Group swimming significantly decreases time to passage success for a galaxiid species

Abstract

We conducted laboratory experiments to explore the potential benefits of group behaviour on passage performance for a small‐bodied migratory species, inanga Galaxias maculatus. An artificial velocity barrier was created to compare the fish passage success of groups of fish to solitary individuals. Passage success was measured using several metrics, including binomial success models and time‐to‐event analysis. Active metabolic rates were also measured as indices of energetic expenditure. Our findings revealed that fish swimming in groups have faster entry and passage rates compared to solitary individuals, but there was no difference in the proportion of fish successfully passing. Successful fish in groups displayed lower metabolic rates, suggesting the potential benefit of reduced energy expenditure for groups of fish. While group swimming did not enhance overall passage success, it significantly reduced the time required for successful passage compared to solitary swimming individuals. These findings underscore the importance of designing fish passes that accommodate gregarious species, ensuring improved success for fish populations overall. By considering the advantages of group behaviour on passage performance, fish passage structures can be tailored to better facilitate the movement of diverse fish species through aquatic environments.

1. INTRODUCTION

Globally, millions of instream structures create fragmentation in river networks, acting as impediments to fish movements (Belletti et al., 2020; Grill et al., 2015; Franklin et al., 2022). These impediments to access for fishes have major ecological consequences, drastically changing fish populations, causing population declines and even extirpation (Lucas & Baras, 2001; Radinger & Wolter, 2014; Silva et al., 2018; Warren & Pardew, 1998). Consequently, it is common to see retroactive additions of fish passage structures or remediation efforts aimed at increasing the upstream migration of fishes (Birnie‐Gauvin et al., 2019; Katopodis & Williams, 2012; Silva et al., 2018). Meta‐analyses have continued to show highly variable and often poor passage efficiency (e.g. Bunt et al., 2012; Hershey, 2021; Roscoe & Hinch, 2010). To improve the passage efficiency of fishways, we need to consider both the environmental and behavioural factors influencing the swimming performance of fishes as they swim upstream (Roscoe & Hinch, 2010).

Swimming in groups (shoaling or schooling), has been shown to provide several benefits to fish, including hydrodynamic advantages, navigational aids and physiological effects, that may have important implications for improving performance through fishways (Johansen et al., 2010; Marras et al., 2015; Okasaki et al., 2020; Ward et al., 2011; Weihs, 1973). Schooling fish gain hydrodynamic advantages, which may help them overcome velocity challenges created by instream barriers. Some schooling fish experience reduced drag and increased lift, potentially reducing overall metabolic demand (Marras et al., 2015; Nadler et al., 2016; Parker, 1973). There are three potential hydrodynamic benefits to explain this: the vortex hypothesis, the channelling hypothesis and the wingtip‐uplift hypothesis (Abrahams & Colgan, 1987; Weihs, 1973). The vortex hypothesis states that the vortices within the group reduce velocity relative to the individual. This is supported by more recent research showing that fish actively alter their body mechanics when swimming in vortices created by other members of the school, which reduces the energy required to swim at a given speed (Liao, 2003; Lindsey, 1978; Marras et al., 2015). The channelling hypothesis states that neighbouring fish in a group improve the thrust efficiency of an individual (Lemasson et al., 2014; Weihs, 1973). The wingtip‐uplift hypothesis states that neighbouring wakes provide lift to an individual (Weihs, 1973). These methods of decreasing metabolic demands in schooling fish could increase their ability to overcome speed constraints.

Group size or local density may also influence the motivation of individual fishes to swim through an instream structure. When local density is high, some fishes may be less motivated to swim upstream (Dominy, 1973; Goerig & Castro‐Santos, 2017). This is particularly problematic when large‐bodied fishes like salmon are overcrowded and less likely to migrate through an instream structure (Dominy, 1973; Goerig & Castro‐Santos, 2017; Johnson et al., 2012; Lemasson et al., 2014). Many large‐bodied species can have strong social hierarchy, with larger individuals occupying dominant positions that make them less likely to swim upstream as density increases (Goerig & Castro‐Santos, 2017). Yet the role of social hierarchy on motivating the movement of smaller or habitually shoaling fishes remains unknown. Migratory juvenile Galaxiids, such as inanga Galaxias maculatus (Jenyns 1842), are relatively similar in size and shoal in the thousands, which likely reduces the effect of social hierarchy (McDowall et al., 1994).

