Comparative study of optomotor response (OMR) in zebrafish and medaka larvae.

(a) Zebrafish and medaka species comparison. Adult fish (left; top: medaka, bottom: zebrafish) and schematic drawings are shown. Scale bar: 1 cm. (b) Larval specimens used in this study. Larvae at 5-6 days post-fertilization (dpf) for zebrafish and 1 day post-hatch (dph) for medaka were examined. Scale bar: 3 mm. (c) Free-swimming behavioral experiment rig. Individual larvae were placed in a 6 cm-radius circular arena. A high-speed camera above the dish took the whole image of the behavioral rig at 90 fps. The center of mass position (x, y) and head angle (hA) of the focal larvae are recorded. Visual stimuli were generated based on these parameters and projected from beneath the arena. (d) Swimming dynamics of zebrafish (left) and medaka (right) larvae. Zebrafish exhibit discrete swimming bouts alternating with interbout periods, whereas medaka demonstrate continuous undulatory swimming with rhythmic acceleration phases. Turn angles were quantified as the deviation in head angle before and after acceleration time points, as illustrated in the bottom right inset of (c). (e) Optomotor response to whole-field sine-grating stimuli (top left). The red trace represents the temporal stimulus pattern (bottom left), where positive values indicate leftward motion and negative values indicate rightward motion. Each trial consisted of 5 seconds open-loop converging sine-grating baseline, followed by 20 seconds closed-loop sine-grating stimuli, and 5 seconds static grating. Turn angles for all swimming bouts are plotted (right, top row; blue: zebrafish, orange: medaka), with averaged responses with shades of the standard errors of the means (SEMs) shown below (right, bottom row) (zebrafish: n=53, medaka: n=42). (f) Optomotor response to whole-field random dot motion stimuli (top left). Protocol identical to (e) 5 seconds open-loop converging sine-grating baseline, followed by 20 seconds closed-loop random dot motion stimuli (bottom left). Individual bout turn angles (top row) and averaged responses with shades of SEMs (bottom row) are shown (zebrafish: n=21, medaka: n=32).

Spatial integration of the whole-field motion stimuli in teleost larvae.

(a) Schematic of experimental arena. Turns were categorized based on distance from the wall where each turn occurred. (b) Sine-grating stimulus configurations: whole-field stimuli (left) or restricted area around the fish (right, 3 cm radius). (c) Turn responses at different distances from the wall. Raw angular data (top) and averaged turn angles across spatial bins (bottom) during trials with 5 s open-loop converging stimuli followed by 20 s closed-loop stimuli. Top rows: whole-field sine-grating stimuli (solid frames) (zebrafish n=42, medaka n=53). Bottom rows: restricted-area stimuli (dotted frames) (zebrafish n=33, medaka n=31). Blue: zebrafish, orange: medaka. (d) Averaged turn angles during closed-loop stimuli from (c). Solid lines: whole-field stimuli; dotted lines: restricted stimuli. Green dotted lines indicate the values of the responses at the wall-nearest area in the whole-rig stimuli (longer dot) and restricted stimuli (shorter dot) condition. Mean turning angles (± SEM) across six spatial bins from wall (bin 0) to center (bin 50). Statistical significance determined by Bonferroni-corrected t-tests following significant stimulus × spatial bin interaction in two-way mixed ANOVA (*p < 0.05, p < 0.01, p < 0.001). Orange arrowhead indicates persistent wall effect. (e) “Wall effects” calculated as deviation between each area and wall-nearest area. With whole-field stimuli, wall effects exist significantly in both species; with restricted stimuli, only medaka show significant wall effects. (f) Moving dots presented in disks of varying radii beneath fish, with stationary dots in surrounding areas. Visual occupying angle (αdisk) estimated for fish swimming 5 mm above sandblasted dish bottom (See Supplementary figure 2d). (g) Turn responses to different motion disk sizes. Raw angular data (top) and averaged angles (bottom) (zebrafish n=10, medaka n=12). (h) Averaged turn angles during 5-25 s across motion disk sizes. Shaded areas: standard errors. Dotted lines: fitted curves based on motion energy calculations. (i) Motion energy calculation for larvae at height h. Each dot at position (x, y) occupies a visual field angle (θ, φ). For x-axis motion, angle changes depend on position (left: dθ/dt, middle: dφ/dt), and motion energy (dθ/dt × dφ/dt) varies by position (right). (j) Larger disks produce greater turning angles, but zebrafish reach 95% maximum response at smaller sizes than medaka (zebrafish: 17.8 ± 4.8 degrees, medaka: 32.8 ± 12.9 degrees; Welch’s t-test, p=0.0044). (k) Inconsistent motion stimuli with opposing directions in inner and outer regions. Response measured across different inner disk angular sizes. (l) Turn responses to varying inner disk sizes. Raw bout data (top) and averaged angles (bottom) (zebrafish n=6, medaka n=11). (m) Average turn responses versus inner disk size. Black dotted line indicates “effect cancel size” where inner and outer motions balance. (n) Effect cancellation size comparison. Medaka showed significantly larger cancellation angular size than zebrafish (zebrafish: 27.0 ± 1.2 degrees, medaka: 36.6 ± 2.0 degrees; Student’s t-test, p=0.00099).

