Diet and annual production of two boreal river fishes following clearcut logging

Synopsis Diet and annual production of two cyprinid fishes in the Piscataquis River, Maine, were investigated before and after extensive deforestation of the surrounding watershed. Observed patterns were evaluated relative to post-logging changes in the benthic macroinvertebrate community, and to differences in ecomorphology between the two fishes. Prey selection was generally density-dependent for both predator species throughout the study, but significant exceptions to density-dependence were observed for several aquatic insect lifemode groups (e.g., swimmers, climbers, burrowers). This apparent preference/avoidance for certain prey taxa by fish predators could be adequately explained by differential availability (vulnerability) among prey taxa. Compared to the year before logging, significant post-logging changes in diet composition and gut fullness were observed for both blacknose dace,Rhinichthys atratulus, and creek chub,Semotilus atromaculatus, concomitant with post-logging changes in the availability of benthic prey. Annual production of blacknose dace was significantly lower (52.7 vs. 37.7 kg ha^–1 y^–1) in the year following logging. In contrast, annual production of creek chub increased significantly after logging, from 8.6 to 17.4 kg ha^–1 y^–1. Because post-logging changes in the physical environment were probably not sufficient to directly affect fish populations, we hypothesize that blacknose dace were morphologically constrained to feed on a declining aquatic prey resource, with a resultant decline in production. In contrast, the generalist creek chub switched to a non-aquatic prey resource (terrestrial arthropods) and no decline in production was observed for that species.     

Urbanization drives contemporary evolution in stream fish

Human activities reduce biodiversity but may also drive diversification by modifying selection. Urbanization alters stream hydrology by increasing peak water velocities, which should in turn alter selection on the body morphology of aquatic species. Here, we show how urbanization can generate evolutionary divergence in the body morphology of two species of stream fish, western blacknose dace (Rhinichthys obtusus) and creek chub (Semotilus atromaculatus). We predicted that fish should evolve more streamlined body shapes within urbanized streams. We found that in urban streams, dace consistently exhibited more streamlined bodies while chub consistently showed deeper bodies. Comparing modern creek chub populations with historical museum collections spanning 50 years, we found that creek chub (1) rapidly became deeper bodied in streams that experienced increasing urbanization over time, (2) had already achieved deepened bodies 50 years ago in streams that were then already urban (and showed no additional deepening over time), and (3) remained relatively shallow bodied in streams that stayed rural over time. By raising creek chub from five populations under common conditions in the laboratory, we found that morphological differences largely reflected genetically based differences, not velocity–induced phenotypic plasticity. We suggest that urbanization can drive rapid, adaptive evolutionary responses to disturbance, and that these responses may vary unpredictably in different species.     

Ontogenetic changes in diel activity,food habits and spatial distribution of juvenile and adult creek chub,Semotilus atromaculatus

Synopsis Juvenile creek chub, Semotilus atromaculatus, differed ecologically from adults. They were essentially diurnal, fed on small prey (adult diptera and aquatic adult coleoptera) and were found mainly in shallow littoral water (0.8 m depth) of a Québec oligotrophic lake. In contrast, adults were principally nocturnal, fed on larger prey (Gammaridae, diptera larvae, diptera pupae and ephemeroptera larvae) and were found in deeper water (1.2, 2.0 and 2.9 m depth). The overlap in diet between juvenile and adult fish was 10 percent. We found that predation by brook charr, Salvelinus fontinalis, upon creek chub was low, as only 5 chart stomachs of the 302 examined contained a total of 16 juvenile chub and none contained adult chub. Laboratory observations revealed that under low light intensity (0.17 lux) adult creek chub fed as efficiently as under high light intensity (22 lux). Several possible hypotheses to explain the evolution of the two different strategies exhibed by adult and juvenile creek chub are discussed.     

Repeat UCrit and endurance swimming in juvenile shortnose sturgeon (Acipenser brevirostrum)

Previous results show that juvenile shortnose sturgeon are steady swimmers and, compared with salmonids, generally have low critical swimming (UCrit) and endurance swimming capacities. Most studies on swimming capacities of sturgeon, and other fishes, include those where fish have only been swum once and the metrics of swimming performance are assessed (e.g., time swum, speed achieved). Under natural conditions, there are ample instances where fish undergo multiple swimming cycles when traversing fish ways, culverts and other sources of fast water flow. While some evidence exists for salmonids, the effects of repeat swimming are not well known for sturgeon. The current study consisted of two experiments. The first examined the UCrit of juvenile shortnose sturgeon following three consecutive swimming trials with a 30 min recovery period between subsequent tests. The second examined the endurance swimming capacities of juvenile shortnose sturgeon following three consecutive swimming trials with a 60 min recovery period between subsequent tests. Our findings indicate that (i) UCrit was consistent (~2 body lengths/s) among swimming trials; (ii) significant individual variation exists between individuals in the endurance swimming trials; and (iii) consistent results exist for individuals across swimming trials in both the UCrit and the endurance swimming tests. These results suggest that juvenile shortnose sturgeon have a high recovery capacity, and their behaviour and morphology likely reflect aspects of their swimming capacities.     

