results suggest that the additional structure presented in exclusion cages, as compared to partial cages, does not further attract syngnathids. Therefore, the provision of structure per se, not necessarily the amount, may be more important in determining ®sh abundances, and the differences in numbers of ®sh between our cage controls and exclusion cages are more likely to re¯ect predation than simply attraction of ®sh. 170 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

Further data are required which (a) evaluate how the predation pressure imposed by con®ned ®sh differs from that imposed by uncaged ®sh, and (b) establish the importance of varying levels of structure in determining ®sh abundances. As Virnstein (1980) commented, a single line of evidence by itself is weak and, therefore, a pluralistic approach to caging experiments, i.e., exclusion and enclosure cages in concert with other measures of cage effects, potentially provides researchers with a rigorous evaluation of the importance of predation versus cage effects. On the basis of the syngnathid data and the results from the enclosure/exclusion experiments, the patterns in ®sh abundances between cage treatments can be interpreted as representing discernible contributions by both cage and predation effects. Biomass of vegetation is positively associated with the abundance and diversity of animals (Pollard, 1984; Bell and Westoby, 1986a,b; Edgar, 1990; Edgar and Robertson, 1992; Connolly, 1994). In marine environments, ®sh faunas associated with seagrass are often more diverse and abundant than those in nearby unvegetated soft sediments (Pollard, 1984). While food resources, level of physical structure, number of mi- crohabitats, reduction of environmental disturbance and stabilisation of sediments may help to explain these patterns (Lewis, 1984), the mediation of predation by aspects of the seagrass may also be important in determining ®sh abundance (Orth et al., 1984). If predation by ®sh is important in structuring assemblages of small ®sh, and structural complexity mediates this predation, then we can expect predation pressure to be greatest in areas where structural complexity is low, such as unvegetated sand. When predators are excluded from habitats which differ in structural complexity, the effect of excluding predatory ®sh should be less in seagrass, whose structural complexity interferes with foraging by ®sh. And if predation produced the pattern of greater ®sh density in seagrass than unvegetated sand habitats, then one would predict a greater effect of predators and a larger relative change in prey abundances in unvegetated sand than seagrass (as shown by Summerson and Peterson, 1984). Our results suggest that patterns are at least partially explained by predation. Small ®sh under pressure from ®sh predation (no cage) are more abundant in seagrass than unvegetated sand. When predators are excluded, the relative increase in small ®sh is greater over unvegetated sand than seagrass, suggesting that the structural complexity generated by aspects of the seagrass may be mediating predation. However, regardless of habitat, the numbers of ®sh varied between cage treatments in qualitatively similar ways. These results suggest therefore, that while predation in¯uences abundances of ®sh in both seagrass and unvegetated sand habitats, it is relatively more important in habitats without any form of refuge. Therefore, we consider that a substantial portion of the variability in ®sh abundance between seagrass and unvegetated sand may be related to differential predation pressure between these two habitats. If the only role of habitat complexity is to mediate predation, then we would expect the numbers of ®sh associated with habitats of variable structural complexity to be similar when released from predation. But our results showed that more ®sh occurred in areas of unvegetated sand than seagrass from which predators were excluded. Therefore, ®sh may prefer alternative, unvegetated sand habitats, but their distributions are restricted to structurally complex seagrass habitats which, in addition to the processes J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 171 suggested by Lewis (1984), provide a refuge from predation. Bologna and Heck (1999) found that scallops experience highest predation pressure in the areas favourable to growth, and ®sh may alter their foraging patterns amongst structurally variable habitats in the presence of predators (Gotceitas and Brown, 1993). Whether our results were due to the selection of habitats which either offer refuge or have low numbers of predators, or whether they were the result of direct mortality, i.e., being eaten, is not known. Theoretically, either alternative is possible Ð Levin et al. (1997) showed that predatory ®sh can directly alter the population structure of ®sh by consuming small individuals, while Jordan et al. (1996) demonstrated that behaviourally mediated predator avoidance modi®es the habitat use in pin®sh. Dietary analysis of caged Arripis truttacea and extensive dietary data on predatory ®sh at Grand Scenic (Hindell et al., 2000) demonstrate the potential for predation, but further research is required to separate the importance of direct predation versus anti-predator behaviour.