In terms of navigational benefits, group travel during migration can have major benefits for a fish's ability to move to its intended destination by reducing directional errors (Larkin & Walten, 1969). Social interactions and group cohesion significantly enhance each individual's ability to align toward and reach a target or direction, particularly in environments with low levels of environmental turbulence (Codling et al., 2007). This may be due in part to individuals improving accuracy by combining their individual estimates, referred to as the “Many Wrongs” hypothesis (Larkin & Walten, 1969). Other features of collective navigation include emergent sensing, leadership, social learning and collective learning (Berdahl et al., 2018). Schooling behaviours improve each individual's fitness and survival (Johansen et al., 2010; Marras et al., 2015; Nadler et al., 2016; Weihs, 1973). Schooling reduces predation risk and increases foraging and migration success (Abrahams & Colgan, 1987). Physiologically, swimming in groups can have “calming effects” on fish, reducing stress levels and potentially improving overall performance (Lefrançois et al., 2009; Nadler et al., 2016; Parker, 1973). The presence of conspecifics may provide a sense of safety and familiarity because it both lowers stress hormones as well as increasing regulation of antioxidant genes that mitigate stress (Gilmour & Bard, 2022; Schumann et al., 2023). This effect is known as “social buffering” (Gilmour & Bard, 2022; Schumann et al., 2023). These findings, taken together, suggest that fish living in groups experience reduced stress, with overall lower metabolic demand and reduced energy expenditure compared to solitary individuals (Lefrançois et al., 2009; Nadler et al., 2016; Parker, 1973). However, the benefits of shoaling fish compared to solitary individuals are at risk because anthropogenic disturbances that force typically shoaling fish into social isolation could cause negative physiological consequences (Nadler et al., 2016). Thus, poorly designed fish passes may be not just ecologically, but physiologically detrimental to the fishes that they block (Jones et al., 2021).

The failure to comprehend the importance of fish behaviour and, consequently, the variability in fish swimming performance, poses a challenge in designing effective fish passage solutions. Collective navigation can have a major effect on passage through dams, with lasting implications for survival, reproductive success and population stability (Okasaki et al., 2020). However, current practices often neglect the nuances of social interactions in fish passage design. Experimental data collected from studying individual fish swimming in isolation raise questions about the transferability of such findings to real‐world conditions. In natural environments, the behaviour and physiology of groups of fish may differ significantly from that of solitary fish.

A recent review by Mozzi et al. (2023) emphasized the importance of incorporating group behaviour into fish passage research, design, management and monitoring. While extensive research has demonstrated the benefits of schooling—such as reduced metabolic costs, lower stress levels and improved navigation—Mozzi et al. (2023) highlighted the need to apply this research to fish passage contexts. Ineffective fish passage structures can fragment fish groups, potentially compounding the challenges faced by solitary individuals traversing these barriers. This fragmentation not only increases the physiological and navigational stress on isolated fish but also raises the risk of mortality. The need for understanding group behaviour and fish passage structures is especially true for gregarious amphidromous and catadromous species (Coutant & Whitney, 2000). Mozzi et al. (2023) suggested that laboratory experiments could provide valuable insights into how group behaviour influences passage performance. They emphasized the importance of understanding how flow conditions, velocity and turbulence impact group structure, cohesion, behaviour and, ultimately, passage success. Supporting this, some researchers have identified links between collective movement and fish passage performance in laboratory settings (e.g. Albayrak et al., 2020; Nyqvist et al., 2024). For example, Nyqvist et al. (2024) found that groups of five fish passed an experimental weir at significantly higher rates than solitary individuals, underscoring the importance of conspecifics for success at fish passage structures.

Island freshwater fish communities have a high prevalence of obligatory amphidromy and in New Zealand many migratory species are both amphidromous and small‐bodied (Franklin & Gee, 2019). This life history and size constraint poses additional challenges to migration through anthropogenic structures. In this study we tested whether group swimming improves the ability of fish to traverse an artificial barrier. We used Galaxias maculatus as an experimental species as they have a wide distribution throughout the temperate Southern Hemisphere and are an amphidromous species often used as a benchmark in New Zealand for evaluating fish passage barriers (Crawford et al., 2025; Franklin et al., 2024). Additionally, G. maculatus display gregarious behaviour, shoaling in the thousands during their upstream migration as small‐bodied juveniles (McDowall et al., 1994). We hypothesised that groups of fishes would be more successful at traversing a barrier than individual fishes.

2. METHODS

The fish collection for this research was carried out under the National Institute of Water and Atmospheric Research (NIWA) MPI special permit SP666‐4. All experimental trials for this study were run with approval to manipulate live animals for the purpose of research by the NIWA Animal Ethics Committee (AEC204) in accordance with the requirements of section 83 of the New Zealand Animal Welfare Act 1999.