Deception Divergence and Sexual Selection

The hypothesis that blacknose dace Rhinichthys atratulus discriminate between piscivorous creek chub Semotilus atromaculatus and nonpiscivorous common suckers Catostomus commersoni was tested in a seminatural, artificial stream. Dace avoided the nonpredatory suckers less than they did the predatory chub, which they avoided during both day and night. However, dace significantly avoided the nonpredator at night. Further, experienced and inexperienced dace avoided the nonpredator similarly, suggesting that they had not learned to discriminate predator from nonpredator in their natural habitat. These results are consistent with two hypotheses (1) that individuals varied in ability to discriminate predatory from nonpredatory fishes, and (2) that the predator was more active than the nonpredator, thereby “frightening” the dace more. At temperatures of < 6°C the nonpredator was inactive and dace no longer avoided it at night, but continued to avoid the active chub, suggesting that movement constituted part of the avoidance-releasing stimulus. Because dace avoided nonpredators, they may waste time in natural streams avoiding harmless fish. It is suggested that the dace cannot afford the risk of a mistaken identity, which may be high in the stream environment, and selection favors those that avoid all large, moving fish.     

Seasonal movements and residency of small-bodied fish in a north temperate urban watershed demonstrate connectivity between a stream and stormwater drain

Despite their often physical connection, neighbouring stormwater drains and urban streams are typically managed quite differently; with drains mostly regarded as poor fish habitat. The goal of this study was to evaluate the interconnectedness of an urban stream (Watts Creek) and adjoining earthen surface stormwater drain (Kizell Drain) from the perspective of fish residency and movements over an entire year. Using a stationary passive integrated transponder (PIT) array, we quantified and compared the direction of movements among Watts Creek, Kizell Drain, and the area downstream of their confluence (herein termed Main) for four common stream fishes. We also determined the residency time (percentage of total time in days) within each of these reaches by combining data from the array and recaptured (with electrofishing and identified with hand-held PIT reader) or portably detected (with mobile PIT reader) fish. While the movements of creek chub (Semotilus atromaculatus) and central mudminnow (Umbra limi) varied across seasons, creek chub resided significantly longer in Watts, while central mudminnow spent more time in Kizell and Main. Longnose dace (Rhinichthys cataractae) moved into and resided most often within Watts. The movements and residency time for white sucker (Catostomus commersonii) did not vary among the reaches. We conclude there is a high degree of connectivity between Watts Creek and Kizell Drain and that, with the exception of longnose dace, the three other species utilize the habitat available in Kizell. This study demonstrates the biological potential of earthen stormwater drains and as a result we recommend these systems be managed as a functional component of urban watersheds.     

Complex physiological traits as biomarkers of the sub-lethal toxicological effects of pollutant exposure in fishes

Complex physiological traits, such as routine aerobic metabolic rate or exercise performance, are indicators of the functional integrity of fish that can reveal sub-lethal toxicological effects of aquatic pollutants. These traits have proved valuable in laboratory investigations of the sub-lethal effects of heavy metals, ammonia and various xenobiotics. It is not known, however, whether they can also function as biomarkers of the complex potential range of effects upon overall functional integrity caused by exposure to mixtures of chemicals in polluted natural environments. The current study used portable swimming respirometers to compare exercise performance and respiratory metabolism of fish exposed in cages for three weeks to either clean or polluted sites on three urban European river systems: the river Lambro, Milan, Italy; the rivers Blythe, Cole and Tame, Birmingham, UK; and the river Amstel, Amsterdam, The Netherlands. The UK and Italian rivers were variously polluted with high levels of both bioavailable heavy metals and organics, and the Amstel by mixtures of bioavailable organics at high concentrations. In both the UK and Italy, indigenous chub (Leuciscus cephalus) exposed to clean or polluted sites swam equally well in an initial performance test, but the chub from polluted sites could not repeat this performance after a brief recovery interval. These animals were unable to raise the metabolic rate and allocate oxygen towards exercise in the second trial, an effect confirmed in successive campaigns in Italy. Swimming performance was therefore a biomarker indicator of pollutant exposure in chub exposed at these sites. Exposure to polluted sites on the river Amstel did not affect the repeat swimming performance of cultured cloned carp (Cyprinus carpio), indicating either a species-specific tolerance or relative absence of heavy metals. However, measurements of oxygen uptake during swimming revealed increased rates of routine aerobic metabolism in both chub and carp at polluted sites in all of the rivers studied, indicating a sub-lethal metabolic loading effect. Therefore, the physiological traits of exercise performance and metabolic rate have potential as biomarkers of the overall sub-lethal toxic effects of exposure to complex mixtures of pollutants in rivers, and may also provide insight into why fish do not colonize some polluted environments.     

