Bigger is not necessarily better: empirical tests show that dispersal proxies misrepresent actual dispersal ability.

Tests for the role of species' relative dispersal abilities in ecological and biogeographical models rely heavily on dispersal proxies, which are seldom substantiated by empirical measures of actual dispersal. This is exemplified by tests of dispersal-range size relationships and by metacommunity research that often features invertebrates, particularly freshwater insects. Using rare and unique empirical data on dispersal abilities of caddisflies, we tested whether actual dispersal abilities were associated with commonly used dispersal proxies (metrics of wing size and shape; expert opinion). Across 59 species in 12 families, wing morphology was not associated with actual dispersal. Within some families, individual wing metrics captured some dispersal differences among species, although useful metrics varied among families and predictive power was typically low. Dispersal abilities assigned by experts were either no better than random or actually poorer than random. Our results cast considerable doubt on research underpinned by dispersal proxies and scrutiny of previous research results may be warranted. Greater progress may lie in employing innovative survey and experimental design to measure actual dispersal in the field.

A large-scale field experiment across six rivers illustrates how the effects of resource enrichment are context dependent.

Resource supplementation can increase species richness and change the faunal composition of communities, but experiments have produced variable outcomes. An often overlooked element is that species richness can only increase if new taxa can disperse to resource-rich locations and invade established, local communities. We experimentally increased a basal resource (detritus) in six rivers in south-eastern Australia by driving wooden stakes into the riverbed to increase retention of detritus. Control sites were left untreated. Sites were located in agricultural sections with mostly cleared vegetation, but with intact (uncleared) reference sites upstream to provide sources of prospective colonists. We measured channel retentiveness and sampled benthic detritus and invertebrates before and after manipulation. We tested whether: greater retentiveness increased detritus densities, species richness and abundances and altered faunal composition; manipulation sites reached bio-equivalence with reference sites; new species arose from upstream reference areas; and whether outcomes were consistent across rivers. Only three rivers gained increases in detritus densities. All had low pre-existing amounts of in-stream wood compared with rivers that did not respond to treatment. Two rivers (Hughes Creek, Seven Creeks) gained higher species richness and invertebrate densities within 12 months and reached bio-equivalence with reference sites. In contrast, Turtons Creek showed species turnover through replacement of individuals. Only in Hughes Creek was there evidence of successful dispersal from the upstream reference area. The outcomes show that the effects of resource supplementation vary between rivers and suggest that pre-existing conditions (e.g. channel retentiveness) may cause these differences, providing clear evidence of context dependence.

Multiyear resource enrichment creates persistently higher species diversity in a landscape-scale field experiment.

Short-term resource enrichment can increase species diversity in communities, but prolonged resource enrichment may result in either a diversity collapse or persistent high species diversity if fluctuation-dependent mechanisms of species coexistence are triggered. We tested the effects of resource enrichment on stream invertebrates by boosting densities of benthic detritus. In a 22-km stream length, we used wooden stakes to enhance retention of detritus at 40-m-long sites; other sites acted as controls. Detritus and invertebrates were sampled prior to treatment and then 1, 2, and 5 yr later. Previously, we reported that detrital densities, species diversity, and densities increased at enrichment sites after 12 months. Here we report that similar increases occurred 2 and 5 yr after manipulation. Prolonged resource enrichment produced persistently higher species diversity without loss of any taxa from the species pool, despite strong shifts in faunal composition in response to environmental variation, including a 1-in-100-yr flood. Detritus densities set upper limits to the densities of common taxa. Positive relations between invertebrate and detritus densities (density-resource relationships) took a variety of forms and showed that detritus was an essential resource for some taxa and a substitutable resource for others. Species varied in the minimum amount of detritus required for presence at a site, and population densities increased strongly from low densities when detritus was increased. These outcomes suggest that fluctuation-dependent mechanisms of coexistence enabled new taxa to coexist at manipulation sites, with relative nonlinear averaging of competition and the storage effect most likely to be in play. Two characteristics of the study stream underpin diversity increases with resource enrichment: overall low background densities of detritus and species that are able to disperse successfully from upstream areas where detritus is more abundant. Thus, the effects of resource enrichment are context dependent.

Just add water: rapid assembly of new communities in previously dry riverbeds, and limited long-distance effects on existing communities.