Comparative study in temporal domain in OMR using flickering dot motion.

(a) Flickering dot motion stimuli presented to free-swimming larvae in a restricted area beneath the fish (left). Dots have defined lifetimes and random spatial locations (spatiotemporal patterns shown on right). (b) Turn responses to varying dot lifetimes during trials. Raw bout data (top) and averaged turn angles (bottom). Blue: zebrafish (n=7), orange: medaka (n=12). (c) Average turn angle increases more slowly in medaka (orange) than zebrafish (blue). Solid lines: experimental data. (d) The time constant of the fitting curve was shown. The values of each fish are shown dots and the averages are shown in horizontal lines. The time constants were significantly larger in medaka than zebrafish (zebrafish: 0.09 ± 0.01 seconds, medaka: 2.26 ± 0.23 seconds (Mean ± STD), Welch’s t-test, p=0.000001). (e) Schematic model of how visual stimuli could be processed in zebrafish (left) and medaka (right) larvae. In zebrafish, we hypothesized that the motion stimuli is averaged globally in a big receptive field (blue circle) and the top part of the retina. On the other hand, in medaka, we hypothesized that the motion stimuli is processed in a local receptive field which is tiled up in a bigger retina and temporally filtered and passed to the averaging step.

Divergent responses to unidirectional motion pulse stimuli reveal species-specific temporal dynamics.

(a) Turn dynamics to unidirectional pulse stimuli. Trials begin with 5 s converging stimuli (black dotted lines), followed by dot-motion pulse stimuli (red trace, top). Individual (middle) and averaged (bottom) turn angles for zebrafish (blue, n=52) and medaka (orange, n=21; shaded areas: standard error). (b) Single-cycle stimulus-response analysis. One complete cycle is shown for zebrafish (left) and medaka (right). Dot colors correspond to temporal bins. (c) Normalized pulse responses and kinetics. Time zero in the ‘Rise’ plot indicates when the motion started, while time zero in the ‘Decay’ indicates when the motion stopped. Rise time constants: no species difference (zebrafish: 0.48 ± 0.02 seconds, medaka: 0.55 ± 0.06 seconds; p=0.52, n.s., mean ± SEM; Mann-Whitney U test). Decay time constants: significantly longer in medaka (zebrafish: 0.41 ± 0.02 seconds, medaka: 1.24 ± 0.14 seconds, mean ± SEM; p=4.0×10⁻¹⁰, Mann-Whitney U test). Blue: zebrafish, orange: medaka. Thin lines represent temporal dynamics of normalized average turn angles for individual fish; thick lines represent population averages. Inset dot plots show individual fish time constants for rise and decay phases. (d) Three-component Gaussian decomposition: large leftward (blue), large rightward (red), and small turns (green). Temporal dynamics in 100 ms bins with fitted curves overlaid. (e) Temporal evolution of behavioral components. Large-turn amplitudes (top), small-turn means (bottom), and large-turn asymmetry (left minus right amplitude, middle, purple). (f) Species comparison of temporal dynamics. Time constants (mean ± SEM): zebrafish big turns (rise: 0.302 ± 0.004 seconds, decay: 0.384 ± 0.037 seconds), zebrafish small turns (rise: 0.701 ± 0.04 seconds, decay: 1.130 ± 0.04 seconds, medaka big turns (rise: 0.363 ± 0.014 seconds, decay: 0.684 ± 0.141 seconds), medaka small turns (rise: 0.894 ± 0.096 seconds, decay: 2.041 ± 0.059 seconds). Zebrafish n=53, medaka n=20. Statistical comparisons (Mann-Whitney U test): big turn rise p=0.93, big turn decay *p=0.013, small turn rise p=0.13, small turn decay **p=0.0067 (asterisks denote *p<0.05, **p<0.01).