5. Conclusion

Our study cautiously suggests that piscivorous ®sh in¯uence the numbers of ®sh within and between seagrass and unvegetated habitats, and therefore, predation by ®sh may be a signi®cant contributor to the widely documented variability in ®sh abundances observed between vegetated and unvegetated habitats. The changes in abundance of small ®sh in areas of seagrass and unvegetated sand where predatory ®sh were excluded suggest that habitat complexity may mediate predation. However, the relative impor- tance of predation versus antipredator behaviour in generating patterns in assemblages of small ®sh is equivocal, and further research is required to determine how patterns are generated. The importance of cage versus predation effects may be evaluated by manipulating predatory ®sh with a combination of enclosure, exclosure and partial cages. Additional information about how cage effects per se in¯uence abundances of ®sh may be gleaned from descriptions of how the numbers of ®sh that associate positively with habitat complexity vary in relation to different levels of arti®cial structure. Because, at least in our experiment, whether or not predation and/or cage effects are observed depends on the time at which the treatments in an experiment are measured, multiple sampling of experimental treatments will enhance our understanding of the dynamic nature of predation impacts in estuarine mosaics of unvegetated sand and seagrass habitats.

Acknowledgements

This manuscript was greatly improved by comments from S. Connell, C. Styan, J. Carey, and two anonymous reviewers. Thanks also to M. Hendricks, R. Watson, L. McGrath and M. Wheatley for assistance in the ®eld and at the research station. This research was funded by a Melbourne Research Scholarship and was conducted using the facilities at the Queenscliff Marine Station. [AU] 172 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

