When feeling is better than seeing : Adult Zebrafish Ignore Wide-Field Optic-Flow in Laminar, but not Turbulent Hydrodynamic Environments

13 Many animals navigate their world largely by seeing and feeling it. To disentangle these visual 14 and mechanosensory contributions, we developed a virtual reality assay targeting the 15 optomotor response in adult wild-type zebrafish swimming against flow. By projecting dynamic 16 visual patterns onto the walls of a variable-speed flow tank, we decoupled wide-field optic flow 17 from hydrodynamic velocity. We then tested fish responses to abrupt visual perturbations 18 while they held station in the unsteady wake behind a bluff body. These perturbations reliably 19 elicited compensatory optomotor responses, with fish aligning to the direction of the moving 20 stimulus. Notably, this behavior was absent in uniform flows, suggesting that fish prioritize 21 visual input when predictive lateral line signaling is compromised. We propose that this 22 sensory shift serves to optimize swimming energetics in turbulent wakes. Extending this 23 framework, we further show that zebrafish swimming against flow, whether alone or in groups, 24 exhibit heightened escape responses to looming visual stimuli. Together, our findings reveal 25 that fish sensory strategies are not fixed but dynamically tuned to hydrodynamic context: 26 favoring visual cues in turbulent environments and lateral line input in uniform flows. 27 28 Graphical Abstract 29 30 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 3

31 Sensory cues from different modalities often arrive simultaneously or overlap in sequence, 32 providing animals with a rich and redundant source of information to navigate novel and 33 complex environments. A growing acknowledgment of the integration and conflict dynamics of 34 multiple sensory modalities (vision is coupled with olfaction, air flow, vestibular, etc.) has led 35 to new insights in interpreting and understanding behavioral responses at the organismal 36 level. 37 Animals moving through water experience vastly different challenges from terrestrial animals. 38 In particular, the higher density and viscosity of water, the forces currents can create, and the 39 relative lack of light characterize aquatic and marine habitats where vertebrate life originated. 40 For over 400 million years, fishes have evolved to integrate mechanosensory and visual 41 information to shape their behavior and ecology. The lateral line and visual system, detecting 42 water flow and light, play a critical role in this integration, enabling fish to navigate and 43 respond to their environment. The lateral line is an ancient mechanosensory system that 44 predates the evolution of visual systems in vertebrates, and detects predators, prey, 45 conspecifics and water currents. Specifically, genes associated with mechanosensation 46 appear earlier in evolution than those related to visual processing (Šestak et al., 2013). This 47 suggests that sensing water flow and pressure was a foundational capability for early aquatic 48 vertebrates. However, despite the growing interest in understanding how multiple sensory 49 systems orchestrate behavior (Dallmann et al., 2023; Sharma & Sponberg, 2023), our 50 understanding of mechanosensation and vision in aquatic animals remains static and poorly 51 understood. 52 In contrast to the highly stochastic nature of ambient turbulent flows, which, when interacting 53 with environmental structures (rocks, vegetation, other animals), produce vortices of widely 54 varying spatial and temporal scales, the boundary layer flow over a swimming fish's skin (and 55 thus lateral line system) is relatively predictable and repeatable (Gray, 1968). When flow 56 interacts with a simple geometric shape like a cylinder, a vortex street can be generated. Fish 57 can recapture the energy of these vortices to hold station in flow (e.g. maintain position 58 relative to the Earth frame of reference). Fish holding station behind a cylinder show 59 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 4 drastically reduced oxygen demands, saving up to half the cost of swimming compared to 60 when swimming in laminar flow (Taguchi & Liao, 2011). Due to the turbulent regime in a 61 vortex street, flow unpredictability causes destabilizing movements in these surfing fish (Liao 62 et al., 2003b; Tritico & Cotel, 2010). Consistent with this, fish with an ablated lateral line avoid 63 turbulence vortex streets and prefer to station holding in areas of smoother flow, suggesting a 64 reliance on detecting flows to maintain position while station holding (Liao, 2006). Turbulence 65 may limit the capabilities of the lateral line to signal normal swimming movements, where 66 muscle commands align with sensory expectations. Predictable flow across the body is critical 67 for the lateral line to generate an image of efficient swimming, enabling a proprioceptive 68 function (Skandalis et al., 2021). 69 How does the hydrodynamic environment influence the reliance on visual information for 70 station holding in adult zebrafish? We hypothesize that fish swimming in turbulence shift their 71 sensory reliance from the lateral line towards vision. Here, we develop a novel virtual reality 72 assay in a flow tank that decouples visual and hydrodynamic sensory inputs for freely 73 swimming fishes. We use this approach to investigate the interplay between vision and the 74 lateral line during optomotor and loom behaviors across laminar and turbulent flow conditions. 75 We address previously inaccessible questions on the effect of wide-field visual on fish 76 swimming and escape behaviors in flow. By decoupling and placing into conflict visual and 77 hydrodynamic stimuli, our approach allows investigation into multi-agent and multisensory 78 integration of fish behavior. 79 80

81 Animals 82 We used adult (>60 dpf) WT zebrafish, Danio rerio (body-length, mean± SE = 38.6 ± 0.7 mm, 83 n=20 fish) raised in the UC Santa Barbara zebrafish facility and transferred to the 84 experimental room >2 weeks prior to experiments. There, fish were maintained in two 10 L 85 freshwater tanks maintained at 23 ± 0.5 °C) with a commercial aquarium heater (Eheim Co.), 86 kept on a 12:12 light:dark cycle and fed commercial pellets ad libitum daily. Prior to the start 87 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 5 of an experimental trial, an individual fish was introduced into the flow tank and left for 10 88 minutes to acclimatize at a current velocity of 10 cm s-1 (e.g. ~2.5 body-lengths s -1). We found 89 this flow velocity would elicit the most reliable station holding response. All experimental trials 90 were conducted in the afternoon (12:00-17:00 PST) in a room enclosed by blackout curtains. 91 After data were collected from swimming trials, fish were euthanized with an over-dose of MS-92 222. 