Barriers preventing species from dispersing to a location can have a major influence on how communities assemble. Dispersal success may also depend on whether dispersers have to colonise an established community or a largely depauperate location. In freshwater systems, dams and weirs have fragmented rivers, potentially limiting dispersal of biota along rivers. Decommissioning aqueducts on two weirs, each within a tributary of different regulated rivers, delivered flow to previously dry riverbeds and additional flows to the main stem, regulated rivers further downstream. This provided an opportunity to test how removal of dispersal constraints affected community assembly in new habitats and whether changed dispersal can alter existing communities. The results were very similar for the two systems. Even with dispersal constrained via reduced drift rates, the new communities in the newly formed habitat in tributaries rapidly resembled unimpacted reference communities that were the source of colonists. For established communities (regulated rivers), greater flow increased the densities of filter feeders but this was due to greater areas of fast-flowing habitat (a change in environmental constraints) rather than higher dispersal rates. Our study illustrates that communities can quickly re-assemble when natural channels that have been dry for decades are re-wetted by flows that deliver dispersers from intact locations upstream. Nevertheless, boosting flows and concomitant densities of dispersers had no strong effects on existing communities. Instead, increased discharges effected a reduction in environmental constraints, which altered trophic structure. Thus, increases in discharge and dispersal produced different outcomes in new versus established communities.

Avoidance and aggregation create consistent egg distribution patterns of congeneric caddisflies across spatially variable oviposition landscapes.

Amongst oviparous animals, the spatial distribution of individuals is often set initially by where females lay eggs, with potential implications for populations and species coexistence. Do the spatial arrangements of oviposition sites or female behaviours determine spatial patterns of eggs? The consequences of spatial patterns may be context independent if strong behaviours drive patterns; context dependent if the local environment dominates. We tested these ideas using a guild of stream-dwelling caddisflies that oviposit on emergent rocks, focussing on genera with contrasting behaviours. In naturally occurring oviposition landscapes (riffles with emergent rocks), we surveyed the spatial arrangement and environmental characteristics of all emergent rocks, identified and enumerated egg masses on each. Multiple riffles were surveyed to test for spatially invariant patterns and behaviours. In landscapes, we tested for spatial clumping of oviposition sites exploited by each species and for segregation of congeneric species. At oviposition sites, we characterised the frequency distributions of egg masses and tested for species associations. Genus-specific behaviours produced different spatial patterns of egg masses in the same landscapes. Congregative behaviour of Ulmerochorema spp. at landscape scales and an aggregative response at preferred oviposition sites led to clumped patterns, local aggregation and species overlap. In contrast, avoidance behaviours by congeners of Apsilochorema resulted in no or weak clumping, and species segregation in some landscapes. Spatial patterns were consistent across riffles that varied in area and oviposition site density. These results suggest that quite different oviposition behaviours may be context independent, and the consequences of spatial patterns may be spatially invariant also.

Aquatic versus Terrestrial Insects: Real or Presumed Differences in Population Dynamics?

The study of insect populations is dominated by research on terrestrial insects. Are aquatic insect populations different or are they just presumed to be different? We explore the evidence across several topics. (1) Populations of terrestrial herbivorous insects are constrained most often by enemies, whereas aquatic herbivorous insects are constrained more by food supplies, a real difference related to the different plants that dominate in each ecosystem. (2) Population outbreaks are presumed not to occur in aquatic insects. We report three examples of cyclical patterns; there may be more. (3) Aquatic insects, like terrestrial insects, show strong oviposition site selection even though they oviposit on surfaces that are not necessarily food for their larvae. A novel outcome is that density of oviposition habitat can determine larval densities. (4) Aquatic habitats are often largely 1-dimensional shapes and this is presumed to influence dispersal. In rivers, drift by insects is presumed to create downstream dispersal that has to be countered by upstream flight by adults. This idea has persisted for decades but supporting evidence is scarce. Few researchers are currently working on the dynamics of aquatic insect populations; there is scope for many more studies and potentially enlightening contrasts with terrestrial insects.

Dispersal traits may reflect dispersal distances, but dispersers may not connect populations demographically.

