Mortality of juvenile damselfish: implications for assessing processes that determine abundance
Russell J. Schmitt
INTRODUCTION
Replenishment of demographically open populations has been examined extensively in marine reef fishes since, like most benthic marine species, they typically have early developmental stages that disperse in the plankton (Doherty and Williams 1988, Mapstone and Fowler 1988, Olafsson et al. 1994, Booth and Brosnan 1995). Despite considerable attention, there currently is little agreement regarding the relative contributions of a planktonic supply of propagules and of subsequent demographic processes in determining densities of local populations (Doherty 1991, Jones 1991, Caley et al. 1996). Clearly, some potentially large fraction of the spatial and temporal variation observed emanates from variability in the number of planktonic stages that settle (Doherty 1981, 1991, Victor 1983, 1986, Mapstone and Fowler 1988, Doherty and Fowler 1994a, b). Less obvious is the degree to which postsettlement processes alter patterns originating at settlement (Jones 1991, Caley et al. 1996). To resolve this uncertainty requires knowledge of larval supply relative to potentially limiting reef resources, as well as the magnitude and form of postsettlement mortality.
One current model for marine reef fishes is that the supply of propagules typically is too low to saturate reef resources with the consequence that local population densities reflect a balance between the external supply of colonists and subsequent density-independent mortality (Doherty 1981, 1991). In such “recruitment limitation” (e.g., Doherty 1981) or “recruitment determination” (Forrester 1990) models, per capita loss rates after settlement are unaffected by the density of individuals present (Doherty and Williams 1988, Forrester 1995). If mortality is density independent, the effect of postsettlement loss on density of a cohort mostly will be quantitative (though perhaps large), and there will be relatively little alteration of the pattern of variation that originates at settlement (Doherty 1991, Jones 1991). Although variance will be reduced by density-independent loss (due to a reduction in mean abundance), such loss will not modify qualitatively a linear relationship between the size of a settlement cohort and the subsequent number that survives to an older life stage. By contrast, when the fraction of a cohort that reaches an older life stage declines with density, the relationship between the densities of successive life stages of a cohort will deviate from linear. The existence and strength of density dependence are crucial because temporal density dependence regulates a local population by bounding fluctuations.
There have been two general ways the recruitment limitation model has been tested using reef fishes. The first has been to determine from natural variation in densities of a species the relationship in abundance of different life stages (e.g., Victor 1983, 1986, Jones 1990, Milicich et al. 1992, Robertson 1992, Meekan et al. 1993, Doherty and Fowler 1994a, Schmitt and Holbrook 1996), and to infer from these patterns whether mortality likely scaled with density. Studies like these can indicate that postsettlement processes were not strong enough to eradicate the “supply” signal, but they do not necessarily eliminate the possibility that postsettlement losses involved density dependence. Indeed, the density of a cohort will remain positively correlated with its initial density except under the stringent conditions of complete or overcompensation of all density-independent losses (Caley et al. 1996, Nisbet et al. 1996; also see Warner and Hughes 1988). In fact, in several studies (Forrester 1990, Jones 1990, Schmitt and Holbrook 1996, Steele 1997) density-dependent losses in a cohort did not obscure a positive relationship between initial and final densities.
The second approach has been to test directly whether per capita mortality rates of reef fishes are dependent on density, which typically has involved experimental manipulations of fish densities. Although density dependence in growth rates of individual fish has been found commonly (e.g., Doherty 1982, 1983, Jones 1984, 1987a, b, 1988, Forrester 1990, Anderson 1993, Booth 1995), evidence of compensatory mortality in the juvenile stages of reef fishes has been found less frequently despite a number of attempts (Doherty 1982, 1983, Jones 1984, 1987a, Victor 1986, Behrents 1987, Pitcher 1988, Sale and Ferrell 1988, Robertson 1992, Anderson 1993, Booth and Beretta 1994, Levin 1994, Connell 1997; but see Jones 1987a, 1988, Robertson 1988, Forrester 1990, 1995, Tupper and Hunte 1994, Carr and Hixon 1995, Tupper and Boutilier 1995, Schmitt and Holbrook 1996, Hixon and Carr 1997, Steele 1997). The failure to detect routinely density dependence in mortality of juvenile reef fishes has been viewed as support for recruitment limitation (Doherty 1991, Robertson 1992, Doherty and Fowler 1994a, b).
Our ability to draw meaningful conclusions regarding the nature of postsettlement mortality in juvenile reef fishes, and thus its potential role in population regulation, is hampered by a number of issues. Among these is the fact that the assessment of density dependence is a notoriously difficult task for a number of logistical and statistical reasons (Hassell 1986, Jones 1991, Murdoch 1994). There is, however, a shortcoming that is relatively specific to the study of reef fishes: a frequent failure to examine patterns of mortality of newly settled individuals (Caley et al. 1996, Steele 1997). To date most experimental and observational studies of postsettlement mortality have focused on juveniles that had settled weeks to months earlier (Williams et al. 1994; see review by Caley et al. 1996). Density-dependent processes that are important for local regulation may occur at or just following settlement, yet this possibility cannot be evaluated fully until very early postsettlement mortality is explored more consistently.
We examined the effect of several factors on per capita mortality rates of the youngest age classes of three species of damselfish and the consequences of these factors in modifying local patterns of abundance. We focused on two age classes: the newly settled cohort (followed from the day of settlement) and the next older cohort that had been on the reef [approximately] 14-28 d at the beginning of observation. We assessed the influence of body size (related to species and time since settlement), density, and interactions within and between different age classes of juveniles on (1) the magnitude of postsettlement mortality, (2) the fraction of each cohort that died in 2 wk, (3) the contributions of each mortality component to reducing spatial variance in juvenile abundance, and (4) the contributions of primary recruitment limitation (sensu Victor 1986), density-independent and density-dependent mortality in determining the average abundance of settlers that survived to 2 wk after arrival.
The study organisms
The three-spot dascyllus (Dascyllus trimaculatus), yellow-tail dascyllus (D. flavicaudus), and humbug dascyllus (D. aruanus) are diurnal planktivores. Adults shelter opportunistically in reef crevices (three-spot dascyllus) or in live coral heads (yellow-tail and humbug dascyllus) (Allen 1991). After a 22-24 d planktonic period (Wellington and Victor 1989), young settle to specific reef habitats: sea anemones for three-spot dascyllus and branched corals (primarily species of Pocillopora and Acropora) for yellow-tail and humbug dascyllus (Sale et al. 1984, Forrester 1990, Schmitt and Holbrook 1996, Holbrook and Schmitt 1997).
Several features of these Dascyllus make them ideal for examining early postsettlement mortality. The occurrence of young in discrete units of microhabitat facilitates accurate censuses. Early postsettlement movement, including unseen replacement, can be problematic to the estimation of mortality (Frederick 1997), but it is not an issue for this system. Young rarely move from the particular coral head or anemone on which they settle for at least the first few weeks (Forrester 1990, Schmitt and Holbrook 1996). For example, we estimated the movement rate of individuals [less than] 20 mm standard length (SL) among a series of empty microhabitats spaced from 5 to 25 m apart; the estimated known movement rates ranged from a low of 0 in 339 observation days for humbug dascyllus to a high of 8 in 1007 observation days for three-spot dascyllus (also see Schmitt and Holbrook 1996). Most movements were to the closest microhabitat. Also, new settlers have distinctive morphological characteristics that identify them as having settled during the previous night (Holbrook and Schmitt 1997), making it easy to distinguish new colonists from individuals that have been on the reef for longer than 1 d. Finally, settlement of these species primarily occurs in discrete pulses lasting 3-5 d near the quarter moon phases of the lunar cycle, with very little or no settlement occurring during intervening days (Holbrook and Schmitt 1997, Schmitt and Holbrook, in press). Because of rapid body growth after settlement, successive cohorts of the youngest age classes are easily distinguishable.