2.1. Fish collection and holding

Juvenile Galaxias maculatus (inanga) were collected using fyke nets and gee minnow traps from Parker Stream, New Zealand (37°36′10.6″S, 175°08′44.5″E). Fish were transported to the NIWA Hamilton fish laboratory in coolers filled with aerated water from where they were collected. Fish were kept in the coolers overnight until the water temperature equilibrated with the laboratory tank water temperature and were then transferred to two 60‐L quarantine tanks with 6 ppt salinity to prevent disease, with ~150 fish kept in each tank. After 1 week, fish were transferred to five 60‐L freshwater tanks, with ~50 fish per tank. Fish were held in a temperature‐controlled room and kept on a 12‐h light, 12‐h dark cycle. The tanks were kept on a recirculating water system and the temperature held at 18 ± 0.5°C. Fish were fed bloodworms every other day and were fasted for 24 h prior to experimentation to ensure a postabsorptive state. Ammonia and pH concentrations were checked regularly, and water changes were completed when ammonia levels were greater than or equal to 0.25 mg L⁻¹.

2.2. Experimental setup

Experiments were conducted using an indoor recirculating flume as previously detailed in Miller et al. (2002). The flume was constructed from acrylic and measured 7.5 m in length, 0.5 m in width and 0.5 m in depth, holding up to 1500 L of water in the working section of the flume, with a return pipe of 0.4‐m diameter running beneath it. A flow straightener consisting of stacked PVC tubes measuring 0.02 m in diameter and 0.3 m long was placed at the upstream end of the flume. The water speed was controlled by an impeller located in the descending section of the return pipe. This impeller was driven by a variable‐speed AC motor. Spot water temperature measurements were taken at the beginning and end of each trial using a handheld thermometer. Temperature ranged from 17.9 to 22.2°C with a mean of 21.1°C (standard deviation 0.96) across all trials.

A raceway was installed to constrict the flow within the flume to create a water speed challenge for the fish. The raceway was designed so that the maximum water speed in the constricted section ranged between 0.45 and 0.50 m s⁻¹. This was based on the findings of Crawford et al. (2025), which showed that only the top 10% of G. maculatus individuals achieved critical swimming speeds of greater than 9.5 bl s⁻¹ (0.45 m s⁻¹). The average Reynolds number in this constricted section was 220,588, indicating that the fish were exposed to turbulent flow conditions.

The test section was delineated from 0 to 5.9 m in the working section of the flume. The raceway was constructed of PVC (Figure 1) and consisted of two sets of 1.2‐m long angled‐wingwalls on the upstream and downstream ends of the raceway, set at a 40° angle relative to the flume wall. The body of the raceway was 2.9 m long and 0.2 m wide. The depth of water in the raceway was maintained at 0.2 m throughout all trials. Mesh screens were placed upstream and downstream of the raceway to keep the fish inside the test area. The base of the raceway was marked with tape every 0.1 m starting from 0.0 at the entrance to the downstream wingwalls. The raceway had additional markings denoting the approach, entry and passage thresholds in the raceway (0.6, 1.8 and 4.7 m, respectively), marked from the end of the downstream working section within the flume (Figure 1).

FIGURE 1.

Aerial view experimental setup, focusing on the working section containing the raceway. (a) Line marking the “approach” to the raceway, (b) line marking the “entry” to the raceway, (c) line marking successful “passage” of the fishway, (d) mesh screens (visible as dark bars) to confine fish within of the flume and (e) removeable mesh screen, containing fish in the “starting pool” for acclimation before the start of the trial.

For analysis of passage, the raceway was delineated at three points (Figure 1):

• approach threshold, marking the point where fish crossed the threshold into the approach of the raceway (Figure 1a)

• entry threshold, marking the point of entry into the raceway (Figure 1b)

• passage threshold, marking the point where fish pass through the upstream end of the raceway (Figure 1c).

To characterize the variation in water speed throughout the raceway, a Sontek Flow Tracker 2 Acoustic Doppler Velocimeter (ADV) was used. Three‐point vertical profiles of water speed were made at 0.5‐m intervals along the centre of the channel, along the axis of the flume in the direction of flow. Measurements were made at heights of 20%, 60% and 80% of the water column above the raceway floor, starting at 0.2 m from the downstream end of the raceway. At each height, speed measurements were collected for 40 s at a sampling frequency of 10.0 MHz, measured in m s⁻¹. After sampling, data were downloaded in FlowTracker2 desktop software, where water speed was averaged across each depth at each station to map out the water speeds experienced by the fish (Figure 2).

FIGURE 2.

Acoustic Doppler velocimeter measurements throughout the raceway taken at 20%, 60% and 80% depth of the water column. The grey dashed lines labelled a, b and c correspond to the approach, entry and passage lines depicted in Figure 1.

Five Swan security cameras were set up 1.1 m above the flume floor, evenly distributed across the length of working section, creating a slight overlap between camera frames. The video recording of every camera was synchronized throughout the duration of the trial to track fish movement. Videos were recorded in colour and all cameras were set at 30 frames per second.