Ecological traits that reflect movement potential are often used as proxies for measured dispersal distances. Whether such traits reflect actual dispersal is often untested. Such tests are important because maximum dispersal distances may not be achieved and many dispersal events may be unsuccessful (without reproduction). For insects, many habitat patches harbour 'resident' species that are present as larvae (sedentary) and adults (winged and dispersing), and 'itinerant' species present only as adults that have dispersed from elsewhere and fail to reproduce. We tested whether itinerancy patterns were temporally consistent, and whether itinerant and resident species differed in wing morphology, a strong correlate of flight capability. Over 3 years and at multiple locations in a 22 km stream length, we sampled larvae and adults of caddisflies in the genus Ecnomus to categorize species as residents or itinerants. Flight capacity was measured using wing size (length and area) and shape parameters (aspect ratio and the second moment of wing area). Three species of Ecnomus were residents and three species were itinerants, and patterns were consistent over 3 years. On average, itinerant species had larger wings, suggesting a greater capacity to fly long distances. Wing shape differed between species, but did not differ systematically between residents and itinerants. Wing morphology was associated with actual but not effective dispersal of some species of Ecnomus. Morphological traits may have weak explanatory power for hypotheses regarding the demographic connectedness of populations, unless accompanied by data demonstrating which dispersers contribute new individuals to populations.

A landscape-scale field experiment reveals the importance of dispersal in a resource-limited metacommunity.

Dispersal may play a strong role in driving species diversity across landscapes. Theoretically, dispersal permits species to remain extant within a metacommunity, even if they are periodically excluded from some local communities. Field tests of dispersal effects are difficult, and most non-experimental data suggest that environmental conditions play the predominant role in setting species diversity. However, most such studies cannot differentiate between patterns caused primarily by dispersal constraints vs. abiotic factors vs. biotic constraints (e.g., priority effects). In 22 km of a sand-bed stream in southeastern Australia, strong longitudinal gradients in the abiotic environment and detritus densities (resources) mean that downstream locations have abiotic conditions that may be physiologically stressful, low resource densities and low species diversities. We experimentally increased the retention of detritus over 40 m stream lengths along the gradient, with other sites acting as controls. If dispersal is constrained, or abiotic or biotic factors primarily control community structure, then we predicted that increasing resources would result in no change in species composition. Alternatively, if dispersal is common, we predicted that species diversity would increase at treatment sites through colonization by species able to tolerate abiotic conditions downstream and able to invade established communities. Invertebrates were sampled prior to manipulation and then four times (1, 4, 9 and 12 months) following manipulation. Detrital standing stocks increased by an order of magnitude at treatment sites. Over 1 yr, invertebrate densities and species richness also increased in treatment sites. Effect sizes were strong in middle and downstream areas, which were colonized by upstream species. Thus, faunal composition of the downstream treatment sites became more similar to upstream locales, and ß-diversity across treatment sites declined as α-diversity increased. Out of 54 common taxa, roughly half responded to the experiment; responders and non-responders had similar proportions of upstream specialists and of different functional feeding groups. Contrary to many non-experimental studies, our results demonstrate that extensive dispersal can be very important and, for many species, potentially more important than the abiotic environment or biotic constraints in affecting community structure when adequate resources are available.

Population densities and density-area relationships in a community with advective dispersal and variable mosaics of resource patches.

Many communities comprise species that select resources that are patchily distributed in an environment that is otherwise unsuitable or suboptimal. Effects of this patchiness can depend on the characteristics of patch arrays and animal movements, and produce non-intuitive outcomes in which population densities are unrelated to resource abundance. Resource mosaics are predicted to have only weak effects, however, where patches are ephemeral or organisms are transported advectively. The running waters of streams and benthic invertebrates epitomize such systems, but empirical tests of resource mosaics are scarce. We sampled 15 common macroinvertebrates inhabiting distinct detritus patches at four sites within a sand-bed stream, where detritus formed a major resource of food and living space. At each site, environmental variables were measured for 100 leaf packs; invertebrates were counted in 50 leaf packs. Sites differed in total abundance of detritus, leaf pack sizes and invertebrate densities. Multivariate analysis indicated that patch size was the dominant environmental variable, but invertebrate densities differed significantly between sites even after accounting for patch size. Leaf specialists showed positive and strong density-area relationships, except where the patch size range was small and patches were aggregated. In contrast, generalist species had weaker and variable responses to patch sizes. Population densities were not associated with total resource abundance, with the highest densities of leaf specialists in sites with the least detritus. Our results demonstrate that patchy resources can affect species even in communities where species are mobile, have advective dispersal, and patches are relatively ephemeral.

Lasting effects of maternal behaviour on the distribution of a dispersive stream insect.