Shifting thermal regimes influence competitive feeding and aggression dynamics of brook trout (Salvelinus fontinalis) and creek chub (Semotilus atromaculatus)

The natural distributions of freshwater fish species are limited by their thermal tolerances via physiological constraints and increased interspecific competition as temperatures shift toward the thermal optima of other syntopic species. Species may mediate stress from temperature change physiologically, behaviorally, or both; but these changes may compromise competitive advantages through effects on feeding and social behavior. In the Appalachian Mountains of North America, creek chub (Semotilus atromaculatus) are found in warm‐water and cold‐water streams and overlap in range with brook trout (Salvelinus fontinalis) across lower thermal maxima, where they compete for food and space. As stream temperatures continue to increase due to climate change, brook trout are under increasing thermal stress which may negatively affect their ability to compete with creek chub. To examine the influences of temperature on competitive interactions between these species, we observed feeding behavior, aggression, and habitat use differences at three temperatures approaching brook trout thermal maxima (18°C, 20°C, and 22°C) among dyad pairs for all combinations of species in experimental flow‐through tanks. We also examined feeding and habitat use of both species under solitary conditions. We found as temperature increased, feeding and aggression of brook trout were significantly reduced in the presence of creek chub. Creek chub pairs were more likely to occupy benthic areas and refugia while brook trout pairs used surface water more. Space use patterns significantly changed by pairing treatment. Aggression and space use shifts allowed increased exploitative and interference competition from creek chub when paired with brook trout that was not present in conspecific pairs. The decreased dominance of a top predator may lead to diverse impacts on stream community dynamics with implications for the future range restriction of brook trout and demonstrate possible mechanisms to facilitate competitive advantages of warm water generalist species under thermal stress.     

Nontroglobitic Fishes in Bruffey-Hills Creek Cave, West Virginia, and other Caves Worldwide

Six species of fishes were tagged and released in September and November 1995 and on five dates between November 1998 and October 1999 inside Bruffey-Hills Creek Cave. Most of the tagged fishes were creek chub, Semotilus atromaculatus, and green sunfish, Lepomis cyanellus. The overall recapture rate was 2.6% as only three of 117 fishes were recaptured. Forty-nine days was the longest time of residence by two L. cyanellus and one S. atromaculatus, and movement of 83.4 m was observed only for the creek chub. Tag loss was confirmed as one factor in the low recapture rate. Nine species of fishes were collected in the cave, including Phoxinus oreas and Pimephales promelas, two species never collected in a cave. Most of the fishes were pigmented normally, but many creek chubs were depigmented and appeared white or pigmentless when first observed in the cave stream. Four species, Phoxinus oreas, Pimephales promelas, Ameiurus nebulosus, and Lepomis macrochirus, were collected for the first time in Bruffey-Hills Creek Cave and in the Bruffey and Hills creeks drainage. Although data from this study did not shed light on residence time of fishes in the cave, the consistent occurrence of epigean fishes in this cave was shown. A list of epigean fishes from caves worldwide was included along with a discussion of aspects of the ecology of epigean fishes in caves and of evolution of troglobitic fishes.     