2.3. Experimental protocol

Due to fish availability and timing of migration, a total of 37 group (five fish per group) trials and 58 solitary (single fish) trials were conducted (Tables 1 and S1–S4). Before being placed in the flume, the body length and weight of all fish were measured (Tables 1 and S1–S4). The condition factor of the fish was calculated using Fulton's condition factor (Froese, 2006; Table 1). The fish were placed in the downstream pool before the start of the trial, with a mesh screen preventing them from entering the raceway (Figure 1). The fish were allowed to acclimate in the pool for 30 min at a water speed of 0.0 m s⁻¹ to reduce the effects of handling stress.

TABLE 1.

Data for group vs. solitary trials, including the number of replicates tested in each category, mean total lengths, mean total weight, mean Fulton's condition factor, water temperature and standard errors.

Category	Number of replicates	Mean length (mm)	Mean weight (g)	Mean condition factor (k)	Mean water temperature (°C)
Group	37	57.83 ± 4.05	0.81 ± 0.22	0.41 ± 0.05	21.45 ± 0.52
Solitary	58	54.36 ± 7.62	0.62 ± 0.38	0.34 ± 0.10	20.14 ± 1.34

Note: Further detailed information can be found in Tables S1–S4.

After acclimation, the cameras were turned on to record fish movement for the 30‐min trial. The impeller motor was set to 31 Hz, resulting in a water speed between 0.45 and 0.5 m s⁻¹ in the working section of the raceway (Figure 2). Once the speed in the flume was set, the downstream mesh screen was removed. During the trial, the times that fish approached the raceway, entered the raceway and finished the raceway were recorded. For trials using groups of fish, the time of each event was collected for every individual. The times at which the fish crossed the approach, entry and passage thresholds were recorded from the time‐stamped video recordings to the nearest second. Only the first attempt for each fish was recorded to prevent duplicating data, but, in general, fish did not stage multiple attempts for each threshold. For fish swimming in groups, we individually tracked and recorded the behaviour of each fish. To minimize the effect of holding time on swimming performance, a stratified trial design was used where solitary and group tests were conducted each trial day.

2.4. Active metabolic rate

At the end of each trial, fish that successfully traversed the raceway were placed into a closed circuit respirometry unit to measure the post‐exercise active metabolic rate, following the methods of Parisi et al. (2020). For each group trial, all successful fish (i.e. all the fish that had made it to the upstream end of the flume) from that group were placed in a single respirometer together. Total fish weight was recorded prior to the respirometry experiment. The respirometer consisted of a 115 mL plastic container connected to a recirculating flow system. The respirometer was placed in a water bath held at ambient trial temperature. A Loligo System oxygen sensor was connected to the respirometer, measuring oxygen concentration (% air saturation) at 1‐min intervals for a period of 15 min. The metabolic rate (MR) of fish was calculated using the following equation from Schurmann and Steffensen (1997):

$$\mathsf{MR}=-\mathsf{1}\times {\mathsf{\Delta O}}_{\mathsf{2}}\times \mathsf{V}\mathsf{x}{{\mathsf{M}}_{\mathsf{\text{fish}}}}^{-\mathsf{1}}\times {\mathsf{\beta O}}_{\mathsf{2}}$$

where ΔO₂ is the rate of change of oxygen (% air saturation per hour), V is the volume of the respirometer in litres, without the mass of the fish (assuming a density of 1 g mL⁻¹), M _fish is the mass of the fish in grams and βO₂ represents the solubility of oxygen in water (mg L⁻¹) at the ambient trial temperature (ranging from 9.47 mg L⁻¹ at 18°C to 8.11 mg L⁻¹ at 22°C).

3. ANALYSIS

3.1. Binary success model

Success was modelled with binary logistic regression, where a fish was given a 1 for successfully crossing a threshold or a 0 for failing to cross a threshold. To model the effect swimming in groups had on the success of fish passing each threshold (approach, entry, passage), we compared the success of groups to individuals, with the response variable defined as the proportion of fish in each replicate that were successful (i.e. groups 0, 0.2, 0.4, 0.6, 0.8 or 1, solitary 0 or 1). A binomial generalized linear mixed‐effects model was used, with proportion successful as the response variable and trial number as a random predictor variable. Fish length and weight were not included in the model due to our inability to uniquely identify these characteristics for individual fish within groups in the flume. Water temperature was added as a covariate but was found not to be statistically significant and was therefore removed from the model (p > 0.05). For this model, a logit link function was selected without adjusting for overdispersion. A separate model was applied to the approach, entry and finish thresholds of the raceway. Analysis was carried out using the glmmTMB package in R (Brooks et al., 2017). A Hosmer Lemeshow test was used to assess model goodness of fit using the Resource Selection package (Lele et al., 2023). All analyses were carried out in R version 4.2.0 (R Core Team, 2020).