METHODS
Field work was conducted on Moorea, French Polynesia (17 [degrees] 30 [minutes] S, 149 [degrees] 50 [minutes] W) during June-September 1996, a period of heavy settlement of the Dascyllus species. Moorea is encircled by barrier reefs situated 0.8-1.3 km from shore, forming a system of lagoons averaging [approximately] 5 m in depth (for more details, see Galzin and Pointier 1985). Patch reefs are interspersed with sand within each lagoon.
Patterns of mortality
We estimated mortality rates of young Dascyllus by determining the fate of individuals on coral heads or anemones for 14-d periods beginning at the peak of a settlement pulse. To minimize confounding influences, we followed individuals on a standard amount of suitable habitat (coral heads and anemones) transplanted without fish to each of our study sites. In all but one case, we used fishes that naturally settled to transplanted habitat; the exception involved manipulation of new settlers and the next older age class of three-spot dascyllus to verify patterns observed for natural levels of settlement. Unless otherwise noted below, we made daily counts of individuals. This procedure only missed losses of settlers that occurred during the interval between arrival (at night) and first observation. Settlement events observed using in situ infrared video technology all occurred between 0000 and 0500, and of these [approximately] 25 settlers, all remained on the microhabitat to which they had settled and all survived until daylight (Holbrook and Schmitt 1997, Holbrook and Schmitt, in press). As daily counts were completed by 1200, the longest probable interval where the loss of a new colonist could have occurred and would have been missed was 12 h.
Cohort-specific loss rates were estimated for the youngest two age classes: individuals that had just settled (hereafter the “newly settled” cohort) and young from the previous settlement pulse (“the next older cohort”). Spatial patterns of mortality of these two age classes were explored at two spatial scales. The first included eleven sites distributed around the 60-km perimeter of Moorea. The second spatial scale considered mortality among individual corals or sea anemones within a site. Because movement rates of these age classes are so low, we equate losses with death although movement could account for a small but unknown fraction of the overall loss rate.
Among-site spatial scale. – Patch reefs were constructed by placing cinder blocks in isolated sandy bottom areas of lagoons at depths of 2.5-3 m and affixing corals or anemones. Each site Was situated about mid-distance between the barrier reef and shore, and each was in the same relative proximity to a pass in the barrier reef (the outlet of water that enters over the reef crest) and to major formations of naturally occurring suitable habitat. The nearest natural live coral formation was [approximately] 25 m away. At all sites, 10 heads of live Pocillopora, each [approximately] 30 cm in diameter and 20 cm tall, and 10 anemones (Heteractis magnifica), each [approximately] 30 cm in diameter, were transplanted to separate cinder blocks placed 5-7 m apart. Coral heads were glued singly to the cinder blocks with Z-Spar Splash Zone Compound (Kop-Coat, Incorporated, Los Angeles, California), and anemones were allowed to attach naturally.
Transplanted habitats were allowed to accumulate fish for a period of just under 1 mo (i.e., one settlement pulse). Just prior to initiation of daily observations for the subsequent settlement event, resident fish were removed from half of the corals and anemones at each site. This allowed us to estimate mortality of new settlers in the presence and absence of the next older age class. We estimated the death rate of the newly settled cohort and the next older cohort during the settlement pulse that occurred in the second half of August 1996. For the settlement pulse examined, a total of 1046 Dascyllus individuals settled and comprised the newly settled cohorts. The next older cohorts consisted of 273 fish that remained following removal from half of the microhabitats.
The dependent variate used in analyses of density dependence was the fraction of a cohort at a site that survived 14 d. Although densities declined over the 2 wk, we used the maximum density of conspecifics at each site as the independent variable; this occurred on the final day of the settlement pulse, which typically was [approximately] 3 d after the start of daily observations. The fractional loss of a cohort was regressed against maximum density at the site; as these patterns appeared nonlinear, we used two nonlinear models as well as linear regression. The nonlinear models were the In(maximum density) and one based on the logistic regression: fractional loss = [exp(a + b x maximum density)/(1 + exp(a + b x maximum density))].
Within-site spatial scale. – To estimate patterns of mortality at the smaller spatial scale, we transplanted 40 live Pocillopora coral heads to a sandy bottom in the Vaipahu Lagoon on the north shore of Mootea. As before, corals were affixed with epoxy to cinder blocks that were 5-7 m apart. Forty anemones were transplanted to nearby hard substrata with the same spacing as for corals. Corals were allowed to accumulate fish (yellow-tail and humbug dascyllus) for [approximately] 12 mo, and anemones accumulated three-spot dascyllus for [approximately] 6 mo. When the settlement pulse examined occurred (late July 1996), corals were occupied by several juvenile age/size classes of both yellow-tail and/or humbug dascyllus, although no adults were present. Consequently, for humbug and yellow-tail dascyllus, partial regressions were used to examine the relationship between the fraction of the new settler or next older cohort that died in 2 wk and the maximum density of different age and species groups present on a coral head. In the partial regressions, the dependent variable was the fraction of a yellow-tail or humbug dascyllus cohort (either the new settler or the next older age class) that died on each coral head. The four independent variables in each partial regression were the density on each coral head of humbug settlers, yellow-tail settlers, humbug juveniles [greater than or equal to] 2 wk old, and yellow-tail juveniles [greater than or equal to] 2 wk old. For each independent variable that was statistically significant, a partial regression leverage plot is presented, which shows the relationship between the dependent variable and the regressor after they have been made orthogonal to the other regressors in the model. Regressions were calculated using raw and In-transformed densities.
Experimental tests of within-site effects of density on mortality. – While our estimates of mortality from the surveys utilized natural variation in densities of settlers, the trials were not experiments because densities and size structures were not manipulated and assigned at random among observational units. Further, both the number of individuals and range in densities of the next older cohorts were much lower than for the settler cohorts. Hence, it is possible that a difference in the mortality pattern between the age classes reflected differences in statistical power. Consequently we performed field experiments to evaluate the major patterns of mortality found in our observations of natural settlement. We focused on three-spot dascyllus because it was possible to obtain the great number of new settlers needed. The effects of density and size-structured interactions at the within-site scale were examined in these experiments using 40 anemones that were transplanted to a reef and spaced [greater than] 5 m apart. One experiment examined mortality of new settlers and of the next older cohort when they co-occurred on anemones (both with similar absolute and ranges in densities). A second experiment examined mortality of the new settler cohort due solely to within-cohort interactions (i.e., older juveniles absent). New settlers were collected during a settlement peak and placed (with the appropriate number of older conspecifics) on experimental anemones (from which residents were removed) within 2 d of settlement. Fish were counted daily for 2 wk.