Optomotor response to the alternating stimuli.

(a) Turn dynamics to alternating stimuli. Each trial begins with converging stimuli for 5 seconds (indicated by black dotted lines), followed by presentation of moving dots that alternate between leftward and rightward motion with velocities following a sinusoidal pattern (red, top row). (b) Frequency-response characterization. Average turn angles plotted with colored dots indicating time bins within cycles and corresponding responses. Input and response frequencies show a linear relationship. (c) Response amplitude analysis. Amplitudes of fitted sine waves plotted versus stimulus frequency. (d) Turn angle distribution analysis within cycles. Fitted curves for 200 ms bins during one complete cycle (2π) for zebrafish (left) and medaka (right). Line colors correspond to temporal positions. Three Gaussian components are fitted: large rightward (red), large leftward (blue), and small turns (green) (top right). Amplitudes of large-turn distributions (third row) and means of small-turn distributions (bottom, green) across cycles. Large-turn asymmetry dynamics (left minus right amplitude, fourth row, purple) and small turns fitted with sine waves (dotted pink: large turns, dotted yellow: small turns). (e) Frequency-domain analysis of behavioral components. Following sine wave fitting, amplitude and phase of large and small turn components plotted to characterize frequency-amplitude and frequency-phase response profiles. Shaded regions: confidence intervals (blue: zebrafish, orange: medaka).

Temporal integration model captures species-specific optomotor responses

(a) Model architecture and fitting. Schematic shows sensorimotor control for zebrafish (blue) and medaka (orange). Visual input drives two pathways: proportion control (big turn probabilities) and bias control (small turn direction). Each implements first-order dynamics with gain (g0) and decay (g1), followed by saturation. Components shown: big turns left (blue), right (red), difference (purple), and small turn bias (green). Fitted parameters displayed above plots. Dashed lines: model predictions; dots: experimental data. Average angles (right) emerge from component combinations. Bar graphs show R² fitting quality. (b) Parameter sensitivity analysis. Model responses across parameter ranges (proportion gain g₀, decay g₁, bias gain fg₀, decay fg₁, saturation). Top: temporal dynamics with line shading corresponding to parameter values (x-axes, bottom). Bottom: rise and decay time constants versus parameters. Horizontal lines: experimental values (blue: zebrafish, orange: medaka). (c) Species-specific parameter estimates. Fitted values reveal distinct temporal processing strategies between species. (d) Simulated dynamics for alternating stimuli using parameters from (c). Experimental data from Figure 5 (shaded lines) compared to simulations (darker lines). (e) Frequency domain validation. Model predictions (dashed) using parameters from (c). Amplitude and phase responses versus frequency for big turn difference and small turn components. Experimental data (shaded: mean ± CI) validate predictions, revealing species differences: zebrafish show linear amplitude decay while medaka exhibits sharp cutoff around 0.5 Hz.