93 Experimental Setup 94 Flow tank: All experiments were conducted using a custom-built 5 L recirculating flow tank 95 with a working section of 22 x 7 x 7 cm (length x width x depth). Water flow was generated by 96 a variable speed AC-DC series motor (Dayton model 2MO37A, 115v 1.5 Amp, Lake Forest 97 Illinois USA) driving a propeller that circulated water through 2 sets of honeycomb collimators 98 (1/8" aperture diameter) to generate uniform flow. Flow velocity was set at 10cm/s and verified 99 by tracking suspended plastic particles within the working section of the flow tank. The cross-100 sectional area of the fish was less than 5% of the cross-sectional area of the flow tank, 101 minimizing any solid blocking effects (Bell & Terhune, 1970). Water was filtered, aerated, and 102 maintained at room temperature of 22.1 °C (SE=0.1°C) throughout the experiment. (See 103 Figure-1A) 104 Visual Projection: We used a system of mirrors (See Figure-1B) to project visual stimuli on 105 customized rear-projection screens mounted to the flow tank walls. For wide-field grating 106 patterns on the sidewalls, we varied orientation and motion (vertical gratings moving 107 downstream/upstream; horizontal gratings moving up/down) and set the optic-flow speed as 108 ca. 10 cm s-1 (spatial frequency=1.8 cm cycle-1, temporal frequency as 5.5 cycle s-1) to match 109 hydrodynamic flow velocity. We recorded these moving projection patterns using a high-110 speed camera at 1000 frames s-1 to verify optic flow rates. For visual looming stimulus, we 111 used an exponentially expanding (doubling-time: diameter=50ms, area=25ms) dark-circle with 112 a bright background presented on the top-wall. Fish were acclimatized for 5 minutes before 113 each trial with either horizontal or vertical gratings on sidewalls in case of station-holding, or 114 plain white background on the top-wall in case of escape trials. For each fish trial, the 115 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 6 treatments order was randomized to evoke a reflexive response and avoid contribution due to 116 learning and memory. 117 Filming and Digitization: A monochromatic Chronos 1.4-HD high-speed video camera (1024 118 x 1024, 1000 frames/s, Burnaby BC Canada) was aimed at a front-surface mirror angled at 119 45° and placed directly below the working section to film the swimming kinematics and 120 position of zebrafish (Figure-1B). An LED panel with white lights and overlayed diffuser was 121 used to optimize the image contrast of the fish. In loom experiments, this panel was replaced 122 with a projection screen for the top-projected loom stimulus with against a uniform 123

illumination. We used a semi-supervised machine-learning based tool, 124 DeepLabCut (Mathis et al., 2018), that allowed us to reliably track 4 points on fish (head-tip, 125 tail-tip, right- and left-pectoral fin base). To create a training-set, we manually labelled these 4 126 points in a total of 800 frames, randomly picked from 40 videos, and trained a neural network 127 model for ~1 million iterations to minimize the tracking error. We subsequently used this 128 model to track points on more than 400,000 frames from 80 video recordings. Tracked videos 129 using DeepLabCut were visually verified and approved before further analysis (Mathis et al., 130 2018). 131 Experimental Design and Data Analysis 132 We studied the effects of hydrodynamic and visual conditions on freely swimming adult 133 zebrafish using two widely studied behaviors, Station-holding and Escape behavior. Below 134 are the experiment specific details for each behavior. 135 (i) Station-Holding Behavior: 136 Water-flow conditions – Steady / Unsteady: Many fish routinely swim against water current to 137 navigate, a behavior known as rheotaxis. These fish may experience Steady (laminar) or 138 Unsteady (turbulent) water-flow. However, station holding fish would naturally experience 139 unsteady water-flow as they swim behind rocks or bluff bodies to get help from eddies and 140 conserve energy. We studied sensory prioritization in the position-maintaining fish in a flow 141 tank and challenged them with steady and unsteady water-flow conditions. Steady flow 142 conditions were generated by a custom 3D printed honeycomb flow-straightener. D-section 143 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 7 cylinders (0.5-1 cm diameter) were added in an upstream location in the working section to 144 create a distinct flow refuge to generate suitable unsteady flow (Liao et al 2003b). 145 Linear perturbations – Optical-Pull / Push: We first tested the role of wide field visual 146 perturbation on a station holding adult fish. As the fish is swimming against water-flow, we 147 externally provided wide-field optic-flow by suddenly moving the surroundings in (i.e., 148 vertically oriented visual gratings) forward (upstream) or backwards (downstream) directions. 149 The fish typically experience optic-flow either during self-movements or when external factors 150 (such as water-flow) move them. Since our treatments induced the optic-flow analogous to 151 fish moved by external factors, we call them ‘optical-push’ and ‘optical-pull’, based on the 152 direction of optic-flow generated. ‘Optical-push’ is the treatment of moving visual surroundings 153 from back to front of the fish, mimicking the optic-flow direction if fish are pushed backwards 154 externally (e.g., sudden increase of water-flow speeds). Similarly, optical-pull treatment will 155 suddenly move wide-field visual patterns from front to back (i.e., in a downstream direction) 156 mimicking a scenario when a station holding fish is pulled forward due to external factors 157 (e.g., sudden decrease of water-flow speeds). The prefix “optical” here refers to the fact that 158 it’s a purely visual perturbations, rather than changing waterflow speed to move the fish. 159 Rotational perturbations – Optical-Roll: We further checked the effects of a conflicting optic-160 flow on station holding fish, where the optic-flow presented along rotational axis whereas the 161 water-flow remained linear (along the body-axis). We moved visual surroundings (i.e., 162 horizontally oriented visual gratings) in opposite directions on each side wall at the same 163 spatial and temporal rates as the ‘Optical-Push / Pull’ treatments (see Methods: Visual 164 Projections). We call them ‘optical roll’ treatments that would mimic optic-flow generated to 165 the fish if it is externally rotated in Roll-direction (around the longitudinal-axis). The rotational 166 optic-flow is either in clockwise (CW: right-side moving down, left moving up) or 167 counterclockwise (CCW: right-side moving up, left moving down) directions from the fish’s 168 perspective. 169 Data Analysis: To examine the effect on station holding behavior, we compared 1 second of 170 fish trajectory data before and after the stimulus and quantified changes in body position, 171 swimming velocity, and tail beat. We excluded the first 500ms of data after the stimulus onset 172 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 8 from this analysis. Swimming velocity was divided into longitudinal (along the direction of flow) 173 and lateral (side-ways) directional components to better explain the behaviors observations. 174 Changes in trajectory pattern were quantified using Spearman’s rank correlation (rho), which 175 measures monotonic changes in the position (where a constant upstream movement = 1, and 176 downstream = -1). We also tracked the distal end of tail and quantified its average cyclic 177 movement amplitude and frequency over a given duration for each trial. 