The experimental design we used differed from that typically done to explore the effects of density on mortality. A standard approach is to have several replicates of several density treatments. A problem with this design is that the fate of an individual is binomial (it survived or died) such that one death contributes more to the fractional loss at lower than at higher abundance; this causes variance in the fractional death rate to scale inversely with abundance, making detection of density dependence difficult statistically. One solution is to have many more replicates at lower than higher densities (e.g., Forrester 1995). We used an alternative approach of having the same number of individuals (new settlers or the next older age class) represented in each density treatment. Our dependent variable was the fraction of all focal individuals in a density treatment that died (i.e., there was no replication within a density treatment). Hence, estimates of fractional loss were based on the same number of individuals in each density treatment (25 for new settlers, 20 for the next older age class), while the number of anemones assigned to different treatments varied from two (highest density) to seven (lowest density treatment). It should be pointed out that differences in the number of groups among the density treatments may have resulted in variation in the quality of the mortality estimates. This issue is not unique to our design as it also pertains to approaches like that used by Forrester (1995).
Nine treatment densities were used to explore between-age class interactions, and eight were used when new settlers were alone. Treatment densities, which ranged from 0.005 to 0.1 individuals/[cm.sup.2] of anemone, were based on natural densities observed on nearby unmanipulated anemones. The fraction of focal individuals that died in 2 wk was regressed against (raw and In-transformed) treatment density.
Effects of density-dependent mortality
Relationships between the numbers of settlers and subsequent survivors. – For both spatial scales examined, the number of new settlers still alive after 2 wk was regressed against the initial number of settlers. Data for each species were fit to both a linear model and a curvilinear model with the shape of a Type II functional response. The equation of the nonlinear model was
[N.sub.14] = a[N.sub.0]/(1 + (a/b)[N.sub.0]) (1)
where [N.sub.14] is the number of settlers still alive after 14 d at a site or microhabitat and No is the initial number that had settled.
Contributions of mortality components to reducing spatial variance in new settlers. – The excellent fits of the nonlinear regression model (Eq. 1) to our settler – survivor data for the among-site scale provided the basis of a framework to estimate the contribution of various mortality components to the observed reduction in spatial variance in abundance of new settler cohorts. Parameter a in Eq. 1 is the initial slope of the relationship (i.e., as No [right arrow] 0), which represents an estimate of the density-independent rate of per capita mortality. Error (95% confidence intervals) associated with these parameter estimates was used to place upper and lower bounds on the contributions of the density-independent and -dependent components of mortality.
For each species, the overall proportional reduction in spatial variance of new settler cohorts in 2 wk was estimated as ([Var.sub.initial] – [Var.sub.final])/[Var.sub.initial]. The contribution of each mortality component to this reduction was done by subtraction. Values for parameter a (including the upper and lower 95% CI estimates for each species) were used to calculate the variance in abundance 2 wk after settlement that was due to density-independent loss lowering mean abundance. Density-dependent loss affects variance through two pathways: by lowering mean abundance further, and by an unequal distribution of that added mortality among densities (i.e., disproportionately fewer and greater losses at lower and higher densities, respectively). The contribution of the added mortality component from density dependence was estimated using the overall mean loss rate of a species to calculate the variance due to the effects of both density-independent and -dependent loss on lowering mean abundance, then subtracting this from the estimated effect on variance due to the density-independent component. The contribution of density dependence due to an unequal distribution of the added mortality was estimated using the full nonlinear model to calculate the variance due to all components and subtracting this from the effect of lowering mean abundance. The differences between the actual reduction in variance and that explained by the full nonlinear model (i.e., “residual”) were quite low (all [less than] 1.6%), indicating that the functional form of the equation used was an excellent approximation of the actual relationships.
Relative contributions of pre- and postsettlement processes in setting average abundance of 2-wk old recruits. – We also used the excellent fits to Eq. 1 of our among-site data for the numbers of new settlers that arrived and survived to estimate the relative importance of larval supply and subsequent density-independent and density-dependent mortality in determining the average abundance of 2-wk-old recruits. Parameter b in Eq. 1 is the estimated asymptotic number of new settlers that can be supported by the available microhabitat at each site. This provides a means to estimate the relative contribution of primary recruitment limitation (sensu Victor 1986) to setting average recruit abundance. Primary recruitment limitation occurs when the input of colonists is too low to saturate resources in the absence of postsettlement mortality. Hence, for the average amount of settlement, the difference between the estimated asymptotic number of 2-wk-old recruits that the microhabitats can support and the mean number of initial settlers (i.e., no mortality to 2 wk) is a measure of the contribution of primary recruitment limitation to setting the average abundance of 2-wk-old recruits. The contribution of subsequent density-independent loss can be estimated using parameter a in Eq. 1 to determine the average number of survivors to 2 wk, and subtracting this from the mean number of initial settlers. Similarly, the density-dependent contribution to the mean abundance of 2-wk-old recruits can be derived using Eq. 1 by subtracting the mean abundance of recruits present from the number that would have survived if only density-independent mortality had operated. Expressing the contribution of each component as a proportion of the total “abundance cost” yields the relative importance of each factor to setting the average abundance of 2-wk-old recruits.
TABLE 1. (A) The proportion of dascyllus young that died at 11
sites around Moorea in the 2-wk period immediately after a
settlement pulse and (B) ANOVA results.
A) Proportion of dascyllus young that died
Species New settler cohort Next older cohort
Three-spot 0.314 (0.046) 0.115 (0.034)
Yellow-tail 0.371 (0.036) 0.152 (0.060)
Humbug 0.547 (0.061) 0.193 (0.061)
B) Differences in fractional death rates
Source df MS F P
Among species 2 0.131 5.12 [less than] 0.01
Between age classes 1 1.003 39.21 [less than] 0.001
Species X age class 2 0.036 1.41 [greater than] 0.25
Error 55 0.026
Notes: (A) Each site contained the same amount of juvenile habitat.
For each damselfish, the mean fractional mortality rates (1 SE in
parentheses) over the 11 reefs are given for the two age classes
present: newly settled individuals and those from the previous
pulse [approximately] 1 mo earlier. (B) Also given are the ANOVA
results for differences in fractional death rates among species and
age classes.
In addition to considering average settlement levels, we used the framework above to estimate the fraction of all 11 sites where settlement was sufficiently high to have saturated sites (1) in the absence of any postsettlement mortality (i.e., no primary recruitment limitation), (2) if only density-independent mortality had occurred, and (3) given the estimated strengths of both density-independent and -dependent mortality.
RESULTS
Patterns of mortality
Species-, age-, and size-specific mortality rates of juvenile Dascyllus. – At the among-site scale, the average percentage loss of young Dascyllus in 14 d differed both between the two cohorts and among the three species (Table 1). However, the species all showed the same qualitative difference in per capita mortality rates between age groups (i.e., nonsignificant ANOVA age x species interaction term; Table 1), with settler cohorts having substantially greater losses than the next older cohorts. Averaged across species, 41% of new settlers and 15% of the next older cohort died during the 2-wk period.
Body size differed among the species and age classes, and this aspect appeared to contribute to the observed differences in per capita death rates. Average standard length of fish within an age class was an excellent predictor of the mortality as fractional losses declined exponentially with increasing body length [ILLUSTRATION FOR FIGURE 1 OMITTED].