Diel foraging in relation to available prey in an Adirondack Mountain stream fish community

The diets of the fish community of Trucka Brook, a small stream located in the central Adirondack Mountains in northern New York, were examined in relation to the bottom fauna and invertebrate drift. Measures of overlap were calculated between the diets of each fish species examined, brook trout (Salvelinus fontinalis), blacknose dace (Rhinichthys atratulus), creek chub (Semotilus atromaculatus) and pearl dace (Semotilus margarita). Overlap was also examined between the fish diets and bottom and drift samples. Blacknose dace, pearl dace and brook trout had the most similar diets which were closely associated with the benthos. Creek chub had the most distinctive diets which did not compare well with any other fish species during either diurnal or nocturnal periods. The mayfly nymph Litobranchia recurvata was the most abundant bottom invertebrate and was the major prey of benthic feeding fishes. The invertebrate drift did not compare favorably with any of the fishes' diets because of the predominance of large cased limnephilid larvae (primarily Psychoglypha sp.) which were not readily consumed by fish.     

Colonization of abandoned swimming pools by larval mosquitoes and their predators following Hurricane Katrina

Thousands of flooded swimming pools were abandoned in New Orleans following Hurricane Katrina and provided a natural experiment to examine colonization of a novel aquatic habitat by mosquito larvae and their aquatic predators. We conducted a randomized survey of flooded swimming pools in two neighborhoods in January 2006 and found that 64% contained mosquito larvae, 92% contained predatory invertebrates, and 47% contained fishes. We collected 12,379 immature mosquitoes representing five species, primarily Culiseta inornata, and secondarily, the arboviral vector Culex quinquefasciatus. Dragonfly nymphs in the families Aeshnidae and Libellulidae were the most common predatory invertebrates collected among a total of 32 non-mosquito invertebrate species. Eleven species of fishes were collected, with Gambusia affinis accounting for 76% of the catch. Diversity of fishes in swimming pools was positively correlated with proximity to a levee breach and the fish assemblage found in swimming pools was similar to that found along shorelines of Lake Pontchartrain and drainage canals that flooded the study area. Mosquito larvae were rare or absent from pools containing fishes; however, path analysis indicated that the presence of top predators or abundant competitors may somewhat mitigate the effect of Gambusia affinis on mosquito presence.     

Sound the alarm: A meta‐analysis on the effect of aquatic noise on fish behavior and physiology

The aquatic environment is increasingly bombarded by a wide variety of noise pollutants whose range and intensity are increasing with each passing decade. Yet, little is known about how aquatic noise affects marine communities. To determine the implications that changes to the soundscape may have on fishes, a meta‐analysis was conducted focusing on the ramifications of noise on fish behavior and physiology. Our meta‐analysis identified 42 studies that produced 2,354 data points, which in turn indicated that anthropogenic noise negatively affects fish behavior and physiology. The most predominate responses occurred within foraging ability, predation risk, and reproductive success. Additionally, anthropogenic noise was shown to increase the hearing thresholds and cortisol levels of numerous species while tones, biological, and environmental noise were most likely to affect complex movements and swimming abilities. These findings suggest that the majority of fish species are sensitive to changes in the aquatic soundscape, and depending on the noise source, species responses may have extreme and negative fitness consequences. As such, this global synthesis should serve as a warning of the potentially dire consequences facing marine ecosystems if alterations to aquatic soundscapes continue on their current trajectory.     

Age structure and growth of Semotilus atromaculatus(Mitchill) in PCB‐contaminated streams

Creek chub Semotilus atromaculatus from two PCB contaminated streams (Clear Creek and Richland Creek) at three locations and a reference stream (Little Indian Creek), Indiana, U.S.A., were examined to determine if age class structure and growth variables were correlated with in‐situ PCB exposure. Approximately five to 15 fish were captured weekly during the spring spawning season and monthly thereafter for a 12 month period. Fish collected ranged from 25 to 267 mm total length (L_T). Throughout the course of this study, no spawning activity was observed at either location in Clear Creek, although some very small young‐of‐the‐year (YOY) creek chub fry were observed at the downstream location by late summer. Creek chub nests were observed in both Richland Creek and Little Indian Creek but YOY were common only in Little Indian Creek. Exposure to PCBs was shown to both enhance and decrease growth in varied laboratory tests; subtle but significant gender‐specific differences in the growth of creek chub populations between the sites were observed. Creek chub up to 24 months in age from Clear Creek and Richland Creek were significantly larger (both L_T and mass for males and L_T for females) than reference site creek chub. This trend was reversed for creek chub aged ≥24 months as the reference site fish were consistently larger with reference males weighing significantly more. Older age classes of creek chub were missing in areas of higher PCB contamination. Female population growth rates and individual instantaneous growth rates were consistently higher at the reference site in comparison to the PCB‐contaminated sites. Calculation of ‘functional b’(as a condition factor) did indicate that growth enhancement in young males did occur at the most contaminated site and reductions in growth (mass relative to L_T) occurred in females from all contaminated sites. Furthermore, long‐term survivorship for females was reduced in the PCB‐contaminated streams. All of these subtle alterations in growth would not have been observed if males and females had not been analysed separately.     