178 (ii) Escape Behavior: 179 Experimental Treatments: We studied another naturalistic behavior that is known to involve 180 vision, an escape behavior, to further compare how fish use vision in different hydrodynamic 181 conditions. We presented an exponentially expanding (looming) stimuli (see Methods: Visual 182 Projections) at the top wall of swimming chamber while fish were challenged to swim against 183 water-flow (Flow) and when at rest (No-Flow). We studied fish’s response when alone in the 184 chamber (Single) or when in a group of five individuals (Group). The looming stimulus 185 generally originated towards the upstream side of the chamber, giving better visibility to the 186 fish swimming against the flow. However, we also carried out a treatment where the position 187 is shifted towards the downstream side of the chamber (Downstream Loom), to study any 188 possible effects arising due to the stimulus position. 189 Data Analysis: We studied fish’s reflexive escape responses involving sudden bending of the 190 body, known as C-Start, upon presenting the looming stimulus. From the video recordings, we 191 find fish’s instantaneous position w.r.t. the stimulus origin in horizontal plane ([px,py] cm) and 192 the instantaneous stimulus radius (R cm) at the time of an escape response (i.e., onset of C-193 start reflex). Vertical position of the fish (pz cm) is considered as the flow-tank height, as the 194 fish were found swimming at the bottom of the tank. From the escaping fish’s three 195 dimensional position (P=[px,py,pz] cm) and stimulus radius (R), we can then calculate the 196 threshold stimulus angle (theta) from the fish’s perspective that elicited the escape response 197 (using the formula: theta = 2*arctan( R / Norm(P) ). We call this angle (theta) the Perceived 198 Loom Angle. We also measured the escape response delay as the time difference between 199 beginning of the looming stimulus and the C-start. 200 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 9

201 Prioritizing vision depends on hydrodynamic environment 202 We quantified fish swimming performance and their responses to wide-field visual 203 perturbation (Optical-Pull and -Push treatments, see methods: Exp. Design) while swimming 204 against steady and unsteady water currents in the flow-tank. For each treatment, we observed 205 responses from 4 adult zebrafish and repeated 4 trials for everyone, making a total 16 unique 206 trials per treatment. We used Wilcoxon signed-rank test to compare behavior responses 207 between the pre-stimulus and post-stimulus values. 208 Fish do not rely on optic-flow to maintain swimming position in steady flows 209 During trials with steady (laminar) flow, fish’s swimming velocities in the streamwise direction 210 remain unchanged after the stimulus onset when compared to the pre-stimulus adaptation 211 period for Optical-Pull (velocities mean ± s.e.m.: pre-stim = -0.2 ± 1.2 cm/s, post-stim = 0.2 ± 212 1.2 cm/s; p = 0.90, n.s.) and, also for Optical-Push treatments (velocities mean ± s.e.m. : pre-213 stim = -0.1 ± 1.0 cm/s, post-stim = -0.9 ± 1.1 cm/s; p = 0.74, n.s.). Conversely, presenting 214 exactly the same visual stimulus but now with unsteady (turbulent) currents elicited 215 stereotypical compensatory optomotor response (Figure-2 C,D). In unsteady flow, swimming 216 velocities in streamwise-directions reduces for Optical-Pull (velocities mean ± s.e.m.: pre-stim 217 = -0.2 ± 0.4 cm/s, post-stim = -4.2 ± 0.6 cm/s; p = 0.0004), whereas it increases for Optical-218 Push treatment (velocities mean ± s.e.m.: pre-stim = -0.3 ± 0.6 cm/s, post-stim = 2.3 ± 0.6 219 cm/s; p = 0.009). This behavior can be explained as positive optomotor response (OMR), 220 since the fish’s post-stimulus displacement is in the same direction as the visual projections 221 on sidewalls, for both Optical-Pull and Push perturbations in unsteady water currents. A 222 compensatory positive OMR may help station-holding fish to maintain its position using visual 223 feedback. Such OMR is missing for both the treatments in steady currents, suggesting a 224 prominently non-visual sensory mechanism to maintain position. 225 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 10 Fish move monotonously with wide-field optic flow in unsteady flows 226 We compared fish’s trajectories with a purely linear model trajectory (with a slope=1) and 227 quantified Spearman’s correlation coefficient (rho) to study the monotonicity of their 228 responses for each treatment. Rho values of 1 and -1 represent a perfectly monotonous 229 change of position in the forward and in backwards direction, respectively. We found that fish 230 swimming in unsteady flows showed a higher monotonous change in their body position post-231 stimulus onset. Rho value for Optical-Pull (rho ± s.e.m : pre-stim = -0.2±0.2, post-stim = -232 0.8±0.1, p=0.009) representing a monotonous downstream drift, and for Optical-Push (rho ± 233 s.e.m : pre-stim = -0.2±0.2, post-stim = 0.6±0.1, p=0.04) representing largely monotonous 234 upstream surge, post stimulus onset in unsteady flows. However, when in steady flows, the 235 rho values are spread across the spectrum with averages close to zero representing a lack of 236 overall monotonous movements (Optical-Pull: pre-stim = 0.0±0.2, post-stim = 0.0±0.2, p=0.99, 237 n.s.; Optical-Push: pre-stim = 0.1±0.2, post-stim = -0.1±0.2, p=0.74) (Figure-2). 238 Compensatory OMR only present for streamwise (longitudinal) direction 239 For both steady and unsteady hydrodynamic conditions, the post-stimulus change in position 240 happened in the streamwise direction (Figure-2, longitudinal), in the same direction as the 241 visual stimulus. Cross stream (e.g. lateral) position, velocity, and Spearman rho values remain 242 unchanged post-stimulus (Figure-S1, all p>0.4, n.s.). We further found that when presenting 243 optical-roll perturbations (i.e., clockwise and counter-clockwise), fish did not display an OMR 244 either in longitudinal or lateral directions (Figure-S2, all p>0.4, n.s.). We also note that fish did 245 not turn, change its swimming direction or show an escape response (e.g., C-start) to any of 246 our wide-field visual stimuli. Raw trajectories and summary statistics values of the means, 247 s.e.m and p-values are included in supplemental material (Supp. Figure-1,2, Supp.Table-S1). 248 Thus, our results of visual perturbation experiments during station holding suggest that adult 249 zebrafish use optic-flow to maintain position while swimming against unsteady water streams, 250 and the compensatory optomotor response to optic-flow depends on the direction of optic-flow 251 and hydrodynamic conditions. 252 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 11 Fish sensitivity to visual threats alters with hydrodynamic conditions 253 As our previous experiment showed that fish swimming against water currents did not trigger 254 escape responses (e.g. C-start) to wide-field visual perturbations, next we presented an 255 exponentially expanding, purely visual looming stimulus on the dorsal wall of the flow tank and 256 studied their behavioral responses. We quantified fish’s escape attempts responses (C-start) 257 due to looming stimulus presented during swimming against water current (Flow) and no 258 currents (No-Flow), both for the individual trails (Single) or when schooling (Group) (Figure-3). 