Body size was not the only determinant of the per individual death rate, at least for new settlers. Fractional losses of new settlers were nearly twice as great in the presence than absence of older conspecifics (Table 2). These differences did not result from disparities in densities of new settlers between microhabitats with and without the next older age class (all P [greater than] 0.2). Rather, both overall density and age (size) structure varied between habitat units that had or lacked the older cohort, indicating that per capita loss rates of newly [TABULAR DATA FOR TABLE 2 OMITTED] [TABULAR DATA FOR TABLE 3 OMITTED] settled Dascyllus of all three species were influenced by one or both of these factors.
Density effects on per capita mortality at the among-site spatial scale. – At the among-site scale, the effect of peak density on per capita mortality of young (both age classes combined) was positive but quite weak (Table 3). The relationship between the fraction of juveniles that died in 2 wk and density was statistically significant for just one of the species.
A considerably different view emerged when the two age classes were considered separately. Fractional mortality of new settlers was strongly and positively related to peak density on a reef ([ILLUSTRATION FOR FIGURE 2 OMITTED]; Table 3). By contrast, [TABULAR DATA FOR TABLE 4 OMITTED] the fraction of the next older cohort that died did not vary significantly with density ([ILLUSTRATION FOR FIGURE 2 OMITTED]; Table 3). These results indicated that per capita mortality rates were strongly density dependent for new settlers, but were density independent for the next older age class. It should be noted that, relative to the new settler cohorts, the lower overall abundances and smaller ranges in densities of the next older cohort raise the possibility that failure to detect density dependence for this age class could have been due to low statistical power.
Per capita mortality rates of new settlers did not appear to scale linearly with increasing density, particularly for three-spot and yellow-tail dascyllus. For them, considerably more variance in fractional mortality of settlers was explained by a In(density) model than a linear model (Table 4). The nature of the relationship with density was less clear for humbug dascyllus; a ln(density) model explained somewhat less variance than the linear model. However, a second two-parameter nonlinear model based on logistic regression provided the same, high amount of explained variance in the fractional death rate of humbug settlers as did the linear model (Table 4). Thus for all three species, a model where the rate of increase in per capita mortality declined with increasing density explained [greater than] 75% of the variance in the fractional mortality of new settlers. With respect to the next older age class, the only substantial improvement of a nonlinear over a linear model was for humbug dascyllus (Table 4), which still accounted for only a third of the variance.
The apparent diminution in the rate of increase in per capita mortality with increasing density within a settler cohort was well approximated by a semilog model ([ILLUSTRATION FOR FIGURE 2 OMITTED]; Table 4). The slopes of the semilog models are estimates of the strength of density dependence for each of the six age and species groups. The relationship between the intensity of density-dependent mortality and the average standard length of individuals in a cohort was inverse and apparently linear [ILLUSTRATION FOR FIGURE 3 OMITTED]. Thus, both the magnitude of per capita mortality [ILLUSTRATION FOR FIGURE 1 OMITTED] and intensity of density dependence [ILLUSTRATION FOR FIGURE 3 OMITTED] scaled with body size of these juvenile damselfish. These relationships also suggest that the intensity of density dependence is relaxed gradually as individuals grow in size over the first week or so on the reef.
Effects of density, size structure, and species composition on per capita mortality at the within-site scale. – Because young Dascyllus settle to and occupy single coral heads or anemones, their populations not only are subdivided spatially at the scale of discrete reefs, but also at the level of suitable microhabitats within a location. Here we present data from a site that was not used in the trials reported above to address the effects of within- and between-cohort densities on per capita mortality rates.
1. Three-spot dascyllus. – At the within-site scale, the overall fraction of new three-spot settlers that died in 2 wk (46.7%) was substantially greater than that of the next older cohort (15.9%). The per capita mortality rate of new settlers varied positively with peak density (size classes combined) on an anemone ([F.sub.1,19] = 32.34; P [less than] 0.001), whereas that for the next older cohort was unrelated to variation in density among anemones ([F.sub.1,28] = 2.75; P [greater than] 0.75). The amount of variance explained in mortality of new settlers was improved slightly when density was In-transformed ([r.sup.2] = 0.67) compared with a linear model ([r.sup.2] = 0.63). These patterns within a site were qualitatively identical to those found at the among-site scale.
With respect to size (age)-specific density effects, per capita mortality of new three-spot settlers increased significantly with the In(density) of older juveniles on an anemone (P [less than] 0.001), but not with ln(density) of settlers alone (P [greater than] 0.9). Indeed, the density of young [greater than or equal to] 2 wk old on an anemone explained more variance in the per capita death rate of new settlers ([r.sup.2] = 0.72) than did total juvenile density (which included settlers; [r.sup.2] = 0.67). The amount of additional variance explained by settler density essentially was zero (partial [r.sup.2] = 0.0001) despite representing [greater than] 25% of all fish present at the start of observations. Per capita mortality of the next older age class was unaffected by the density of settlers (P [greater than] 0.98) or juveniles [greater than or equal to] 2 wk old (P [greater than] 0.7).
In field experiments, we manipulated both densities and size structure of new settlers and older juveniles among anemones to examine patterns of mortality when both age classes had similar average abundance and range in densities. For new settlers, the average fraction that died in the field experiments was about twice as great in the presence of older recruits (43%) than when alone (23%). Increases in the density of the next older age class resulted in increased per capita mortality of new settlers ([ILLUSTRATION FOR FIGURE 4 OMITTED]; [F.sub.1,7] = 44.9; P [less than] 0.001); again ln(density) was a better predictor ([r.sup.2] = 0.87) than a linear model ([r.sup.2] = 0.80). These experimental results are qualitatively identical to those obtained from our surveys. However, unlike the survey results, the density of settlers in the absence of older juveniles had a positive and statistically significant effect on their per capita mortality rate ([ILLUSTRATION FOR FIGURE 4 OMITTED]; [F.sub.1,6] = 55.8; P [less than] 0.001). The within-cohort effect was better described by a In(density) ([r.sup.2] = 0.90) than linear model ([r.sup.2] = 0.83). The intensity of density dependence (i.e., slope of fractional death of settlers vs. In(density)) was, however, greater in the presence (slope = 0.22) than absence (0.10) of older individuals (Table 5; [ILLUSTRATION FOR FIGURE 4 OMITTED]).
For the next older cohort, the fraction that died in the experiments was independent of density in both the presence (slope = 0.042; [r.sup.2] = 0.097; [F.sub.1,7] = 0.75; P [greater than] 0.4) and absence (slope = -0.007; [r.sup.2] = 0.005; [F.sub.1,6] = 0.03; P [greater than] 0.85) [TABULAR DATA FOR TABLE 5 OMITTED] of new settlers. These results suggest that the failure to detect density dependence in the loss rates of the next older cohorts at both the among- and within-site scales was not due to lower average abundance or a smaller range in densities.
2. Yellow-tail dascyllus. – Coral heads surveyed for the within-site scale contained different juvenile age (size) classes of both yellow-tail and humbug dascyllus, so potential species- and size-specific effects on per capita mortality were explored. We report only results of regressions where density was In-transformed since they always explained greater variance in per capita loss rates than did raw densities.