Competition for food between an introduced crayfish and two fishes endemic to the Colorado River basin

Synopsis Crayfish are not native to the Colorado River basin (CRB), however they are now established in portions of the mainstem and in many tributaries. I used density manipulation experiments in a laboratory setting to determine intra- and interspecific competition for food between Orconectes virilis, an aggressive polytrophic crayfish now common in the CRB, and two native fishes: Gila chub, Gila intermedia, and flannelmouth sucker, Catostomus latipinnis. I tested each fish species in separate trials. Growth of Gila chub decreased when animal densities increased, however they were more affected by intraspecific competition than by crayfish presence. In contrast, growth of flannelmouth suckers was more affected by crayfish than by intraspecific competition. Crayfish growth was not significantly altered by presence of either fish. Crayfish thus reduced fish growth by competition for food, but the effect differed markedly between the two species. An erratum to this article can be found at .     

Fish species associations in riffle habitat of streams of varying size and acidity in New Brunswick and Nova Scotia

Factors related to stream size and alkalinity were most influential in determining fish species associations in three catchments of New Brunswick (NB) and Nova Scotia (NS) Canada, as determined by discriminant function analysis of TWINSPAN stream site classification.
In the circumneutral Saint Croix (NB) catchment, the creek chub, Semotilus atromaculatus (Mitchill), brook trout, Salvelinus fontinalis (Mitchill), blacknose dace, Rhinichthys atratulus (Hermann), Atlantic salmon, Salmo salar Linnaeus, eel, Anguilla rostrata (Lesueur) and fallfish. Semotilus corporalis (Mitchill) dominated stream sites of progressively greater discharge and higher median mid-summer temperature. In the acidic Gold (NS) system, the brook trout, Atlantic salmon, and eel exhibited a distribution pattern in relation to stream size and temperature similar to that for the Saint Croix, with the eel relatively more abundant in the Gold at large stream sites. The creek chub was excluded from the smallest tributaries by low pH. The ranges of blacknose dace and fallfish do not extend to southwestern Nova Scotia. In the Medway (NS) system (slightly more acidic than the Gold), the relative abundance of Atlantic salmon is reduced, and that of eel increased as compared with the Gold and Saint Croix systems. The lower limiting mid-summer pH levels for creek chub, salmon, brook trout, and eel are 5.2, 5.0, 4.7 and <4.5 respectively.     

Assessment of errors associated with harbour seal (Phoca vitulina) faecal sampling

Six harbour seals, ages 4–8 years, were held as pairs in a 10 times 20 times 2 m tank filled with sea water, and on 60 occasions were fed a meal of a specific species of fish or cephalopod of known size. The tank was drained periodically, and harbour seal faeces were collected on a 0.5 mm sieve. Number and size of otoliths and beaks found in faeces were determined. Fifty-eight percent of 670 fish and 37% of 36 cephalopods fed to harbour seals were represented by their otoliths or beaks in faeces. Estimated number of prey consumed was determined from the greatest number of left or right otoliths or upper or lower beaks collected in faeces. Estimated length ofprey was determined from measurements of otoliths and beaks recovered in the tank and relationships of otolith and beak measurements to prey length. Estimated number of fish eaten was not significantly different among pairs of harbour seals, but was different among species of fishes. Only 24–35% of fish species with small otoliths were represented in faeces, whereas more robust otoliths from other species were less apt to be completely dissolved. Estimated length of fishes was significantly less than lengths of fishes fed to harbour seals in 39 (76.5%) of 51 trials. Cephalopod beaks were not affected by passage through the harbour seal digestive tract. Amount of otolith dissolution was not related to species of fish; estimated fish length was underestimated by an average 27.5%. Although some (7.4%) of the otoliths were collected within 100 h after the fish were ingested, more than 90% were recovered within 24 h after the fish was eaten. Correction factors were developed which will allow researchers to estimate more reliably number and size of fish and cephalopod prey eaten by harbour seals.     