259 We used Wilcoxon rank-sum test to compare quantities between Flow and No-Flow 260 treatments. 261 Escape Distance-Delay relationship depends on hydrodynamic conditions 262 Adult zebrafish escape responses for both Single and Group trials showed that the escaping 263 fish were positioned closer to the stimulus origin as compared to the non-escaping fish. This 264 relationship between the escape response presence and the distance from stimulus are 265 consistent across hydrodynamic conditions, Flow (distance mean ± s.e.m.: Escape = 48.5±6.3 266 mm, n=36; No-Escape = 93.7±7.2 mm, n=53; p=9.1e-6) and No-Flow (Escape = 41.8±4.2 267 mm, n=41; No-Escape = 60.3±6.4 mm, n=48; p=0.002) (Figure-3A). Thus, we considered 268 fish’s position when triggering a C-start response in our subsequent analysis of the behavior. 269 We also quantified the escape response delay since the start of looming stimulus expansion 270 for each of the escaping fish in all trials and studied its relationship with fish’s distance from 271 stimulus to find whether closely positioned fish escaped faster. Flow trials show a stronger 272 positive relationship (slope m = 0.20 for Single, m = 0.24 for Group) when fishing swimming 273 against current than in No-Flow trials (slope m = 0.05 in Single, m = 0.04 in Group) (Figure-3 274 B). We then performed linear regression analysis because raw slope values are susceptible to 275 any mismatch in axis ranges for distance and delay. Combining escape events from Single 276 and Group trials, we found a stronger escape distance-delay correlation for the Flow 277 treatment (Pearson coefficient = 0.67, n=36, p<0.001) than in No-Flow conditions (Pearson 278 coefficient = 0.21, n=41, n.s.). This analysis shows that the fish that were swimming against 279 flow triggered C-start sooner when located closer and later when further away from the 280 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 12 stimulus. Whereas the non-swimming fish in general did not show such response, suggesting 281 they just responded to the presence of the threatening stimulus. 282 Station holding fish show a lower threshold-angle to looming stimulus 283 The escape behavior may be better studied by quantifying the threshold stimulus angle for 284 each escaping individual because it would account for both the response delay and fish’s 285 distance from the stimulus. We calculated the threshold stimulus angle of the looming 286 stimulus from the perspective of the fish in three-dimensional space at the time of escape to 287 test whether station holding fish (Flow trials) responded to the angular expansion faster than 288 in No-Flow trials, as it would be predicted by their slope-values (Figure-3B). To maintain 289 consistency across each responses, we considered all trials for which the looming stimulus 290 was originating within the lateral-visual field (153° field on right and left side each; n=50 fish) 291 for the given fish positions, and excluded data with the stimuli in their blind-spot (21° field on 292 rear side; n=5 trials) and binocular (33° field on front; n=22 trials) (Pita et al., 2015). This 293 allowed us to better compare our data across Flow and No-Flow trials because fish’s body 294 orientations are not always facing upstream when not swimming, as they would while 295 swimming against water currents (No-Flow trials). 296 We found that station holding fish triggered an escape at a lower stimulus angle in Flow as 297 compared to stationary fish in No-Flow (Figure-3C). These angles are plotted on a model 298 looming stimulus to visualize their median threshold angles on the expanding visual stimulus 299 in temporal axis. Finally, combining the threshold stimulus angles from each escaping fish 300 across all the Single and Group trials, we found a lower threshold for fish swimming against 301 water current (Flow: 15.9°±3.9° deg, n=21) than fish in stationary water (No-Flow: 32.6°±6.8° 302 deg, n=29) (Figure-3D). A lower threshold angle (p=0.017) to trigger a C-start reflex suggest 303 increased sensitivity to the visual threat stimuli. 304 Together, our results highlight the relationship between the optic-flow and hydrodynamic flow, 305 where fish swimming in complex water currents elevate their response to optic-flow 306 perturbations. 307 308 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 13

309 The OMR is crucial for an animal's ability to maintain its position, and is particularly vital for 310 fish navigating dynamic aquatic habitats. Our research offers new insights into the adaptive 311 strategies of sensory modalities in fishes, revealing how they prioritize visual and 312 mechanosensory cues based on the predictability of their hydrodynamic environment, and 313 how these strategies influence other crucial behaviors like escape responses. 314 Developmental and Hydrodynamic Influences on OMR 315 Our study of adult zebrafish reveals a distinct OMR compared to that observed in larvae. 316 While larval zebrafish typically exhibit a positive OMR, swimming with moving bars to reduce 317 optic flow (Olive et al., 2016), adults demonstrate a negative OMR, swimming against the 318 visual motion (Bak-Coleman et al., 2015). This developmental shift likely reflects the 319 maturation and calibration of sensory and motor systems (Kohashi et al., 2012). Larvae, with 320 less developed sensory systems and limited proprioception, may primarily rely on visual cues 321 for body displacement. In contrast, fully integrated adult systems allow for the development of 322 robust expectations of flow and the accumulation of error-driven motor learning experiences 323 (Montgomery et al., 2002; Skandalis et al., 2021). This calibration enables adults to form 324 precise expectations of how lateral line (and likely vestibular) inputs correspond with visual 325 inputs in predictable hydrodynamic environments like still water or steady flow. 326 However, this calibrated expectation for sensory information, typically associated with 327 steady swimming kinematics, is violated in turbulent flows. We suggest that the positive OMR 328 observed in larvae arises because they initially rely heavily on visual inputs as they gain the 329 experience necessary to fully calibrate their mechanosensory and visual systems. A 330 compelling avenue for future investigation would be to precisely determine the developmental 331 stage at which this OMR switch occurs, and whether this transition can be accelerated or 332 delayed by specific environmental conditions like light levels or turbulence. 333 Beyond neural development, the hydrodynamic regime itself imposes distinct 334 challenges and opportunities for sensory processing. Our findings indicate that fish 335 dynamically adjust their reliance on mechanosensation versus vision in a context-dependent 336 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 14 manner, revealing a previously unrecognized prioritization strategy driven by hydrodynamic 337 predictability, which correlates with Reynolds number (Re). Larval fish operate at low 338 Reynolds numbers (Re < 100), where viscous forces dominate and flow patterns are highly 339 predictable. In this regime, consistent lateral line reafference from self-motion likely allows 340 larvae to prioritize visual inputs for OMR, as mechanosensory input provides a stable internal 341 reference. Conversely, adult zebrafish operate at high Reynolds numbers (Re > 1000), 342 routinely encountering turbulent flows characterized by complex, unpredictable vortices in 343 natural aquatic environments. When uniform flow interacts with bluff bodies, the resulting 344 vortices can significantly alter swimming kinematics (Liao et al., 2003a; Sutterlin & Waddy, 345 1975). Consequently, the lateral line system receives less predictable input than during self-346 generated swimming in uniform flow (Crapse & Sommer, 2008). While vortices can be 347 detected (Chagnaud et al., 2007), their presence likely diminishes the lateral line's capacity to 348 provide a consistent signal for self-motion or external flow, substantially reducing its reliability 349 as an accurate indicator of effective swimming in turbulence (Skandalis et al., 2021). 