In the 2-wk period, 47% of the new settlers but just 7% of the next older cohorts of yellow-tail dascyllus died. A substantial amount of variance in per capita mortality of new yellow-tail settlers ([r.sup.2] = 0.61) was explained when species-specific variation in density of age classes among corals was considered explicitly in partial regression analyses. Fractional mortality of new yellow-tail settlers varied positively with local densities of older conspecifics and new humbug settlers [ILLUSTRATION FOR FIGURE 5a, b OMITTED], [TABULAR DATA FOR TABLE 6 OMITTED] but was unrelated to densities of older humbugs or its own cohort (Table 6). Variation in density of older conspecifics accounted for the largest amount of explained variance (partial [r.sup.2] = 0.43) whereas that for newly settled humbugs accounted for a much lower portion (partial [r.sup.2] = 0.09). The intensity of density dependence on yellow-tail settlers was much greater from older conspecifics ([ILLUSTRATION FOR FIGURE 5a OMITTED]; slope = 0.29) than from new humbug settlers ([ILLUSTRATION FOR FIGURE 5b OMITTED]; slope = 0.16).
For the corals and period examined, the two most abundant age groups were yellow-tail dascyllus [greater than or equal to] 1 mo old and new humbug settlers, each of which were more than twice as abundant as the other two groups (Table 7). Thus, per capita mortality of new yellow-tail settlers scaled positively with the two age groups of fish that had the greatest density.
TABLE 7. Abundances of new settler and older juvenile cohorts of
yellow-tail and humbug dascyllus on corals used to examine
within-reef effects of species- and cohort-specific densities on
the fraction of young that died in 2 wk (see Table 6, [ILLUSTRATION
FOR FIGURE 5 OMITTED]).
Propor-
Species Age class Abundance Range tion
Yellow-tail Older juveniles 7.35 (0.98) 0-23 0.38
New settlers 2.77 (0.36) 0-10 0.14
Humbug Older juveniles 3.30 (0.50) 0-13 0.17
New settlers 5.93 (0.92) 0-23 0.31
Notes: Data are the mean number (1 SE in parentheses) and range of
individuals in each age class present on a coral head at the start
of the observation period. Also given is the proportion of the total
juveniles represented by each age class.
By contrast with new settlers, the fraction of the next older age class of yellow-tail dascyllus that died was not related to variation among corals in the density of any species, age class, or combination of species and age class (Table 6).
3. Humbug dascyllus. – Results for humbugs at the within-site scale were qualitatively similar to those of yellow-tail dascyllus, including the consistently better fit of the data to a ln(density) than linear regression model. A substantially higher fraction of new banded humbug settlers died (53%) than did older juveniles (7%).
As with yellow-tail dascyllus, a substantial amount of explained variance ([r.sup.2] = 0.46) in the per capita mortality of humbug settlers was obtained when species-specific age class densities were used as dependent variables in the regression model (Table 6). The fractional mortality of new settlers scaled positively with the density of the two most abundant age groups (Tables 6 and 7): older yellow-tail juveniles [ILLUSTRATION FOR FIGURE 5c OMITTED] and their own cohort [ILLUSTRATION FOR FIGURE 5d OMITTED]. There was no statistically significant effect of variation in density of new yellow-tail settlers or of older humbugs (Table 6). The within-cohort effect was stronger and accounted for more explained variance (slope = 0.25; partial [r.sup.2] = 0.295) than did that of older yellow-tail humbugs (slope = 0.16; partial [r.sup.2] = 0.10). Again, per capita mortality of new humbug settlers was positively related to the two most abundant groups of juvenile Dascyllus present.
TABLE 8. Relationships between the number of new settlers that
colonized a reef or microhabitat and the number of those individuals
that survived 2 wk for each spatial scale examined.
Explained variance ([r.sup.2])
Linear Nonlinear
Analysis model model
Among sites, natural settlement
Yellow-tail 0.95 0.97
Humbug 0.60 0.87
Three-spot 0.94 0.97
Among corals/anemones within a site, natural settlement
Yellow-tail 0.61 0.59
Humbug 0.25 0.42
Three-spot 0.69 0.72
Among anemone experiments, three-spot dascyllus
Single cohort experiment 0.92 0.92
Mixed cohort experiment 0.53 0.63
Note: The table presents the amounts of variance explained
(r.sup.2) by linear and nonlinear (Type II functional response)
regression models. All regressions were significant at
[Alpha] = 0.005.
The fractional death of the next older age classes of humbug dascyllus was not related statistically to variation in the density of any species, age class, or combination of species and age class (Table 6).
Effects of density-dependent mortality
Relationships between the numbers of settlers and subsequent survivors. – Linear regression of the numbers of settlers and subsequent survivors to 2 wk was highly statistically significant (P [less than] 0.005) for all species at both spatial scales. Moreover, a linear model explained nearly as much variation as a general nonlinear function (with a Type II functional response shape) in five of eight comparisons (Table 8). In only two cases (both involving humbugs) did the nonlinear model provide a substantial improvement in the amount of variance explained. However, the nonlinear form used provided an excellent fit ([r.sup.2] [approximately] 0.9) to the data for four of the eight comparisons, and a moderately good fit ([r.sup.2] [approximately] 0.6- 0.7) to three others.
Much more variance was explained by both regression models at the larger (among site) spatial scale than at the within-site scale (Table 8) despite [approximately] 4 times more data points for the within-site regressions. Three aspects likely contributed to this disparity. First, the amount of variance explained by a regression is sensitive to the range of values, and for the settlement pulses and sites examined, the range spanned two plus orders of magnitude among sites but just one order of magnitude within a site. Second, the contribution of a random death of an individual to unexplained variance will vary inversely with sample size, and the number of individuals that comprised each datum was about an order of magnitude lower at the within- than the amongsite scale. Third, relative to variation among sites, the units of observation (microhabitats) examined at the within-site scale varied tremendously from one another in the number of older cohorts, the densities of those cohorts, and (for corals) the species composition of older fishes present. These factors, which influenced the mortality of new colonists (Table 6), did not covary systematically with the density of new settlers among microhabitats, and the attendant variation in death rates they caused was not accounted for in simple regressions of the initial and subsequent numbers of individuals in a cohort.
Contributions of mortality components to reducing spatial variance in new settlers. – Averaged across the three species, mortality over 14 d reduced the amount of spatial variance in densities of new settler cohorts by 86.6% (1 SE: [+ or -] 4.2%) among the sites (Table 9) and 81.3% ([+ or -] 6.1%) among microhabitats within a site. A portion of these reductions was due to density-dependent mortality, and the patterns were quite consistent among the three species (Table 9). Overall, density dependence was responsible for 41-67% of the reduction in variance. However, given the error around the estimated density-independent mortality rate, the reduction due to density dependence could have been as low as [approximately] 20% or as high as 64-95% (Table 9). The component responsible for most of the density-dependent reduction in spatial variance was the effect on lowering mean abundance (beyond that attributable to density independence) (Table 9), which had a 2-6 times greater effect on reducing variance than did the distribution of that added mortality among densities (Table 9).
Relative contributions of pre- and postsettlement processes in setting average abundance of 2-wk-old recruits. – At the among-site scale, primary recruitment limitation had the largest influence in determining the average abundance of Dascyllus young that were present 2 wk after they colonized (Table 10). At the average settlement density of each species, the relative contribution of this component to setting the mean abundance of 2-wk-old recruits in the initially undersaturated microhabitats was 66-79%. Averaged across the species, density-independent and density-dependent mortality contributed [approximately] 17 and [approximately] 13 % to the mean abundance of 2-wk-old recruits, respectively (Table 10).