1. Predicting population dynamics at large spatial scales requires integrating information about spatial distribution patterns, inter-patch movement rates and within-patch processes. Advective dispersal of aquatic species by water movement is considered paramount to understanding their population dynamics. Rivers are model advective systems, and the larvae of baetid mayflies are considered quintessential dispersers. Egg laying of baetids along channels is patchy and reflects the distribution of oviposition sites, but larvae are assumed to drift frequently and far, thereby erasing patterns created during oviposition. Dispersal kernels are often overestimated, however, and empirical tests of such assumptions are warranted because of the pivotal role distribution patterns can have on populations. 2. We tested empirically whether the egg distribution patterns arising from oviposition behaviours persisted and were reflected in the distribution patterns of larval Baetis rhodani. In field surveys, we tested for associations between egg mass and larval densities over 1 km lengths of four streams. A control species, the mayfly Ephemerella ignita, was employed to test for covarying environmental factors. We estimated drift rates directly to test whether larvae dispersed between riffles (patches of high egg mass density) and whether drift rates were density-dependent or density-related - expected outcomes if drift erases patterns established by maternal behaviours. 3. Positive associations between egg masses and larval benthic densities were found for neonate and mid-stage larvae of Baetis, but not the control species, suggesting persistence of the patchy distribution patterns established at oviposition. Drift rates were high, and riffles were net exporters of neonate and mid-stage larvae, but drift rates were unrelated to benthic densities and few drifters reached the next riffle. Riffles were sinks for large larvae, suggesting ontogenetic shifts in habitat use, but little long-distance dispersal. 4. Overall, the results suggest that most neonate and mid-stage larvae of B. rhodani remain close to the natal riffle, and late-stage larvae disperse shorter distances than routinely assumed. The persistence of maternal effects on distribution patterns well into juvenile life of an allegedly iconic disperser suggests that traditional models of how dispersal influences the population dynamics of many lotic invertebrates may be incorrect.

Environmental constraints on oviposition limit egg supply of a stream insect at multiple scales.

Species with complex life cycles pose challenges for understanding what processes regulate population densities, especially if some life stages disperse. Most studies of such animals that are thought to be recruitment limited focus on the idea that juvenile mortality limits the density of recruits (and hence population density), fewer consider the possibility that egg supply may be important. For species that oviposit on specific substrata, environmental constraints on oviposition sites may limit egg supply. Female mayflies in the genus Baetis lay egg masses on the underside of stream rocks that emerge above the water's surface. We tested the hypothesis that egg mass densities are constrained by emergent rock densities within and between streams, by counting egg masses on emergent rocks. All emergent rocks were counted along 1-km lengths of four streams, revealing significant variation in emergent rock density within streams and a more than three-fold difference between streams. In each stream, egg mass density increased with the density of emergent rocks in 30-m stretches. We used regression equations describing these small-scale relationships, coupled with the large-scale spatial variation of emergent rocks, to estimate egg mass densities for each 1-km stream length, a scale relevant to population processes. Scaled estimates were positively associated with emergent rock density and provided better estimates than methods that ignored environmental variation. Egg mass crowding was inversely related to emergent rock density at the stream scale, a pattern consistent with the idea that oviposition substrata were in short supply in streams with few emergent rocks, but crowding did not compensate entirely for differences in emergent rock densities. The notion that egg supply, not larval mortality, may limit population density is an unusual perspective for stream insects. Environmental constraints on egg supply may be widespread among other species with specialised oviposition behaviours.

Does dispersal control population densities in advection-dominated systems? A fresh look at critical assumptions and a direct test.

1. In advection-dominated systems (both freshwater and marine), population dynamics are usually presumed to be dominated by the effects of migrants dispersing by advection, especially over the small spatial scales at which populations can be studied, but few studies have tested this presumption. We tested the hypothesis that benthic densities are controlled by densities of dispersers for two aquatic insects in upland streams. 2. Our study animals were two species of caddisflies (Hydropsychidae), which become sedentary filter-feeders following settlement onto substrata. Densities of dispersers in the drift (advective dispersal) were quantified using nets placed along the upstream edges of riffles, where the latter abruptly abutted a slower, upstream run. Settlement was estimated at each site using brick pavers, half of which had been fenced to prevent colonization of their top surfaces by walking hydropsychids, thus allowing us to distinguish also the mode of movement during settlement. 3. First through fifth instars of two species, Smicrophylax sp. AV2 and Asmicridea sp. AV1, were abundant and showed disparate results. Drift and settlement were relatively strongly related for Smicrophylax. The best fit lines were shown by second and third instars settling on plain bricks, suggesting that drift played a strong role in settlement, but that some drifters dropped to the bottom and located substrata by walking. Quantile regression suggested that drift sets limits to settlement in this species and that settlement success was highly variable. In contrast, settlement by Asmicridea was poorly related to drift; settlers were mainly individuals re-dispersing within sites. 4. Smicrophylax densities appear to be controlled by dispersal from upstream, but benthic density of Asmicridea is more likely linked to local demography. Our data demonstrate the dangers of assuming that supposedly drift-prone species can all be modelled in the same way. Alternative models emphasizing little or different kinds of movement should be considered. Variability in oviposition coupled with weak dispersal, for example, is a viable alternative hypothesis to explain variation in benthic density along channels. Moreover, the constraints on settlement of Smicrophylax show that immigrants into sites can be in short supply, an hypothesis rarely considered in stream research.