3.2. Time‐to‐event analysis

We used time‐to‐event analysis, also known as survival analysis, to assess the rate of fish attempting the various stages within the raceway (approach rate, entry rate and passage rate) (Castro‐Santos & Haro, 2003; Goerig & Castro‐Santos, 2017). Time‐to‐event analysis enables the inclusion of censored data, allowing us to account for fish that did not successfully pass each threshold of the raceway by the end of the 30 min trial. Time‐to‐event analysis calculates the likelihood that a fish will cross a threshold within the raceway at a particular time, given that it has not yet crossed that threshold (i.e. the event rate). We refer to the event rate at the approach threshold, entry threshold and passage threshold as the “approach rate,” “entry rate” and “passage rate,” respectively. The Cox regression model was used to model the instantaneous event rate at each threshold, which was right censored and modelled as a function of time, as shown in the following equation and explained by Goerig and Castro‐Santos (2017):

$$\lambda \left(t\right)={\lambda }_0\left(t\right)e^{\mathit{X\beta }+\mathit{Zb}}$$

where $\lambda \left(t\right)$ is the baseline hazard function (i.e. event rate) modelled as a function of time (t), X is the matrix of fixed effects, Z is the matrix of random effects, and $\beta$ and b are the fixed‐ and random‐effect coefficients, respectively. Group status (group vs. solitary fish) was the fixed effect, with trial number included as a frailty term to account for random effects. Water temperature was added as a covariate but was found not to be statistically significant and was therefore removed from the model (p > 0.05). Fish length and weight were not included in the model due to our inability to uniquely identify these characteristics for individual fish within groups in the flume. The selected model met Cox proportional hazards model assumptions that were tested using Schonfeld residuals. Based on the above equation, hazard ratios (HRs) were calculated for each treatment (group status), comparing the event rates between groups and solitary fish. A hazard ratio greater than 1 indicates that the treatment (group) has a higher event rate compared to the reference (solitary), a hazard ratio less than 1 indicates the treatment (group) has a lower event rate compared to the reference (solitary) and a hazard ratio equal to 1 indicates no difference in event rate between either treatment.

These models were implemented using the Coxme package in R version 4.2.0 (R Core Team, 2020; Therneau, 2015). The Likelihood Ratio Test, Wald Test, and Score (Logrank) Test were used to identify model fit. We developed three distinct models to answer specific questions:

1. The first model addresses the approach rate: the likelihood of a fish to cross the approach threshold per unit of time.

2. The second model addresses the entry rate: the likelihood of a fish to cross the entry threshold, once they had crossed the approach threshold, per unit of time.

3. The third model addresses the passage rate: the likelihood of a fish to cross the passage threshold, once they had crossed the entry threshold, per unit of time.

We then used the survfit function from the Coxme package (Therneau, 2015) to calculate the proportion of fish that are predicted to cross each threshold at a given time for each event rate model. As time progresses, the predicted proportion of fish crossing the threshold decreases because the number of available fish diminishes over time.

3.3. Active metabolic rate

An analysisy of covariance was used to test for the effects of group vs. solitary swimming on post‐exercise oxygen consumption. Oxygen consumption (mgO₂ g⁻¹ h⁻¹) was the response variable, with group as a categorical predictor variable and average fish weight as the continuous predictor variable. A Likelihood Ratio Test was used to determine the model of best fit for predictor variables starting with a fully saturated model including interactions. All possible interactions between body length and fish weight were not statistically significant and were removed from the model.

4. RESULTS

4.1. Modelling binary success

4.1.1. Approach

Eighty‐four percent of groups had at least one fish successfully approach the raceway, 32% of groups had all five fish approach the raceway and 16% of groups had no individuals approach the raceway (Table 2). Of the 58 solitary fish, 72% of individuals successfully approached the raceway and 27% of individuals did not. There was no statistically significant effect of group on the proportion of fish that successfully approached the raceway (p = 0.41) (Table 3).

TABLE 2.

Number of replicates for every proportion of solitary (n = 58) and groups of (n = 37) fish crossing each raceway threshold.

Raceway threshold	Group proportion						Solitary	Proportion
	0	0.2	0.4	0.6	0.8	1	0	1
Approach	6	1	5	4	9	12	16	42
Entry	6	1	8	2	9	11	26	32
Passage	6	1	8	2	9	11	34	24

TABLE 3.

Model summaries for each of the three binary logistic regression models.

Variable	Coefficient estimates	Standard error	z	p value
Approach success
Group	0.38	0.45	0.83	0.41
Entry success
Group	−0.27	0.43	−0.62	0.54
Passage success
Group	−0.82	0.43	−1.91	0.06

4.1.2. Entry

Eighty‐four percent of groups had at least one fish successfully enter the raceway (84%), 30% of groups had all five fish enter the raceway and 16% of groups had no individuals enter the raceway (Table 2). Of the 58 solitary fish, 55% of individuals successfully entered the raceway and 45% of individuals did not. There was no statistically significant effect of group on the proportion of fish that successfully entered the raceway (p = 0.54) (Table 3).