350 Similar principles of context-dependent sensory reweighting have been observed in 351 terrestrial systems. For example, hawkmoths tracking moving flowers modulate their reliance 352 on vision and mechanosensation depending on ambient luminance (Sharma & Sponberg, 353 2023). In bright light, visual gain increases while mechanosensory gain decreases; in dim 354 light, the reverse occurs. This mirrors our findings in zebrafish, where lateral line input 355 becomes less reliable in turbulent flow, prompting a shift toward visual cues. In both cases, 356 animals preserve behavioral performance—flower tracking in moths, station-holding and 357 escape in fish—by flexibly reconfiguring sensory strategies. These parallels suggest that 358 adaptive sensory prioritization, rather than fixed integration, may be a widespread solution to 359 environmental uncertainty. 360 Energetic Imperatives and Adaptive Sensory Prioritization 361 The diminished reliability of lateral line input in turbulent flows, coupled with the energetic 362 imperative of station-holding, profoundly influences how fish perceive and react to their 363 environment. The drive to conserve energy strongly shapes fish behavioral choices, 364 particularly in complex flow fields, and consequently their reliance on specific sensory cues. 365 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 15 Fish exploit the energetic benefits of station-holding within vortical wakes generated 366 behind bluff bodies (Liao, 2004; Stewart et al., 2016; Taguchi & Liao, 2011). This strategy, 367 termed Karman Gaiting, involves precise positioning to reduce the energetic demands of 368 swimming (Liao et al., 2003a, 2003b). Fish actively return to these energetically favorable 369 regions after displacement, demonstrating a strong motivation to maintain such positions. 370 Optimal station-holding often occurs at a specific saddle point downstream from the obstacle, 371 where flow conditions minimize energy expenditure (Taguchi & Liao, 2011; Zdravkovich, 372 1997). Even a small displacement can lead to ejection into the higher-cost freestream, with 373 oxygen consumption twice as high (Taguchi & Liao, 2011). These substantial energetic costs 374 provide a powerful incentive for fish to actively maintain station in a specific region of the 375 wake; a benefit absent in the viscous regime occupied by larval fish. 376 This drive leads to a marked shift in sensory prioritization. In predictable, uniform 377 laminar flows, where lateral line inputs are reliable, OMR is absent or diminished. This 378 suggests lateral line prioritization due to its direct, rapid, and continuous feedback on the body 379 from the water flow that vision alone cannot replicate with equivalent fidelity or speed. 380 However, when station-holding fish encounter the unpredictable, noisy hydrodynamic cues of 381 a turbulent wake, vision rises to a paramount role. This results in the emergence of a positive 382 OMR, where fish swim in the direction of a moving visual stimulus. For a fish maintaining a 383 fixed position in a turbulent flow-field, such a visual strategy is highly adaptive. Ignoring visual 384 cues would result in costly downstream drift, or risk ejection into the high-cost freestream. 385 Interestingly, stationary near-field visual cues (e.g., the cylinder) proved insufficient to 386 maintain station-holding in turbulence, indicating that broad, wide-field visual stimuli are 387 required for this behavior. 388 Role of the Efferent System and Sensory Conflict 389 A switch to visual reliance in turbulent flows suggests an underlying neural mechanism that 390 actively modulates sensory input. We propose that the efferent system of the lateral line plays 391 a critical role in this dynamic reweighting of sensory information. The efferent system cancels 392 out flow information generated by self-movement (e.g., corollary discharge, (Crapse & 393 Sommer, 2008)), allowing zebrafish to prune afferent information correlated with swimming 394 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 16 motions and the fluid environment (Skandalis et al., 2021). This comparison of expectation 395 and sensing enables the lateral line to convey external proprioceptive abilities in simple flow 396 environments. 397 Building on this, we hypothesize that the visual system is also calibrated into this 398 proprioceptive architecture. Sensing the predictable flow across the body during anticipated 399 undulations suppresses behavioral responses stemming from wide-field visual inputs; hence, 400 fish in steady flows do not react to optical perturbations. When holding station in turbulent 401 flow, the hydrodynamic cues necessary for body positioning during swimming diminish, 402 prompting fish to switch to visual cues to guide their behavior and maintain position. In these 403 conditions, fish in our study show compensatory responses to optical push and pull 404 treatments. 405 Importantly, when station-holding fish follow downstream-drifting bars, they do not turn 406 and swim downstream, nor do they cease swimming and drift passively while still facing 407 upstream. Instead, they maintain an upstream orientation and swim slower, using a reduced 408 tailbeat frequency and amplitude. Active swimming is necessary to activate an efferent motor 409 copy, which also improves control. Swimming, through corollary discharge, could calibrate 410 flow expectations along the body, allowing fish to recognize less turbulent (e.g., more 411 predictable) flows. Such an awareness is harder to execute when swimming downstream, 412 given the anterio-posterior sensitivity bias of neuromasts (Munz, 1985). Through the lateral 413 line efferent/afferent system, fish are continuously in touch with their environment. Undulating 414 the body allows fish to detect changes in turbulence, as exemplified when fish, visually 415 prompted to leave their station-holding region, eventually encounter freestream flow outside 416 the wake. This would be impossible if fish drifted downstream straight-bodied, without the 417 efferent activity to gauge body awareness. 