Depending on the species of damselfish, settlement at 1 or 2 of the 11 sites was sufficiently great that the microhabitats would have become saturated had there been no postsettlement mortality (i.e., no primary recruitment limitation) (Table 10). For two of the three species, settlement was even high enough at one or two sites that resources would have stopped further population growth had there only been density-independent [TABULAR DATA FOR TABLE 9 OMITTED] mortality (Table 10). However, at no site was settlement sufficiently great to actually saturate the available microhabitats given the strengths of both the density-independent and -dependent loss rates. The average densities of 2-wk-old recruits that resulted from the single settlement pulse were between 16 and 25% of the estimated asymptotic number that could have been supported.
DISCUSSION
A central unresolved issue for species with demographically open populations is the extent to which patterns of local abundance and dynamics are shaped by the external input of young relative to processes that operate after colonization. The resolution of this problem is hampered by a lack of information about the settlement process and the events immediately following. Recruitment, defined in this context as the first observation of young on a reef (Keough and Downes 1982), often is measured for reef fishes weeks to months after individuals have settled from the plankton. It is widely accepted that mortality occurs between settlement and when recruitment typically is measured, yet recruitment repeatedly has been used as a surrogate for settlement to examine the influence of larval supply on patterns of abundance or dynamics. There is little information for reef fishes as to whether very early postsettlement mortality varies with density (see Steele 1997), yet this knowledge is crucial to understanding dynamics of these species.
With respect to the Dascyllus species, the form of mortality differed substantially between the two youngest age classes on the reef. For all three species at both spatial scales, per capita mortality rates of new settlers were density dependent. By contrast, we did not detect any effect of density on the per capita mortality of fish that already had been on the reef for just a few weeks. We suspect that the intensity of density dependence declines rapidly as these fish grow in body size over the first week after settling. Although too few examples exist to draw general conclusions, that density dependence may operate on juvenile life stages of reef fishes primarily during a brief period immediately after settlement has a number of important implications. One is that studies that focus on juveniles that have been present for more than a few days may not capture important regulatory processes. A sizable fraction of experimental studies that have sought evidence of compensatory mortality in reef fishes have manipulated juveniles that were weeks to months old, rendering them of uncertain value for assessing the prevalence of early postsettlement density dependence. Evidence that mortality of juvenile reef fish increases with density has been found when observations were initiated at settlement (e.g., Jones 1987b, 1988, Forrester 1990, 1995, Cart and Hixon 1995, Tupper and Boutilier 1995, Hixon and Cart 1997, Steele 1997, although not in all cases (e.g., Victor 1986, Robertson 1992). Accurate inferences regarding the contribution of larval supply require knowledge of mortality from the moment of arrival.
For Dascyllus at Mootea, no strong threshold density appeared to exist below which density-dependent mortality did not occur, which is counter to the predictions of the recruitment limitation or recruitment determination model (e.g., Forrester 1990, Doherty 1991). Although per capita mortality of new settlers rose with density, it did not rise at a constant rate (Table 4). The form of these nonlinearities indicated a relaxation, and not an intensification, of the intensity of density-dependent mortality with density. That is, the effect of adding an individual on elevating the per capita mortality rate of new Dascyllus settlers appeared to be greater at lower than higher densities. This relaxation does not appear to have resulted from the fact that fractional mortality has a fixed upper bound. At present we cannot provide a biological explanation for this pattern, although we suspect that it may involve the nature of competitive interactions between individuals for shelter space from predators. The findings are, however, consistent with a local resource gradually growing limited well before it became totally saturated and stopped further population growth.
TABLE 10. The relative importance of primary recruitment
limitation, density-independent (D-I) mortality, and
density-dependent (D-D) mortality to setting the average abundance
of new settlers that survived to 2 wk at the among-reef scale. Also
given are the proportions of the 11 sites where primary recruitment
limitation did not occur (i.e., microhabitats would have been
saturated had no postsettlement mortality occurred), where
settlement was sufficiently great such that microhabitats would
have been saturated had there been only density-independent losses,
and where net recruitment (i.e., input – all losses to 2 wk) was
sufficiently great to saturate microhabitats; a site can occur in
more than one category.
Dascyllus species,
proportional
contribution(*)
Yel-
low- Hum- Three-
Source of loss tail bug spot
Primary recruitment limitation 0.79 0.66 0.69
D-I loss 0.12 0.20 0.18
D-D loss 0.09 0.14 0.13
Dascyllus species,
proportion of sites
in category
Yel-
low- Hum- Three-
Category tail bug spot
No primary recruitment limitation 0.09 0.18 0.18
Saturated if only D-I losses 0 0.09 0.18
Saturated with D-I and D-D losses 0 0 0
* Contribution to the average abundance of survivors to 2 wk.
With respect to the Dascyllus species, two unresolved issues involve the agent(s) of mortality and the manner by which variation in density influenced per capita mortality rates of new settlers. Although we have not yet fully addressed these topics, we have some evidence that strongly implicates predation. The coral heads and anemones with which Dascyllus young associate serve as shelters from predators (e.g., Coates 1980, Forrester 1990, Holbrook and Schmitt, in press); individuals rapidly retreat into these microhabitats when threatened during the day, and they associate closely with the shelter throughout the night. Successful attacks by predators on small Dascyllus juveniles appear to involve individuals that either had been chased off the shelter by another resident or were located at the margin of the microhabitat (Holbrook and Schmitt, in press). Our observations also suggest that the Dascyllus individuals at greatest risk to predators (i.e., chased off or relegated to the periphery) tend to be the smallest present in a group, and in this study both the magnitude of mortality and the intensity of density dependence varied inversely with body size. This is consistent with the finding that in general, mortality rates of juvenile reef fishes decrease with increasing body size (both within and across species), a pattern that is widely believed to reflect declining vulnerability to predators (Victor 1986, Eckert 1987, Shulman and Ogden 1987, Sale and Ferrell 1988, Forrester 1990, Hixon 1991, Hixon and Beets 1993, Levin 1994, Cart and Hixon 1995). Hixon and Carr (1997) demonstrated that density-dependent mortality of early postsettlement stages of a Caribbean damselfish (Chromis cyanea), which also shelters in coral heads, was caused by the combined actions of both transient and resident predators. The potential for density-dependent mortality in young reef fishes is likely to be great when vulnerable size classes compete for discrete shelters from predators.
Our results highlight potentially serious problems with inferring the form of postsettlement mortality from a relationship between the abundance of a cohort at successive time periods (i.e., settler-recruit relationships). The first problem is underscored by the failure of any simple model to fit our data for the within-site scale, and the second by the good linear fits to data from the among-site scale. In the former case, settler densities did not covary systematically with other factors that contributed to mortality. Such variability, which is probably quite common in nature, can easily mask detection of density dependence. In the latter case, the relationships between the number of settlers and the number alive 14 d later were generally well described by a linear model, despite compelling evidence that compensatory mortality existed and was responsible for a large amount of the reduction in variance of abundance of new Dascyllus settlers.
The relatively good linear fits to our settler-recruit data imply two things. First, comparing linear and nonlinear fits to a settler-recruit relationship can be insensitive to the detection of density dependence. The second implication is that density dependence often may not be sufficiently strong to obviate an approximately linear relationship between larval supply and subsequent recruitment. The degree to which settler – recruit relationships will deviate from linearity is dependent in large part on the intensity of density dependence, so the approach potentially can provide a means to evaluate the contribution of density dependence to the modification of settlement patterns.