4.1.3. Passage

Eighty‐four percent of groups had at least one fish successfully pass the raceway (84%), 30% of groups had all five fish pass the raceway and 16% groups had no fish pass the raceway (Table 2). Of the 58 solitary fish, 41% of individuals successfully completed the raceway. There was no statistically significant effect of group on the proportion of fish that successfully completed the raceway (p = 0.06; Table 3). All three models fit model assumptions based on the Hosmer–Lemeshow test (χ ² values that fall below the 95% confidence interval and high p values).

4.2. Time‐to‐event analysis

4.2.1. Approach rate

Of the 95 trials in this study, six groups of fish and 16 individuals failed to approach the raceway but were still included in the analysis and censored at the maximum trial time (i.e. 30 mins). There was no statistically significant effect of group status on approach rate (p = 0.59) (Table 4). Groups of fish had a slight increase in approach rate compared to solitary fish, with an HR of 1.14 and confidence interval of 0.54–1.41. This indicates that groups are 14% more likely than solitary fish to cross the approach threshold by the next point in time. However, the confidence interval spanning 1 suggests insufficient evidence to support a statistically significant difference in approach rates between groups and individuals. Groups of fish had a median approach time of 644.5 s and solitary fish had a median time of 997 s (Figure 3).

TABLE 4.

Cox proportional hazards event rate model outputs for each raceway threshold.

Parameter	Coefficient ± standard error	Hazard ratio	p value
Approach rate
Group	0.13 ± 0.24	1.14	0.59
Frailty (trial)	–	–	0.93
Entry rate
Group	0.74 ± 0.26	2.10	0.004
Frailty (trial)	–	–	0.93
Passage rate
Group	1.08 ± 0.28	2.95	<0.001
Frailty (trial)	–	–	0.93

Note: Group status (group vs. solitary fish) was the fixed effect, with trial number included as a frailty term to account for random effects.

FIGURE 3.

Event rate models for each raceway threshold, representing the proportion of fish crossing each threshold as a function of time. Zero on the y axis shows that 0% of the population have crossed the threshold at that time (t) and 1 shows that 100% of the fish have crossed the threshold at time (t). Data are stratified by group status, where groups of fish are represented by the solid green line and solitary fish are represented by the dashed purple line. Shaded regions in the corresponding colours represent the 95% confidence intervals. Frailty terms are not included in the figure model.

4.2.2. Entry rate

Six groups of fish and 26 individuals failed to enter the raceway and were censored at the maximum trial time. There was a statistically significant effect of group status on entry rate (p = 0.004) (Table 4 and Figure 3). Groups of fish had a higher entry rate with an HR of 2.22, indicating groups were 122% more likely than solitary fish to cross the entry threshold by the next point in time. Groups had a median entry time of 15.8 s compared to solitary fish with a median entry time of 207 s (Figure 3).

4.2.3. Passage rate

Six groups of fish and 34 individuals failed to pass the raceway and were censored at the maximum trial time. There was a statistically significant effect of group status on passage rate (p < 0.001) (Table 4 and Figure 3). Groups of fish had a higher passage rate, with an HR of 2.49, indicating groups were 149% more likely to cross the passage threshold by the next point in time than solitary fish. Groups had a median passage time of 34.5 s. However, fewer than 50% of solitary fish crossed the passage threshold, meaning that we were unable to determine a median passage time (Figure 3).

4.3. Respirometry

Successful fish swimming in groups had significantly lower active metabolic rates (MMRs) than successful solitary fish (p < 0.001; Table 5 and Figure 4). Fish weight did not have a statistically significant effect on MMR (p = 0.12) but was kept in the model as weight is a known factor influencing metabolic rates (Norin & Clark, 2016; McKenzie & Claireaux, 2010).

TABLE 5.

Summary of Satterthwaite's type III analysis of covariance comparing the active metabolic rates of group vs. solitary swimming fish.

Source	Sum of squares	Mean square	Numerator degrees of freedom	Denominator degrees of freedom	F value	p (>F)
Group	115.60	115.62	1	49	21.47	<0.001
Weight	18.10	18.13	1	49	2.50	0.12

FIGURE 4.

Boxplots of active metabolic rates (mgO₂ g⁻¹ h ⁻¹) of fish swimming in groups compared to individual solitary swimming fish. The data points within each treatment group represent individual replicates and are sized based on average fish length, with larger points representing larger fish. Green represents groups and purple represents solitary fish.