418 Our hypothesis that the efferent system gates visual input is further supported by our 419 sensory-conflict experiments. Our optical-roll experiments reveal that visual input becomes 420 less influential when lateral line input aligns with an expected environmental model. This 421 suggests that continuous hydrodynamic input to the lateral line system overrides sudden 422 visual inputs, causing fish to disregard cross-stream moving bars when flow is moving 423 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 17 downstream (this study, (Bak-Coleman et al., 2015)). In virtual rolling scenarios, the absence 424 of dorsoventral flows typically detected by the lateral line, coupled with semicircular canal 425 inputs that do not align with the rolling movements conveyed by the visual world, further 426 highlights this sensory prioritization. This raises a crucial question: if the prioritization of the 427 lateral line over the visual system is a fundamental strategy for navigating predictable 428 environments, do these principles of context-dependent sensory weighting extend to other 429 fundamental, survival-critical behaviors? To answer this, we next investigated if this sensory 430 flexibility also applies to rapid, ecologically vital behaviors like the escape response, a well-431 characterized behavior fish use to flee predators based on auditory, visual, or lateral line 432 stimuli (Mirjany et al., 2011). 433 Vision-Dependent Escape Responses in Flow and Group Dynamics 434 For over a century, the escape response has been a focal point of research, primarily in 435 individual fish within simplified hydrodynamic environments. While most studies have been 436 conducted in still water, fish in nature often form groups as an anti-predator response (Nadler 437 et al., 2021; Krause & Ruxton, 2002; Magurran, 1990) and frequently inhabit current-swept 438 environments. Less understood, however, is the effect of group dynamics on escape 439 behavior, where individuals must process not only visual threats but also information from 440 moving group members. 441 We observed that fish initiated escape responses more frequently when in closer 442 proximity to a loom stimulus, regardless of whether they were individuals or in a group, and 443 whether they were swimming in flow or still water (Figure-3 A, B). However, the delay in 444 response showed a stronger correlation with stimulus distance when fish were swimming in 445 flow compared to still water (Figure-3 C, D). This positive correlation between distance and 446 response delay suggests that fish are reacting to the angular expansion of the stimulus rather 447 than merely its presence. When accounting for both stimulus distance and response delay, 448 fish in flow exhibited a lower angular threshold for triggering an escape response. This finding 449 aligns with the hypothesis that fish are more sensitive to visual threats while swimming in flow 450 than in still water. 451 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 18 Our results suggest that the lateral line may be sensitized to compensate for potential 452 constraints in escape trajectories within flowing water. Unlike in still water, where escape 453 speeds and energetic costs are relatively equal in all directions, upstream escape paths in 454 flow likely result in slower bursts and higher energetic costs compared to downstream or 455 cross-stream paths (Domenici & Hale, 2019; Kohashi et al., 2012). This implies that faster 456 responses in flow are a prominent component of station-holding behavior. Indeed, faster flows 457 have been shown to elicit faster performance phenotypes in other fish species in the wild 458 (Nadler et al., 2018). 459 The heightened visual sensitivity of zebrafish in challenging hydrodynamic 460 environments demonstrates their adaptive flexibility in sensory processing. However, this 461 behavioral adaptation frequently occurs within a social context, as zebrafish intrinsically 462 associated in groups. This raises further questions about how collective dynamics influence 463 threat perception and response. We found that individuals within a group exhibited a higher 464 angular threshold for escape compared to solitary individuals. This suggests that schooling 465 favors robustness to a stimulus rather than increased sensitivity, a phenomenon also 466 documented in wild fish (Fahimipour et al., 2023). Interestingly, while schooling exposes 467 individuals to unpredictable hydrodynamic stimuli generated by conspecifics, this appears to 468 have a desensitizing effect on vision, a stark contrast to the OMR where individuals become 469 more sensitive to visual threats. 470 Rethinking the Lateral Line: Beyond Simple Flow Sensing 471 To more precisely dissect the independent contributions of the lateral line and vestibular 472 systems to sensing turbulent flow, future experiments should aim to create hydrodynamic 473 conditions that generate unpredictable flow across the body while ensuring that vortex 474 strength and size do not displace the body. This approach would allow for the investigation of 475 unreliable lateral line information concurrent with predictable vestibular input, a distinction not 476 achievable in the present study due to the absence of lateral line ablation experiments. 477 Traditional antibiotic or genetic ablations of the lateral line, by primarily targeting hair 478 cells, selectively remove afferent (incoming) flow information while leaving the efferent system 479 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 19 intact. This suggests that such ablations do not isolate the study of flow sensing alone but 480 instead introduce a more complex scenario that likely involves the unmasking of other 481 sensory modalities, such as body sensing, which complicate behavioral interpretations. 482 Building on this, we predict that fish with a non-functional lateral line, when exposed to 483 uniform flows, will exhibit behaviors analogous to those observed in turbulent flow conditions. 484 In these situations, alterations in the visual wide-field may elicit behavioral responses that 485 would typically be disregarded in a uniform flow environment (e.g., antibiotic studies, (Liao, 486 2006)). This intricate interplay among sensory modalities may represent a fundamental 487 behavioral mechanism that prioritizes robust, rapid signals from the lateral line system, and 488 may have been necessary for animals before advanced visual systems were establish 489 (Šestak et al., 2013). 490

491 In summary, we demonstrate that wide-field visual inputs do not alter the behavior of adult 492 zebrafish swimming in uniform flows. We argue that during uniform flow conditions, where 493 hydrodynamic stimuli can be anticipated and compared to an internal model of movement, 494 fish prioritize flow inputs from the lateral line system and/or the vestibular system over wide-495 field visual stimuli. In contrast, fish holding station in turbulent flows alter their behavior in 496 response to wide-field visual inputs. We reason that the lateral line can no longer reliably 497 predict flow along the body in unsteady flows as it can during uniform swimming. Because 498 these fish may be less certain of their swimming state based on lateral line 499 mechanoreception, vision emerges to play a larger role in directing behavioral responses. The 500 greater energetic consequence of forfeiting position when station-holding behind a bluff body 501 makes it particularly significant that visual inputs are acted upon once lateral line inputs 502 become less predictable. Furthermore, while schooling, fish experience unpredictable 503 hydrodynamic stimuli created by other individuals, yet they exhibit a decreased sensitivity to 504 looming stimuli. Our work supports the idea that the prioritization of sensory modalities, rather 505 than simple integration, is specific to flow environments. The ability to dynamically prioritize 506 sensory inputs underscores the adaptive capacity of fish in complex and challenging 507 environments. These findings emerge from an approach that embraces more complex and 508 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 20 naturalistic experiments, and have important implications for fields ranging from neuroscience, 509 collective behavior, and robotic control. 510 511 512 Contributions 513 Conceptualization: J.C.L., S.D. Methodology: J.C.L., S.D.; Software: S.D. Validation: J.C.L., 514 S.D.; Formal analysis: S.D.; Investigation: J.C.L., S.D.; Resources: J.C.L; Data curation: 515 J.C.L., S.D.; Writing - original draft: J.C.L., S.D.; Writing - review & editing: J.C.L., S.D.; 516 Supervision: J.C.L.; Project administration: J.C.L; Funding acquisition: J.C.L. 517 518