If the essential challenge for species with demographically open populations is to estimate the relative influence of the multiple processes that determine patterns of abundance, it will be necessary to develop appropriate frameworks to estimate the contributions of each process to the pattern of interest (e.g., Steele 1996). Here we used a simple but potentially general framework to address two long-standing questions for reef fishes: to what degree were spatial patterns that originated at settlement modified by subsequent processes, and what was the relative importance of larval supply, and density-independent and density-dependent mortality in determining the average abundance of an older life stage? Because our experimental sites contained much lower densities and different combinations of age classes of Dascyllus than occur on natural reefs in Moorea, the particular results we obtained from the application of the framework are perhaps less important than the general approach and its strengths and limitations.
For the first question we calculated the reduction after 2 wk in the spatial variance in abundance of new settler cohorts that originated at settlement. In all cases, early postsettlement mortality resulted in [greater than]80% reductions in the variance. If there had been no density dependence in the mortality, all of the reductions would have been attributed to the effect of density-independent losses on lowering mean abundance and the modification would only have been quantitative. Given that density dependence occurred, the more difficult issue was partitioning the reductions in variance into the density-independent and -dependent components. Our procedure indicated that, despite juvenile densities that were only 15-25% of ambient, density-dependent mortality accounted for about half of the [greater than]80% reductions in variance of new settler abundance.
We also used the framework to estimate the relative importance of primary recruitment limitation (i.e., too low a supply of larvae to saturate resources in the absence of any postsettlement mortality; Victor 1986), density-independent and density-dependent mortality in determining the average number of settlers that survived to 2 wk. For each species, the proportion of the abundance cost attributable to primary recruitment limitation was the largest ([approximately] 70%), and that due to density-dependent mortality was the smallest ([approximately] 12%). The rather substantial importance of density-dependent mortality was suprising given the degree to which experimental microhabitats were undersaturated. This was not a result of us following an unusually large settlement event as the pulse followed was well within the range of variation of six other settlement pulses we measured at Moorea.
The “intra-cohort” framework we used is a potentially powerful tool to estimate the relative importance of mortality components and larval supply to patterns of abundance of species with demographically open populations. There are, however, important constraints and limitations of the approach. Among them are the need to obtain reliable estimates of the density-independent mortality rate and saturation density (both of which potentially can be estimated quite well in field experiments) and to apply an appropriate functional relationship to the data. These become problematic when data are noisy, which typically is the case when dealing with mortality. However, a larger constraint occurs when intercohort processes and/or competition with other species are important sources of density-dependent loss, as was the case for the Dascyllus species. This limitation is well illustrated by the fact we could not apply the framework to data from our within-site trials because of the importance of intercohort and interspecific interactions. Other techniques will need to be developed when intercohort processes have a substantial influence on mortality.
The growing evidence that newly settled colonists are the life stage where density-dependent mortality is likely to occur or be most intense emphasizes the need to focus attention on this critical but brief period for reef fishes. We also need to understand the general attributes of reef fishes that influence the likelihood of density-dependent mortality beyond the obvious issue of larval supply relative to availability of suitable habitat. There is an even greater need to develop a better suite of tools to estimate the importance of larval supply relative to subsequent processes. Continued failure to do so will hinder our ability to assess the contributions of the multiple processes that determine patterns of abundance and dynamics, and to our understanding of how populations of species with dispersing life stages are regulated.
ACKNOWLEDGMENTS
We thank K. Seydel, C. Shuman and P. Raimondi for assistance in the field, C. Briggs, M. Carr, C. Osenberg, S. Gaines, P. Raimondi, A. Stewart-Oaten for critical discussions, two anonymous reviewers for valuable input on the manuscript, and B. Williamson, W. Holbrook and M. Schmitt for logistical assistance. We especially want to acknowledge discussions with Craig Osenberg which were instrumental in the development of the intracohort framework, with Steve Gaines regarding the contribution of different sources of mortality to reduction of variance and Pete Raimondi about our experimental design. We also appreciate the help and hospitality of Frank Murphy and Steve Strand of the UC Berkeley Gump South Pacific Biological Station. The research was supported by the National Science Foundation (OCE 9503305), and funds for the infrared video equipment were provided by the W. M. Keck Foundation. This paper is Contribution No. 55 of the Gump South Pacific Biological Station.
LITERATURE CITED
Allen, G. R. 1991. Damselfishes of the world. Mergus Publishers Hans A. Baensch, Melle, Germany.
Anderson, T. W. 1993. An evaluation of density dependence in a marine temperate reef fish: competitive and predatory effects. Dissertation. University of California, Santa Barbara, California, USA.
Behrents, K. C. 1987. The influence of shelter availability on recruitment and early juvenile survivorship of Lythrypnus dalli Gilbert (Pisces: Gobiidae). Journal of Experimental Marine Biology and Ecology 107:45-59.
Booth, D.J. 1995. Juvenile groups in a coral-reef damselfish: density-dependent effects on individual fitness and population demography. Ecology 76:91-106.
Booth, D. J., and G. A. Beretta. 1994. Seasonal recruitment and habitat associations in a guild of pomacentrid reef fishes at two sites around St. Thomas, USVI. Coral Reefs 13: 81-89.
Booth, D. J., and D. H. Brosnan. 1995. The role of recruitment dynamics in rocky shore and coral reef fish communities. Advances in Ecological Research 26:309-385.
Caley, M. J., M. H. Carr, M. A. Hixon, T. P. Hughes, G. P. Jones, and B. A. Menge. 1996. Recruitment and the local dynamics of open marine populations. Annual Review of Ecology and Systematics 27:477-500.
Carr, M. H., and M. A. Hixon. 1995. Predation effects on early post-settlement survivorship of coral-reef fishes. Marine Ecology Progress Series 124:31-42.
Coates, D. 1980. The discrimination of and reactions towards predatory and non-predatory species of fish by humbug damselfish, Dascyllus aruanus (Pisces, Pomacentridae). Zeitschrift fuer Tierpsychologie 52:347-354.
Connell, S. D. 1997. The relationship between large predatory fish and recruitment and mortality of juvenile coral reef-fish on artificial reefs. Journal of Experimental Marine Biology and Ecology 209:261-278.
Doherty, P. J. 1981. Coral reef fishes: recruitment limited assemblages? Proceedings of the 4th International Coral Reef Symposium 4:465-470.
—–. 1982. Some effects of density on the juveniles of two species of tropical territorial damselfishes. Journal of Experimental Marine Biology and Ecology 65:249-261.
—–. 1983. Tropical territorial damselfishes: is density limited by aggression or recruitment? Ecology 64:176-190.
—–. 1991. Spatial and temporal patterns in recruitment. Pages 261-292 in P. F. Sale, editor. The ecology of fish on coral reefs. Academic Press, San Diego, California, USA.
Doherty, P. J., and A. J. Fowler. 1994a. An empirical test of recruitment limitation in a coral reef fish. Science 263:935-939.
Doherty, P. J., and A. J. Fowler. 1994b. Demographic consequences of variable recruitment to coral reef fish populations: a congeneric comparison of two damselfishes. Bulletin of Marine Science 54:297-313.