5. DISCUSSION

Our study used a controlled experimental design to assess the impact of group swimming on passage success at a water speed barrier. The results showed that group swimming significantly reduced the time required for fish to successfully pass the barrier. However, there was no effect of group on approach rates. There was also no measurable effect of group on the probability of approach success, entry success or passage success. These findings, along with the observation that successful groups had significantly reduced active metabolic rates compared to successful individuals, suggest physiological and behavioural advantages contribute to the success of collective movement. Although group swimming did not influence the initial stages of approaching and entering the raceway, group swimming significantly improved overall success in raceway passage. We found that fish swimming in groups had significantly lower metabolic rates than successful solitary fish, suggesting potential physiological and social benefits of group swimming (Berdahl et al., 2018; Lefrançois et al., 2009; Nadler et al., 2016; Parker, 1973; Weihs, 1973).

One possible explanation is that group swimming provides hydrodynamic advantages, reducing the relative water velocity experienced by individuals and lowering their energy expenditure. There are three potential hydrodynamic theories to explain this reduction in energy expenditure: the vortex hypothesis, the channelling hypothesis and the wingtip‐uplift hypothesis (Abrahams & Colgan, 1987; Weihs, 1973). These effects allow individual fish in groups to conserve energy while maintaining movement through the high‐velocity zone of the raceway. Additionally, social buffering effects may help reduce stress‐induced metabolic costs (Lefrançois et al., 2009; Nadler et al., 2016; Parker, 1973), further improving passage success. The presence of conspecifics has been shown to lower cortisol levels, regulate antioxidant gene expression and reduce overall physiological stress, which can otherwise elevate metabolic rates in solitary fish (Gilmour & Bard, 2022; Parker, 1973; Schumann et al., 2023). Together, these hydrodynamic and social benefits may explain why fish swimming in groups had a lower energetic cost of passage, allowing them to traverse velocity barriers more efficiently than solitary individuals.

We caveat this hypothesis because all successful fish in a group were placed in a single respirometer. Consequently, we cannot definitively determine whether their lower metabolic rates were due to an actual reduction in individual energy expenditure during passage or were confounded by the calming effect of being tested with conspecifics inside the respirometer. This is an important limitation of our study, as the availability of respirometry equipment restricted our ability to measure individual metabolic rates within groups. To control for this in the future, all fish should be placed into individual respirometers to accurately measure the effects of group swimming and passage success on individual energy expenditure.

There was a significant difference in Fulton's condition factor between group and solitary swimming fish (p = 0.03), with fish in groups having a higher condition factor. This represents a limitation of our study, as fish condition is a known predictor of swimming performance, with larger, heavier fish generally being more successful (Laborde et al., 2016). Therefore, the higher passage rates observed in group fish may be partly due to their more favourable body condition, rather than solely the effects of swimming with conspecifics.

Group swimming did not seem to influence the overall probability of approach or entry success. For Galaxias maculatus, it appears that group behaviour may not serve as an explanatory factor for fishway attraction or for motivation to enter the raceway. However, the advantages of group swimming become apparent during the successful traversing of the raceway. We observed significantly faster times for groups crossing the passage threshold, with the majority of fish in groups passing the raceway in under 60 s. One possible explanation for the lack of a significant difference in approach or entry success is that water velocity at these thresholds was not high enough to provide a clear advantage for individuals in groups compared to solitary fish. The average entry threshold velocity (0.28 m s⁻¹) corresponds to the average critical swimming speed of Galaxias maculatus (Crawford et al., 2025) and is significantly lower than the average velocity in the test section of the raceway (0.46 m s⁻¹), which only the strongest 10% of G. maculatus could sustain. Solitary fish were likely able to cross lower water velocities at the approach and entry thresholds without difficulty. In contrast, the higher water velocities within the raceway may have provided groups with hydrodynamic and social buffering advantages, increasing their likelihood of completing passage, whereas only the fastest and strongest solitary fish were expected to succeed.

Another explanation is that, during the initial stages of entry and approach, individual differences appear to play a more significant role than collective behaviour. We theorize that personality may be involved, and that bolder and more active individuals exhibit a higher likelihood of approaching a challenge compared to shyer ones, consistent with previous studies (Mensinger et al., 2021; Mozzi et al., 2024; Nyqvist et al., 2024; Tang & Fu, 2020). It has been observed that bold individuals tend to commit quickly to actions and exhibit less variability in their movements (Dahlbom et al., 2011). This may contribute to why we did not observe any benefit of group swimming on approach rates or probability of approaching or entering the raceway. While the velocity at the approach and entrance of the raceway was lower, higher turbulence and changes in flow direction (such as the formation of eddies) may have interfered with an individual's ability to sense its neighbours, potentially negating the positive benefits of group swimming (e.g. hydrodynamic advantages) (Chicoli et al., 2014; Codling et al., 2007; Zhang et al., 2024). Conversely, the approach or entrance to the raceway may not have been sufficiently intimidating to deter less bold fish from approaching. As a result, shy solitary fish and fish in groups may have been equally likely to approach or enter the raceway. However, once in the raceway, group swimming behaviour demonstrates more pronounced benefits for the overall success and time reduction in completing the raceway.