References

Anon., 1973. Environmental study of Port Phillip Bay. Report on phase one, 1968±1971. Melbourne and Metropolitan Board of Works, and Fisheries and Wildlife Department of Victoria, Melbourne. Bailey, K.M., 1994. Predation on juvenile ¯at®sh and recruitment variability. Neth. J. Sea Res. 32 (2), 175±189. Bell, J.D., Pollard, D.A., 1989. Ecology of ®sh assemblages and ®sheries associated with seagrasses. In: Larkum, A.W.D., McComb, A.J., Shepherd, S. (Eds.), Biology of Seagrasses: A Treatise on the Biology of Seagrasses with Special Reference to the Australian Region. Elsevier, Amsterdam, pp. 565±609. Bell, J.D., Westoby, M., 1986a. Abundance of macrofauna in dense seagrass is due to habitat preference, not predation. Oecologia 68, 205±209. Bell, J.D., Westoby, M., 1986b. Variation in seagrass height and density over a wide spatial scale: effects on common ®sh and decapods. J. Exp. Mar. Biol. Ecol. 104, 275±295. Bell, J.D., Westoby, M., Steffe, A.S., 1987. Fish larvae settling in seagrass: do they discriminate between beds of different leaf density? J. Exp. Mar. Biol. Ecol. 111, 133±144. Bologna, P.A.X., Heck, K.L., 1999. Differential predation and growth rates of bay scallops within a seagrass habitat. J. Exp. Mar. Biol. Ecol. 239, 299±314. Chesson, P., 1998. Spatial scales in the study of reef ®shes: a theoretical perspective. Aust. J. Ecol. 23, 209±215. Choat, J.H., 1982. Fish feeding and the structure of benthic communities in temperate waters. Annu. Rev. Ecol. Syst. 13, 423±449. Connell, S.D., 1997. Exclusion of predatory ®sh on a coral reef: the anticipation, pre-emption and evaluation of some caging artefacts. J. Exp. Mar. Biol. Ecol. 213, 181±198. Connell, S.D., 1998. Patterns of piscivory by resident predatory ®sh at One Tree Reef, Great Barrier Reef. Mar. Freshwater Res. 49, 25±30. Connell, S.D., Kingsford, M.J., 1997. The utility of descriptive information for assessing the impact of coral reef piscivores on their prey. Proc. 8th Int. Coral Reef Symp. 1, 999±1004. Connell, S.D., Kingsford, M.J., 1998. Spatial, temporal and habitat-related variation in the abundance of large predatory ®sh at One Tree Reef, Australia. Coral Reefs 17, 49±57. Connolly, R.M., 1994. Removal of seagrass canopy: effects on small ®sh and their prey. J. Exp. Mar. Biol. Ecol. 184, 99±110. Doherty, P.J., Sale, P.F., 1985. Predation on juvenile coral reef ®shes: an exclusion experiment. Coral Reefs 4, 225±234. Edgar, G.J., 1990. The in¯uence of plant structure on the species richness, biomass and secondary production of macrofaunal assemblages associated with Western Australian seagrass. J. Exp. Mar. Biol. Ecol. 137, 215±240. Edgar, G.J., Robertson, A.I., 1992. The in¯uence of seagrass structure on the distribution and abundance of mobile epifauna: pattern and process in a Western Australian Amphibolis bed. J. Exp. Mar. Biol. Ecol. 160, 13±31. Edgar, G.J., Shaw, C., 1995a. The production and trophic ecology of shallow-water ®sh assemblages in southern Australia. II. Diets of ®shes and trophic relationships between ®shes and benthos at Western Port, Victoria. J. Exp. Mar. Biol. Ecol. 194, 83±106. Edgar, G.J., Shaw, C., 1995b. The production and trophic ecology of shallow-water ®sh assemblages in southern Australia. I. Species richness, size-structure and production of ®shes in Western Port, Victoria. J. Exp. Mar. Biol. Ecol. 194, 53±81. Englund, G., 1997. Importance of spatial scale and prey movements in predator caging experiments. Ecology 78 (8), 2316±2325. Englund, G., Olsson, T., 1996. Treatment effects in a stream ®sh enclosure experiment: in¯uence of predation rate and prey movements. OIKOS 77, 519±528. Gibson, R.N., Robb, L., Burrows, M.T., Ansell, A.D., 1996. Tidal, diel and longer term changes in the distribution of ®shes on a Scottish sandy beach. Mar. Ecol. Prog. Ser. 130, 1±17. Gomon, M.F., Glover, J.C.M., Kuiter, R.H., 1994. In: The Fishes of Australia's South Coast. State Print, Adelaide. J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174 173