Doherty, P. J., and D. McB. Williams. 1988. The replenishment of coral reef fish populations. Oceanography and Marine Biology Annual Review 26:487-551.
Eckert, G. J. 1987. Estimates of adult and juvenile mortality for labrid fishes at One Tree Reef, Great Barrier Reef. Marine Biology 95:167-171.
Forrester, G. E. 1990. Factors influencing the juvenile demography of a coral reef fish. Ecology 71:1666-1681.
—–. 1995. Strong density-dependent survival and recruitment regulate the abundance of a coral reef fish. Oecologia 103:275-282.
Frederick, J. L. 1997. Post-settlement movement of coral reef fishes and bias in survival estimates. Marine Ecology Progress Series 150:65-74.
Galzin, R., and J. P. Pointier. 1985. Moorea Island, Society Archipelago. Proceedings of the 5th International Coral Reef Symposium 1:73-102.
Hassell, M. P. 1986. Detecting density dependence. Trends in Ecology and Evolution 1:90-93.
Hixon, M. A. 1991. Predation as a process structuring coral-reef fish communities. Pages 475-508 in P. F. Sale, editor. The ecology of fish on coral reefs. Academic Press, San Diego, California, USA.
Hixon, M. A., and J. P. Beets. 1993. Predation, prey refuges, and the structure of coral-reef fish assemblages. Ecological Monographs 63:77-101.
Hixon, M. A., and M. H. Carr. 1997. Synergistic predation, density dependence, and population regulation in marine fish. Science 227:946-949.
Holbrook, S. J., and R. J. Schmitt. 1997. Settlement patterns and process in a coral reef damselfish: in situ nocturnal observations using infrared video. Proceedings of the 8th International Coral Reef Symposium 2:1143-1148.
Holbrook, S. J., and R. J. Schmitt. In press. In situ nocturnal observations of reef fishes using infrared video. Cybium.
Jones, G. P. 1984. Population ecology of the temperate reef fish Pseudolabrus celidotus Bloch & Schneider (Pisces, Labridae). I. Factors influencing recruitment. Journal of Experimental Marine Biology and Ecology 75:257-276.
—–. 1987a. Competitive interactions among adults and juveniles in a coral reef fish. Ecology 68:1534-1547.
—–. 1987b. Some interactions between residents and recruits in two coral reef fishes. Journal of Experimental Marine Biology and Ecology 114:169-182.
—–. 1988. Experimental evaluation of the effects of habitat structure and competitive interactions on the juveniles of two coral reef fishes. Journal of Experimental Marine Biology and Ecology 123:115-126.
—–. 1990. The importance of recruitment to the dynamics of a coral reef fish population. Ecology 71:1691-1698.
—–. 1991. Postrecruitment processes in the ecology of coral reef fish populations: a multifactorial perspective. Pages 294-330 in P. E Sale, editor. The ecology of fish on coral reefs. Academic Press, San Diego, California, USA.
Keough, M. J., and B. J. Downes. 1982. Recruitment of marine invertebrates: the role of active larval choices and early mortality. Oecologia 54:348-352.
Levin, P.S. 1994. Fine-scale temporal variation in recruitment of a temperate demersal fish: the importance of settlement versus post-settlement loss. Oecologia 97:124133.
Mapstone, B. D., and A. J. Fowler. 1988. Recruitment and the structure of assemblages of fish on coral reefs. Trends in Ecology and Evolution 3:72-77.
Meekan, M. G., M. J. Milicich, and P. J. Doherty. 1993. Larval production drives temporal patterns of larval supply and recruitment of a coral reef damselfish. Marine Ecology Progress Series 93:217-225.
Milicich, M. J., M. G. Meekan, and P. J. Doherty. 1992. Larval supply: a good predictor of recruitment of three species of reef fishes (Pomacentridae). Marine Ecology Progress Series 86:153-166.
Murdoch, W. W. 1994. Population regulation in theory and practice. Ecology 75:271-287.
Nisbet, R. M., W. W. Murdoch, and A. Stewart-Oaten. 1996. Consequences for adult fish stocks of human-induced mortality on immatures. Pages 257-277 in R. J. Schmitt and C. W. Osenberg, editors. Detecting ecological impacts. Academic Press, San Diego, California, USA.
Olafsson, E. B., C. H. Peterson, and W. G. Ambrose, Jr. 1994. Does recruitment limitation structure populations and communities of macro-invertebrates in marine soft sediments: the relative significance of pre- and post-settlement processes. Annual Review of Oceanography and Marine Biology 32:65-109.
Pitcher, C. R. 1988. Validation of a technique for reconstructing daily patterns in the recruitment of coral reef damselfish. Coral Reefs 7:105-111.
Robertson, D. R. 1988. Abundances of surgeonfishes on patch-reefs in Caribbean Panama: due to settlement, or post-settlement events? Marine Biology 97:495-501.
—–. 1992. Patterns of lunar settlement and early recruitment in Caribbean reef fishes at Panama. Marine Biology 114:527-537.
Sale, P. F., W. A. Douglas, and P. J. Doherty. 1984. Choice of microhabitats by coral reef fishes at settlement. Coral Reefs 3:91-99.
Sale, P. E, and D. J. Ferrell. 1988. Early survivorship of juvenile coral reef fishes. Coral Reefs 4:117-124.
Schmitt, R. J., and S. J. Holbrook. 1996. Local-scale patterns of larval settlement in a planktivorous damselfish – do they predict recruitment? Marine and Freshwater Research 47:449-463.
Schmitt, R. J., and S. J. Holbrook. In press. Temporal patterns of settlement of coral reef damselfishes in the genus Dascyllus. Cybium.
Shulman, M. J., and J. C. Ogden. 1987. What controls tropical reef fish populations: recruitment or benthic mortality? An example in the Caribbean reef fish Haemulon flavolineatum. Marine Ecology Progress Series 39:233-242.
Steele, M. A. 1996. The relative importance of processes affecting recruitment of two temperate reef fishes. Ecology 78:129-145.
—–. 1997. Population regulation by post-settlement mortality in two temperate reef fishes. Oecologia 112:64-74.
Tupper, M., and R. G. Boutilier. 1995. Effects of conspecific density on settlement, growth and post-settlement survival of a temperate reef fish. Journal of Experimental Marine Biology and Ecology 191:209-222.
Tupper, M., and W. Hunte. 1994. Recruitment dynamics of coral reef fishes in Barbados. Marine Ecology Progress Series 108:225-235.
Victor, B.C. 1983. Recruitment and population dynamics of a coral reef fish. Science 219:419-420.
—–. 1986. Larval settlement and juvenile mortality in a recruitment-limited coral reef fish population. Ecological Monographs 56:145-160.
Warner, R. R., and T. P Hughes. 1988. The population dynamics of reef fishes. Proceedings of the 6th International Coral Reef Symposium 1:149-155.
Wellington, G. M., and B.C. Victor. 1989. Planktonic larval duration of one hundred species of Pacific and Atlantic damselfishes (Pomacentridae). Marine Biology 101:557567.
Williams, D. McB., S. English, and M. J. Milicich. 1994. Annual recruitment surveys of coral reef fishes are good indicators of patterns of settlement. Bulletin of Marine Science 54:314-331.
COPYRIGHT 1999 Ecological Society of America
COPYRIGHT 2000 Gale Group