519 We would like to thank Matteo Adorisio for help with experiments and preliminary analysis, 520 and Eileen Hamilton for fish care. All protocols were approved by the University of Santa 521 Barbara Institutional Animal Care and Use Committee. This research was supported in part by 522 NSF Grant No. PHY-1748958, NIH Grant No. R25GM067110, Gordon and Betty Moore 523 Foundation Grant No. 2919.01, and the Kavli Foundation. This research was additionally 524 supported by the National Science Foundation Grants IOS 1257150, 2321275, and 1856237; 525 NSF MPS/PHY 2102891 and NSF ENG/CMMI 2345913 to J. Liao, and National Institute on 526 Deafness and Other Communication Disorders Grant R56DC020321 to J. Liao. The authors 527 declare no competing interests. 528 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 21 Figures 529 530 Figure 1 – Schematic of experimental setup with the fish swimming flow-tank and visual 531 stimulus projection. (A) A variable-speed flow tank placed on a stand with a filming-mirror 532 placed underneath tilted at 45 angle to film the ventral view of the swimming fish. (B) Cross-533 sectional view of flow tank with two side-mirrors and a top-mirror used for visual stimulus 534 projection, and filming mirror schematic. (C) Schematic of optical stimuli and their motion 535 direction, where the patterns move simultaneously in the direction of sold arrows, and then for 536 a separate treatment, move simultaneously in dashed-arrows directions. Vertical gratings for 537 “optical push/pull” and horizontal gratings for “optical roll” treatments and the expanding loom 538 stimulus on the top wall for escape trials. (D) Perceived threshold stimulus angle (θ) is 539 calculated based on the 3D position of the escaping fish and the instantaneous radius of the 540 loom stimulus. 541 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 22 542 543 Figure 2 – Quantifying station-holding fish’ response to wide-field optical perturbations in 544 steady water-flow (A-B) and unsteady flow(C-D). (A,C) “Optical pull” stimulus entails visual 545 patterns suddenly moving downstream at the stimulus-onset (vertical dashed-line). Top-panel: 546 Fish’s longitudinal positions (along the streamwise direction) is plotted against time, where 547 negative position values show downstream drift for “optical pull” (visual patterns moving 548 downstream) and each color represents one individual fish (n=4 fish, 4 trials each); Bottom-549 panel: Comparing the change in fish’s swimming velocity and Spearman’s correlation co-550 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 23 efficient (representing degree of monotonous movements) before and after the stimulus 551 onset. (B,D) Same plots as in part-A but for the “optical push” stimulus, where visual patterns 552 move upstream at the onset (n=4 fish, 4 trials each). 553 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 24 554 Figure 3 – Comparison of escape response (C-start) due to a purely visual looming stimulus 555 during “flow” (swimming against water current) and “no-flow” (still water) conditions in 556 individual (“Single”) and school (“Group”) of fish. (A) Presence or absence of escape trigger 557 (C-start) is classified for “flow” (top) and “no-flow” (bottom) conditions, and plotted against 558 fish’s swimming distance from the stimulus center point. (B) Similar to part-A but now 559 projecting the looming stimulus on a different part (i.e., towards the downstream end of the 560 flow-tank). (C-D) For the escaping fish, their swimming distance and response delay are 561 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 25 compared between “flow” and “no-flow” conditions. A higher slope indicates stronger 562 correlation between those quantities. (E) Perceived angle of the threshold stimuli (θ), which is 563 found using fish’s 3D position and instantaneous radius of looming stimulus, for all trials are 564 plotted on a representative looming stimulus (Dashed line) that indicates the angular 565 expansion of the looming stimulus over time. (F) The same threshold stimulus angles for all 566 the fish are compared between “flow” and “no-flow” trials. A lower angle represents quicker 567 escape. 568 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 26 Supplementary Materials 569 570 Supp. Figure 1 – Raw trajectory data and velocity data for lateral movements in optical 571 pull/push experiments for steady and unsteady flow (raw data for longitudinal movements, 572 and plot legends are as in Fig-2). 573 574 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 27 575 576 Supp. Figure 2 – Raw data for optical roll perturbations (CW & CCW directions) for 577 longitudinal and lateral direction movements. Plot legends are similar to explained in Fig-2 (A-578 B)) 579 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 28 Supp. Figure 3 – Images of the experiment setup. (A) Flow tank, Projection and Recording 580 setup. (B) Example visual stimulus projected on a side-wall of swimming chamber. (C) A 581 section of an image acquired by the camera of a group of fish swimming with an expanding 582 looming stimulus in the background and the first fish (from upstream side) triggering a C-start 583 reflex. 584 (A) (C) (B) was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 29 Table S1 : Statistical summary of lateral-only motion in response to visual perturbation 585 Lateral-Only Side Velocity mean ± s.e.m (cm/s) P-val Side rho mean ± s.e.m P-val Pre- Stimulus Post- Stimulus Pre vs Post Pre- Stimulus Post- Stimulus Pre vs Post Optical-Pull (Steady) 4.2E-04 ± 3.5E-03 1.5E-03 ± 2.0E-03 0.98 (n.s.) 3.2E-02± 1.8E-01 1.0E-01 ± 1.5E-01 0.82 (n.s.) Optical-Push (Steady) -3.7E-04 ± 2.9E-03 -2.2E-03 ± 1.7E-03 0.63 (n.s.) -1.1E-01 ± 1.7E-01 -2.3E-01 ± 1.6E-01 0.60 (n.s.) Optical-Pull (Unteady) 3.6E-04 ± 1.8E-03 2.3E-03 ± 2.3E-03 0.67 (n.s.) 1.5E-02 ± 1.7E-01 8.0E-02 ± 1.7E-01 0.86 (n.s.) Optical-Push (Unsteady) -1.6E-04 ± 2.5E-03 -1.6E-03 ± 3.8E-03 0.74 (n.s.) 2.0E-01 ± 1.7E-01 2.1E-02 ± 1.9E-01 0.46 (n.s.) Optical-Roll (cw) -2.5E-03 ± 4.2E-03 -1.8E-03 ± 3.7E-03 0.98 (n.s.) -2.6E-02 ± 1.7E-01 6.0E-03 ± 1.7E-01 0.90 (n.s.) Optical-Roll (ccw) 1.2E-03 ± 3.6E-03 -3.3E-03 ± 2.6E-03 0.74 (n.s.) 1.4E-02 ± 1.9E-01 -1.5E-01 ± 1.6E-01 0.86 (n.s.) 586 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 30