Gotceitas, V., Colgan, P., 1989. Predator foraging success and habitat complexity: quantitative test of the threshold hypothesis. Oecologia 80, 158±166. Gotceitas, V., Brown, J.A., 1993. Substrate selection by juvenile Atlantic cod (Gadus morhau): effects of predation risk. Oecologia 93, 31±37. Gotceitas, V., Fraser, S., Brown, J.A., 1997. Use of eelgrass beds (Zostera marina) by juvenile Atlantic cod (Gadus morhua). Can. J. Fish. Aquat. Sci. 54, 1306±1319. Hall, S.J., Raffaelli, D., Basford, D.J., Robertson, M.R., Fryer, R., 1990. The feeding relationships of the larger ®sh species in a Scottish sea loch. J. Fish Biol. 37, 775±791. Hamer, P.A., Jenkins, G.P., 1996. Larval supply and short term recruitment of a temperate zone demersal ®sh, the , Sillaginodes punctata Cuvier, to an embayment in south-eastern Australia. J. Exp. Mar. Biol. Ecol. 208, 197±214. Heck, K.L., Crowder, L.B., 1991. Habitat structure and predator-prey interactions in vegetated aquatic systems. In: Bell, S.S., McCoy, E.D., Mushinsky, H.R. (Eds.), Habitat Structure: the Physical Arrangement of Objects in Space. Chapman and Hall, London, pp. 281±299. Heck, K.L., Able, K.W., Fahay, M.P., Roman, C.T., 1989. Fishes and decapod crustaceans of Cape Cod eelgrass meadows: species composition, seasonal abundance patterns and comparison with unvegetated substrates. Estuaries 12, 59±65. Hindell, J.S., Jenkins, G.P., Keough, M.J., 2000. Variability in the abundances of ®sh associated with seagrass habitats in relation to the diets of predatory ®shes. Mar. Biol. 136, 725±737. Hixon, M.A., 1991. Predation as a process structuring coral reef ®sh communities. In: Sale, P.F. (Ed.), The Ecology of Fishes on Coral Reefs. Academic Press, San Diego, CA, p. 754. Hoedt, F.E., Dimmlich, W.F., 1994. Diet of subadult Australian Salmon, Arripis truttacea, in Western Port, Victoria. Aust. J. Mar. Freshwat. Res. 45, 617±623. Hyslop, E.J., 1980. Stomach contents analysis: a review of methods and their application. J. Fish Biol. 17, 411±429. Jenkins, G.P., Black, K.P., 1994. Temporal variability in settlement of a coastal ®sh (Sillaginodes punctata) determined by low-frequency hydrodynamics. Limnol. Oceanogr. 39 (7), 1744±1754. Jenkins, G.P., Wheatley, M.J., 1998. The in¯uence of habitat structure on nearshore ®sh assemblages in a southern Australian embayment: comparison of shallow seagrass, reef algal and unvegetated sand habitats, with emphasis on their importance to recruitment. J. Exp. Mar. Biol. Ecol. 221, 147±172. Jenkins, G.P., Black, K.P., Wheatley, M.J., Hatton, D.N., 1997a. Temporal and spatial variability in recruitment of a temperate, seagrass-associated ®sh is largely determined by physical processes in the pre- and post-settlement phases. Mar. Ecol. Prog. Ser. 148, 23±35. Jenkins, G.P., May, H.M.A., Wheatley, M.J., Holloway, M.G., 1997b. Comparison of ®sh assemblages associated with seagrass and adjacent unvegetated habitats of Port Phillip Bay and Corner Inlet, Victoria, Australia, with emphasis on commercial species. Estuar. Coast. Shelf Sci. 44, 569±588. Jenkins, G.P., Welsford, D.C., Keough, M.J., Hamer, P.A., 1998. Diurnal and tidal vertical migration of pre-settlement King George whiting Sillaginodes punctata in relation to feeding and vertical distribution of prey in a temperate bay. Mar. Ecol. Prog. Ser. 170, 239±248. Jordan, F., Bartolini, M., Nelson, C., Patterson, P.E., Soulen, H.L., 1996. Risk of predation affects habitat selection by the pin®sh Lagodon rhomboides (Linnaeus). J. Exp. Mar. Biol. Ecol. 208, 45±56. Kemp, W.M., 1989. Estuarine Seagrasses. In: Day, J.W., Hall, C.A.S., Kemp, W.M.,Yanez-Arancibia, A. (Eds.), Estuarine Ecology. Wiley, New York, pp. 226±253. Kennelly, S.J., 1991. Caging experiments to examine the effects of ®shes on understorey species in a sublittoral kelp community. J. Exp. Mar. Biol. Ecol. 147, 207±230. Kingsford, M.J., 1992. Spatial and temporal variation in predation on reef ®shes by coral (Plectropomus leopardus, Serranidae). Coral Reefs 11, 193±198. Kingsford, M.J., 1998. Reef ®shes. In: Kingsford, M.J., Battershill, C. (Eds.), Studying Temperate Marine Environments: a Handbook for Ecologists. Canterbury University Press, Christchurch, pp. 132±165. Laprise, R., Blaber, S.J.M., 1992. Predation by Moses perch, Lutjanus russelli, and blue-spotted trevally, Caranx bucculentus, on juvenile brown tiger prawn, Penaeus esculentus: effects of habitat structure and time of day. J. Fish Biol. 40, 627±635. Levin, P.S., Hay, M.E., 1996. Responses of temperate reef ®shes to alterations in algal structure and species composition. Mar. Ecol. Prog. Ser. 134, 37±47. 174 J.S. Hindell et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 153 ±174