Studies on the impacts of social interactions on fish passage success have largely focused on salmonid species (Dominy, 1973; Goerig & Castro‐Santos, 2017; Johnson et al., 2012; Lemasson et al., 2014; Okasaki et al., 2020). These studies typically focus on existing fishways using fish density as a predictor variable for passage success. Since these structures have predetermined size constraints, much of this research has focused on the negative effects of high fish densities on passage, rather than potential benefits of collective navigation. In contrast, our research took an alternative approach by using an experimental raceway where we controlled the number of individuals in each trial, effectively removing any confounding effects of overcrowding within the structure. This setup also allowed us to directly compare the passage performance of solitary fish to groups of fish, which is not as easily done in a real‐world fishway. Additionally, it is important to understand how social interactions impact collective navigation and passage success for a variety of species to create more tailored fish passage solutions (Birnie‐Gauvin et al., 2019; Mozzi et al., 2024). While researching the density effects of salmonids in a fishway, Okasaki et al. (2020) found that the effects of social interactions are species dependent. Their findings revealed that chinook salmon benefit from higher densities during passage, while sockeye salmon do not show any density‐dependent response. They suggest that some fish species, like chinook salmon, might use their conspecifics as navigational guides or homing signals, enhancing their ability to navigate challenging passages. In contrast, sockeye salmon may not rely on such social cues for navigation, leading to a lack of positive effects from increased density. Similarly, studies have demonstrated links between collective movement and fish passage efficiency (e.g. Albayrak et al., 2020; Nyqvist et al., 2024). For instance, Nyqvist et al. (2024) found that fish in groups of five passed an experimental weir at significantly higher rates than solitary individuals, highlighting the role of conspecifics in successful passage. These findings support the idea that social interactions and hydrodynamic benefits contribute to improved passage outcomes in group‐swimming fish.

The combination of previous research highlighting the negative effects of overcrowding in fishways and our findings demonstrating the positive effect of group swimming on passage success highlight the importance of considering group dynamics in fish passage design. While our experiments were conducted with groups of five individuals, the observed benefits of group passage suggest that appropriately scaled fish passage structures could support larger groups. However, caution is needed when extrapolating these findings to larger groups of fish because the dynamics of group swimming may differ with increasing group size, potentially leading to overcrowding or competition for space.

Given that group swimming did not seem to offer benefits in terms of approach or entrance rates in our experiments, this suggests a need for improved strategies in fishway entrance design to cater to more individualized behaviour. Designing fishways to attract and guide less bold fish requires features such as entrances located where fish naturally aggregate and flows that effectively draw fish toward the entrance (Tan et al., 2021). Additionally, the significant positive impact of group swimming on passage success underscores the need to design fishways that support and encourage group movement, ensuring that fishways accommodate a range of behaviours and social dynamics. Practical examples may include constructing larger resting pools or low‐speed zones capable of accommodating entire shoals of fish simultaneously. Haro et al. (1998) found that the ability of Atlantic salmon and American shad to navigate through weirs decreases as group size increases, which suggests that artificial structures not specifically designed to accommodate large groups of fish may pose challenges for successful passage of these groups. Given that many published studies highlight overcrowding as a hindrance to successful fish passage (Dominy, 1973; Goerig & Castro‐Santos, 2017; Johnson et al., 2012; Lemasson et al., 2014), there is a particular emphasis on designing fish passes that accommodate the natural shoal sizes of fish. This approach aims to mitigate the negative impacts associated with overcrowding, ensuring that fish can navigate through passage systems more effectively and continue to benefit from group swimming. A more nuanced approach to the design of fishways could enhance their efficiency for both individual and group‐oriented swimming behaviours.

AUTHOR CONTRIBUTIONS

R.C., E.G., P.F. and B.H. conceived and designed the study. R.C. conducted the study, managed the data and led the writing of the manuscript. E.G., P.F. and B.H. edited the manuscript. All authors gave their approval for manuscript submission. All authors contributed critically to the drafts and gave final approval for publication.

FUNDING INFORMATION

This work was supported by the Ministry for Business Innovation and Employments (CO1X615) and a University of Waikato Doctoral Scholarship.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Data S1. Supporting Information.

Crawford, R. M. B. , Gee, E. M. , Hicks, B. J. , & Franklin, P. A. (2025). Group swimming significantly decreases time to passage success for a galaxiid species. Journal of Fish Biology, 107(2), 372–383. 10.1111/jfb.70040