Separating natural responses from experimental artefacts: habitat selection by a diadromous fish species using odours from conspecifics and natural stream water.

Animals use sensory stimuli to assess and select habitats, mates and food as well as to communicate with other individuals. One way they do this is to use olfaction, whereby they identify and respond to chemical cues. All organisms release odours, which mix with other chemical substances and ambient environmental conditions. The result is that animals are frequently immersed in a complex, highly dynamic sensory environment where they must identify and respond to only some of the potential stimuli they encounter in the face of significant levels of background noise. Understanding how organisms respond to different chemical cues is therefore dependent on knowing how these responses might be influenced by potential interactions with other stimuli. To test this, we examined whether the diadromous fish Galaxias maculatus was attracted to conspecific odours and whether this response differed when cues were offered in an artificial environment lacking other potential chemical stimuli (tap water) or a more natural background environment (stream water). We found that (1) fish responded to both natural stream water odours and those from conspecifics but the response to the latter was stronger; (2) the attraction to conspecific odours was stronger in tap water than in stream water, which indicates the importance of these odours may be overestimated when they are offered in artificial media. We also conducted a brief literature review, which confirmed that artificial media are commonly used in experiments and that the background environment is often not considered. Our results show that future research testing the responses of organisms to auditory, olfactory and visual cues should carefully consider the context in which cues are presented. Without doing so, such studies may inaccurately assess the importance of sensory cues in natural situations in the wild.

Experimental evidence for physical cues involved in oviposition site selection of lotic hydrobiosid caddis flies.

Cues involved in determining the distribution of invertebrate propagules within a stream landscape contribute greatly to our knowledge of the supply and arrangement of new recruits and thus an improved understanding of factors that might ultimately affect population parameters. Previous observations indicated that both current velocity and rock size were important determinants of the egg mass distribution of certain hydrobiosid caddis flies that lay their eggs in single masses beneath emergent rocks. These observations were tested experimentally in a temperate, upland Australian stream. Manipulations of current speed confirmed that females of Ulmerochorema sp. and the Taschorema complex deposited more eggs on rocks in elevated current speeds (>0.40 ms(-1)) whereas Apsilochorema sp. deposited more eggs on rocks in slow currents (<0.30 ms(-1)). This latter result did not coincide with previous observational data. The anomaly between observational and experimental data, however, was reconciled by the outcome of a further experiment that tested the influence of the emergent or 'landing pad' size of rocks as the abundance of Apsilochorema egg masses increased with landing pad size independent of the prevailing flow conditions. Landing pad size did not influence the abundance of egg masses of Taschorema or Ulmerochorema. Patterns of female visits to rocks indicated that taxa might distinguish between favoured egg-laying sites prior to landing on rocks. Large aggregations of adult male and female Ulmerochorema collected from rocks favoured for oviposition provide indirect evidence for mating induced swarming behaviour associated with oviposition site selection. This study provides a framework for more sophisticated questions relating to the influence of oviposition site selection on structuring populations of lotic macroinvertebrates.

Spatial and temporal variation in recruitment and its effects on regulation of parasite populations.

Variation in recruitment rates of parasites to hosts possibly contributes significantly to fluctuations in parasite numbers, yet is almost never measured directly in the field. I measured the variation in recruitment rates of three species of parasitic mites living in two species of freshwater mussels over several spatial and temporal scales. I also examined separately the effect of spatial dispersion of hosts on mite recruitment. Uninfected hosts of both species were placed out each month, for a period of a month, for 2 years at one site and 12 months at another. Mussels of both species were collected simultaneously each month so that abundance of recruiting mites could be compared to mean abundances of mites in hosts at that time. To test the effect of host dispersion on recruitment rates, mussels were set out in clumped and regular patterns in a separate experiment. Overall, recruitment rates were often high but also varied substantially between sites, seasons, years and months. The likely impacts differed between mite species with one probably affected strongly by recruitment variation, while abundances of the other two were not. Populations of the latter two species are probably regulated by intraspecific competition for mates and egg-laying sites. Sampling data are often used to estimate recruitment rates but the latter should be measured, if possible, by exposing uninfected hosts for a known period of time. This direct method reveals patterns of recruitment that cannot be deduced from sampling data. The lack of information on recruitment variation represents a major gap in our knowledge of parasite populations.