Levin, P., Petrik, R., Malone, J., 1997. Interactive effects of habitat selection, food supply and predation on recruitment of an estuarine ®sh. Oecologia 112, 55±63. Lewis, F.G., 1984. Distribution of macrobenthic crustaceans associated Thalassia, Halodule and unvegetated sand substrata. Mar. Ecol. Prog. Ser. 19, 101±113. Mattila, J., 1995. Does habitat complexity give refuge against ®sh predation? Some evidence from two ®eld experiments. In: Eleftheriou, A., Ansell, A., Smith, C.J. (Eds.), Biology and Ecology of Shallow Coastal Waters. Olsen and Olsen, Fredensborg, pp. 261±268. Minello, T.J., 1993. Chronographic tethering: a technique for measuring prey survival time and testing predation pressure in aquatic habitats. Mar. Ecol. Prog. Ser. 101, 99±104. Orth, R.J., 1992. A perspective on plant±animal interactions in seagrasses: physical and biological deter- minants in¯uencing plant and animal abundance. In: John, D.M., Hawkins, S.J., Price, J.H. (Eds.), Plant±Animal Interactions in the Marine Benthos. Association Special Volumes, Vol. 46. Clarendon Press, Oxford, pp. 147±164. Orth, R.J., Heck, K.L., von Montfrans, J., 1984. Faunal communities in seagrass beds: A review of the in¯uence of plant structure and prey characteristics on predator-prey relationships. Estuaries 7 (4a), 339±350. Pollard, D.A., 1984. A review of ecological studies on seagrass-®sh communities, with particular reference to recent studies in Australia. Aquat. Bot. 18, 3±42. Robertson, A.I., 1980. The structure and organisation of an eelgrass ®sh fauna. Oecologia 47, 76±82. Robertson, A.I., 1982. Population dynamics and feeding ecology of juvenile Australian salmon () in Western Port, Victoria. Aust. J. Mar. Freshwater Res. 33, 369±375. Rosenberg, M., Hodgkinson, R., Black, K., Colman, R., 1992. In: Hydrodynamics of Port Phillip Bay. Field collection and data reduction. Victorian Institute of Marine Sciences, Melbourne. Sale, P.F., 1998. Appropriate spatial scales for studies of reef-®sh ecology. Aust. J. Ecol. 23, 202±208. Sogard, S.M., Able, K.W., 1994. Diel variation in immigration of ®shes and decapod crustaceans to arti®cial seagrass habitat. Estuaries 17 (3), 622±630. Steele, M.A., 1998. The relative importance of predation and competition in two reef ®shes. Oecologia 115, 222±232. Stoner, A.W., 1982. The in¯uence of benthic macrophytes on the foraging behaviour of pin®sh, Lagodon rhomboides (Linnaeus). J. Exp. Mar. Biol. Ecol. 58, 271±284. Summerson, H.C., Peterson, C.H., 1984. Role of predation in organising benthic communities of a temperate- zone seagrass bed. Mar. Ecol. Prog. Ser. 15, 63±77. Virnstein, R.W., 1978. Predator caging experiments in soft sediments: caution advised. In: Wiley, M.L. (Ed.), Estuarine Interactions. Academic Press, New York, pp. 261±273. Virnstein, R.W., 1980. Measuring effects of predation on benthic communities in soft sediments. In: Kennedy, V.S. (Ed.), Estuarine Perspectives. Academic Press, New York, pp. 281±290. Wilkinson, L., Hill, M., Welna, J.P., Birkenbeuel, G.K., 1992. SYSTAT for Windows: Statistics, Version 5 Edition. SYSTAT, Inc., Evanston, IL.