The Ichthyofauna Associated with an Intertidal Creek and Adjacent Eelgrass Beds in the Kariega Estuary,South Africa

The structure and functioning of salt marsh fish communities in the overall ecology of southern African estuaries is poorly understood. This study compares the ichthyofauna associated with a salt marsh creek and eelgrass bed in an attempt to evaluate the relative importance of these habitats to fishes. Taylor's salt marsh creek and adjacent eelgrass bed in the Kariega Estuary were sampled twice per season between the winter of 1992 and the autumn of 1994. The average density and standing stock of fishes were found to be considerably higher in the eelgrass bed than in the intertidal creek. Both habitats had similar fish diversities but were dominated by different taxa, the most notable of which was the dominance of mugilids in the creek and their scarcity in the eelgrass. Taylor's intertidal creek and adjacent eelgrass beds were dominated by juvenile fish, with both habitats functioning as nursery areas for juvenile fish, albeit for totally different ichthyofaunal communities. The similar fish diversities but lower abundances in the intertidal creek compared to the eelgrass beds are in contrast to similar North American studies, and refute the hypothesis that intertidal salt marsh creeks have higher fish densities but lower diversities than eelgrass beds.The first author is also the senior author     

Group Choice as a Function of Group Size Differences and Assessment Time in Fish: The Influence of Species Vulnerability to Predation

The shoal-choice behaviour of two species of fish that differ in their vulnerability to predation was compared. Individuals of threespine stickleback, Gasterosteus aculeatus, and creek chub, Semotilus atromaculatus, were presented with a simultaneous choice of two equidistant stimulus shoals of conspecifics that differed in membership size (5 vs. 6 fish, 5 vs. 7, 5 vs. 8, 5 vs. 9 and 5 vs. 10). Test fish were allowed to view the stimulus shoals from a standard distance for either 10–20 or 120–150s before being frightened with a stimulus from an overhead light and released to join either shoal. We observed which shoal (the smaller or the larger one) the test fish approached. Preference for the larger stimulus shoal generally increased with increasing shoal size difference and with the duration of the assessment period, and was more pronounced in chub (the more vulnerable of the two species). For the short assessment period, chub showed a significantly stronger preference for the larger stimulus shoal than sticklebacks, whereas there was no significant difference between species for the long assessment period. Furthermore, chub responded more readily to small differences in shoal size (of 1–3 fish) than sticklebacks, for both short and long assessment periods. The above results are consistent with the hypothesis that chub, as the more vulnerable of the two species (in terms of predation), should be able to identify the larger of two shoals more quickly and should be more sensitive to small differences in shoal size than sticklebacks.     

Quantitative sampling of fishes in a simple system: failure of conventional methods

In a capture-mark-recapture fish density estimate carried out in a small freshwater pond in southern Ontario, Canada, a lift net and minnow traps were used as catching gear. Both worked well for initial captures, but most recaptures were in the traps. Density estimates within reasonable confidence limits were possible for only two species, creek chub (Semotilus atromaculatus) and common shiner (Notropis cornutus). The other 13 species present were either missed entirely or not captured or recaptured sufficiently so estimates for them were inadequate. Subsequent collection of fishes after rotenone poisoning verified the estimates for the two species, if the size selectivity of the gear and the possible effects of marking and limitations of collecting after poisoning were considered.     

Intraguild indirect effects through trophic cascades between stream-dwelling fishes

1. To examine the strength and role of indirect effects through trophic cascades by omnivorous fish on algal biomass in streams, we introduced one of four fish species (ayu Plecoglossus altivelis altivelis, pike gudgeon Pseudogobio esocinus esocinus, Japanese dace Tribolodon hakonensis and pale chub Zacco platypus) in experimental pools. We also investigated the indirect effects of gudgeon, dace and chub on the growth of ayu. 2. We chose the four fish species across a continuum of omnivory. Ayu fed mainly on benthic algae, and gudgeon predominantly on invertebrates. Dace and chub fed on both algae and invertebrates, but dace preyed on invertebrates more than chub. 3. The presence of gudgeon, dace and chub reduced the number of algal-feeding invertebrates and increased the algal biomass through trophic cascades. Consequently, ayu growth rate over the experimental period in pools with one of the three fish species was 25.9-44.1 times greater than the growth rate in pools with only ayu. 4. The positive indirect effect on ayu growth was large for gudgeon and dace and small for chub, whereas the addition of ayu reduced ayu growth considerably due to intraspecific competition. 5. The relative intensity of indirect effects on ayu growth through trophic cascades was predictable from the food overlap between ayu and the other fishes. However, the similar strength of indirect effects by gudgeon and dace that fed differently on algae and invertebrates suggests that feeding behaviour, prey preference and trait-mediated indirect interactions were also important in the prediction.