Competition between mobile species using patchy resources: an example from a freshwater, symbiotic assemblage.

Three species of freshwater mites that are symbiotic with mussels in St Mark's River, north Florida, have consistently high rates of colonization to and occupy high proportions of mussels. The mites Unionicola poundsi and U. serrata are territorial and have limited numbers/host, whereas the non-territorial U. abnormipes has highly variable numbers/host. U. abnormipes and U. serrata are most common in Villosa villosa and Uniomerus declivis respectively, patterns that can not be explained by host species preferences, whereas U. poundsi is equally abundant in both host species. Field experiments showed that both U. poundsi and U. serrata were limited most by intraspecific competition between adult mites, presumably for access to food and oviposition sites. Additionally, U. serrata did not remain within small hosts, most of which were V. villosa. In contrast, numbers of U. abnormipes were limited by both other mite species although the nature of the interactions differed. U. serrata may prey on U. abnormipes when they co-occur, whereas U. poundsi probably only excludes U. abnormipes from certain areas within hosts. Hence, U. abnormipes occurs mostly in V. villosa because most of these mussels do not contain U. serrata, but even so its numbers are still depressed by U. poundsi. The results were consistent with the general expectation of Holmes and Price (1986) that parasite assemblages where species have high colonization levels should be organized primarily by biotic interactions. However, specific outcomes of competition between mites were consistent with the more general model of Levins (1979) for competition between species using variable resources. Failure of other models to apply to Unionicola pinpointed at least five key biological characters that may form a better basis of comparison than taxonomic or habitat-based contrasts.

Guild structure in water mites (Unionicola spp.) inhabiting freshwater mussels: choice, competitive exclusion and sex.

Unionicolid water mites inhabit freshwater unionid mussels during the nymphal and adult stages of their life-cycle. Regular sampling of mussels from two sites in St. Mark's River, Fl. established that each of four species of water mite (Unionicola abnormipes, U. fossulata, U. serrata and U. formosa) occurred mainly in one or two of the mussel species available at each site.The role of preference for particular mussel species during host location was assessed for the first three mite species by choice experiments, in which mites were offered different mussel species simultaneously. In five out of six experiments, mites entered normally unused mussels as often as they did normally used ones. Additionally, a sexual difference in choice was found for U. fossulata, with males preferring one mussel species and females showing no preference. One mussel species, (Anodonta imbecilis), normally unused but chosen by mite species during the lab. experiments, is inhabited exclusively by the fourth mite species, U. formosa, in the field. An experiment showed that U. formosa excludes other mite species aggressively from Anodonta imbecilis.The results illustrate the sometimes misleading nature of simple sampling data as an indication of host specificity or host preference in parasites. They suggest also that the population dynamics of some parasites might be more fruitfully compared to unrelated, free-living species than to other parasites.

Recruitment of marine invertebrates: the role of active larval choices and early mortality.

Spatial variation in the recruitment of sessile marine invertebrates with planktonic larvae may be derived from a number of sources: events within the plankton, choices made by larvae at the time of settlement, and mortality of juvenile organisms after settlement, but before a census by an observer. These sources usually are not distinguished.A study of the recruitment of four species of sessile invertebrates living on rock walls beneath a kelp canopy showed that both selection of microhabitats by settling larvae and predation by fish may be important. Two microhabitats were of interest; open, flat rock surfaces, and small pits and crevices that act as refuges from fish predators.The polychaete Spirorbis eximus and the cyclostome bryozoan Tubulipora spp. showed no preference for refuges, but settled apparently at random on the available substrata. Tubulipora was preyed upon heavily by fish, while Spirorbis was relatively unaffected. The bryozoans Celleporaria brunnea and Scrupocellaria bertholetti both recruited preferentially into refuges. Scrupocellaria were preyed upon, while Celleporaria juveniles seemed unaffected. Predation by fish modified the spatial distribution (Tubulipora), abundance (Tubulipora), or size distribution (Scrupocellaria) of the juvenile population, or had relatively little effect (Celleporaria, Spirorbis).All of the above events occur within three weeks of settlement. Since inferences about the effect of larval events on the population dynamics of adult organisms are often based on observations of the patterns of recruitment after one or two months, they are therefore likely to be misleading.