587 Bak-Coleman, J., Smith, D., & Coombs, S. (2015). Going with, then against the flow: 588 Evidence against the optomotor hypothesis of fish rheotaxis. Animal Behaviour, 107, 589 7–17. https://doi.org/10.1016/j.anbehav.2015.06.007 590 Chagnaud, B. P., Bleckmann, H., & Hofmann, M. H. (2007). Kármán vortex street detection by 591 the lateral line. Journal of Comparative Physiology. A, Neuroethology, Sensory, Neural, 592 and Behavioral Physiology, 193(7), 753–763. https://doi.org/10.1007/s00359-007-593 0230-1 594 Crapse, T. B., & Sommer, M. A. (2008). Corollary discharge across the animal kingdom. 595 Nature Reviews. Neuroscience, 9(8), 587–600. https://doi.org/10.1038/nrn2457 596 Dallmann, C. J., Dickerson, B. H., Simpson, J. H., Wyart, C., & Jayaram, K. (2023). 597 Mechanosensory Control of Locomotion in Animals and Robots: Moving Forward. 598 Integrative And Comparative Biology, 63(2), 450–463. 599 https://doi.org/10.1093/icb/icad057 600 Domenici, P., & Hale, M. E. (2019). Escape responses of fish: A review of the diversity in 601 motor control, kinematics and behaviour. Journal of Experimental Biology, 222(18), 602 jeb166009. https://doi.org/10.1242/jeb.166009 603 Fahimipour, A. K., Gil, M. A., Celis, M. R., Hein, G. F., Martin, B. T., & Hein, A. M. (2023). Wild 604 animals suppress the spread of socially transmitted misinformation. Proceedings of the 605 National Academy of Sciences, 120(14), e2215428120. 606 https://doi.org/10.1073/pnas.2215428120 607 Gray, J. (1968). Animal locomotion [by] James Gray. In Animal locomotion. Weidenfeld & 608 Nicolson. 609 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 31 Kohashi, T., Nakata, N., & Oda, Y. (2012). Effective Sensory Modality Activating an Escape 610 Triggering Neuron Switches during Early Development in Zebrafish. The Journal of 611 Neuroscience, 32(17), 5810–5820. https://doi.org/10.1523/JNEUROSCI.6169-11.2012 612 Krause, J., & Ruxton, G. D. (2002). Living in Groups. Oxford University PressOxford. 613 https://doi.org/10.1093/oso/9780198508175.001.0001 614 Liao, J. C. (2004). Neuromuscular control of trout swimming in a vortex street: Implications for 615 energy economy during the Karman gait. The Journal of Experimental Biology, 207(Pt 616 20), 3495–3506. https://doi.org/10.1242/jeb.01125 617 Liao, J. C. (2006). The role of the lateral line and vision on body kinematics and hydrodynamic 618 preference of rainbow trout in turbulent flow. Journal of Experimental Biology, 209(20), 619 4077–4090. https://doi.org/10.1242/jeb.02487 620 Liao, J. C., Beal, D. N., Lauder, G. V., & Triantafyllou, M. S. (2003a). Fish Exploiting Vortices 621 Decrease Muscle Activity. Science, 302(5650), 1566–1569. 622 https://doi.org/10.1126/science.1088295 623 Liao, J. C., Beal, D. N., Lauder, G. V., & Triantafyllou, M. S. (2003b). The Kármán gait: Novel 624 body kinematics of rainbow trout swimming in a vortex street. Journal of Experimental 625 Biology, 206(6), 1059–1073. https://doi.org/10.1242/jeb.00209 626 Magurran, A. E. (1990). The adaptive significance of schooling as an anti-predator defence in 627 fish. Annales Zoologici Fennici, 27(2), 51–66. 628 Mathis, A., Mamidanna, P., Cury, K. M., Abe, T., Murthy, V. N., Mathis, M. W., & Bethge, M. 629 (2018). DeepLabCut: Markerless pose estimation of user-defined body parts with deep 630 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 32 learning. Nature Neuroscience, 21(9), Article 9. https://doi.org/10.1038/s41593-018-631 0209-y 632 Mirjany, M., Preuss, T., & Faber, D. S. (2011). Role of the lateral line mechanosensory 633 system in directionality of goldfish auditory evoked escape response. Journal of 634 Experimental Biology, 214(20), 3358–3367. https://doi.org/10.1242/jeb.052894 635 Montgomery, J., Carton, G., & Bodznick, D. (2002). Error-Driven Motor Learning in Fish. The 636 Biological Bulletin, 203(2), 238–239. https://doi.org/10.2307/1543417 637 Munz, H. (1985). Single unit activity in the peripheral lateral line system of the cichlid 638 fishSarotherodon niloticus L. Journal of Comparative Physiology A, 157(5), 555–568. 639 https://doi.org/10.1007/BF01351350 640 Nadler, L. E., Killen, S. S., Domenici, P., & McCormick, M. I. (2018). Role of water flow regime 641 in the swimming behaviour and escape performance of a schooling fish. Biology Open, 642 bio.031997. https://doi.org/10.1242/bio.031997 643 Nadler, L. E., McCormick, M. I., Johansen, J. L., & Domenici, P. (2021). Social familiarity 644 improves fast-start escape performance in schooling fish. Communications Biology, 645 4(1), 897. https://doi.org/10.1038/s42003-021-02407-4 646 Olive, R., Wolf, S., Dubreuil, A., Bormuth, V., Debrégeas, G., & Candelier, R. (2016). 647 Rheotaxis of Larval Zebrafish: Behavioral Study of a Multi-Sensory Process. Frontiers 648 in Systems Neuroscience, 10. https://doi.org/10.3389/fnsys.2016.00014 649 Pita, D., Moore, B. A., Tyrrell, L. P., & Fernández-Juricic, E. (2015). Vision in two cyprinid fish: 650 Implications for collective behavior. PeerJ, 3, e1113. https://doi.org/10.7717/peerj.1113 651 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 33 Šestak, M. S., Božičević, V., Bakarić, R., Dunjko, V., & Domazet-Lošo, T. (2013). 652 Phylostratigraphic profiles reveal a deep evolutionary history of the vertebrate head 653 sensory systems. Frontiers in Zoology, 10(1), 18. https://doi.org/10.1186/1742-9994-654 10-18 655 Sharma, V. P., & Sponberg, S. (2023). Context dependent multisensory integration: 656 Mechanosensation depends on luminance for robust performance. Neuroscience. 657 https://doi.org/10.1101/2023.09.29.560224 658 Skandalis, D. A., Lunsford, E. T., & Liao, J. C. (2021). Corollary discharge enables 659 proprioception from lateral line sensory feedback. PLOS Biology, 19(10), e3001420. 660 https://doi.org/10.1371/journal.pbio.3001420 661 Stewart, W. J., Tian, F.-B., Akanyeti, O., Walker, C. J., & Liao, J. C. (2016). Refuging rainbow 662 trout selectively exploit flows behind tandem cylinders. The Journal of Experimental 663 Biology, 219(Pt 14), 2182–2191. https://doi.org/10.1242/jeb.140475 664 Sutterlin, A. M., & Waddy, S. (1975). Possible Role of the Posterior Lateral Line in Obstacle 665 Entrainment by Brook Trout ( Salvelinus fontinalis ). Journal of the Fisheries Research 666 Board of Canada, 32(12), 2441–2446. https://doi.org/10.1139/f75-281 667 Taguchi, M., & Liao, J. C. (2011). Rainbow trout consume less oxygen in turbulence: The 668 energetics of swimming behaviors at different speeds. Journal of Experimental Biology, 669 214(9), 1428–1436. https://doi.org/10.1242/jeb.052027 670 Tritico, H. M., & Cotel, A. J. (2010). The effects of turbulent eddies on the stability and critical 671 swimming speed of creek chub ( Semotilus atromaculatus ). Journal of Experimental 672 Biology, 213(13), 2284–2293. https://doi.org/10.1242/jeb.041806 673 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint 34 Zdravkovich, M. M. (1997). Flow Around Circular Cylinders: A Comprehensive Guide Through 674 Flow Phenomena, Experiments, Applications, Mathematical Models, and Computer 675 Simulations. Oxford University PressOxford. 676 https://doi.org/10.1093/oso/9780198563969.001.0001 677 678 was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted April 1, 2026. ; https://doi.org/10.64898/2026.03.30.715425doi: bioRxiv preprint