Fig 1.
Experimental apparatus and analysis techniques.
(A) Box-shaped flight arena in which lovebirds performed a U-turn flight maneuver starting and ending at the perch. The maneuver was filmed in stereo with two high-speed cameras at 2000 fps. (B) Illustration of the four time points within a wing beat in which we assessed the head and body orientation in “low resolution” with respect to wing beat phase. Red dots depict marker points on the head used to obtain its position and yaw orientation. Blue dots depict tracked shoulder positions used to obtain the yaw orientation of the body. (C) Schematic camera view into the arena. Note that due to the indirect view via the mirror above the arena, camera images were mirrored around the horizontal axis. To avoid confusion, we will refer to the flight scene as seen from the camera perspective. Red and blue lines show how yaw orientations of the head and body were obtained relative to a horizontal line in the image.
Fig 2.
Individual example of a turning on a dime maneuver.
(A) Head trajectory during a leftward U-turn flight maneuver in the flight arena. Plotted head position (gray dots) and head yaw orientation (red lines) are presented in 25 ms steps (50 frames). Because the turn is performed on a dime, traces after takeoff and before landing overlap. Arena wall grayscale (black, gray, and white) illustrates the inner wall texture seen by the bird. The thick black bar in the arena shows the perch position. (B) Top view of all 15 wing-tip positions at mid stroke during the same turn maneuver as shown in (A). Deflections in wing tip traces (wing beat 5–6 & 11–12) were used to separate the continuous flight maneuver into three consecutive phases: before turn (gray), during turn (blue) and after turn (red). (C) Head (red) and body (blue) yaw orientation angle (Φ) during a left turn maneuver. Φ is calculated in relation to a horizontal axis in the flight arena (see Fig 1 and methods). Φ values of 0° indicate an orientation along the horizontal axis facing away from the perch whereas Φ of 180° indicates an orientation along this axis facing towards the perch. Positive Φ deflections indicate a turn to the left, negative ones a turn to the right. The Φ difference angle between head and body is shown in green. Body data was filtered and interpolated (blue dashed line, see methods) to calculate the Φ head-body angle. Vertical bars represent the downstroke phases and thereby the wing beat timing before, during and after the turn. (D) Saccade detection (black, see text) based on absolute yaw rotation velocity (ω, gray graph). The dashed gray line illustrates the saccade detection threshold of 400°/s that ω had to exceed for at least 12 ms (24 data points) to define a head turn as saccadic.
Fig 3.
Lovebirds performed an intermittent flight style with two wingbeat distributions in which saccades are started during downstrokes.
(A) The downstroke / upstroke phase ratio vs. instantaneous flap frequency distribution for individual wingbeats of five birds. A phase ratio of 1 indicates up- and downstrokes of equal duration, values <1 indicate longer upstrokes, values >1 longer downstrokes. Normalized bimodal Gaussian fits are shown for flap frequency (top) and for downstroke / upstroke time ratios (right). The bird-specific bimodal distribution parameters for the flapping frequency are: 2dg: μ1 = 9.78, σ1 = 1.61, μ2 = 17.26, σ2 = 1.01; 2lg: μ1 = 10.26, σ1 = 1.83, μ2 = 18.14, σ2 = 0.91; 2y: μ1 = 9.39, σ1 = 1.1, μ2 = 17.19, σ2 = 0.86; 1y: μ1 = 8.97, σ1 = 0.6, μ2 = 15.76, σ2 = 0.96; 3g: μ1 = 9.49, σ1 = 2.3, μ2 = 16.72, σ2 = 1; For downstroke / upstroke periods the obtained bimodal distribution parameters are: 2dg: μ1 = 0.5, σ1 = 0.07, μ2 = 1.26, σ2 = 0.27; 2lg: μ1 = 0.56, σ1 = 0.13, μ2 = 1.43, σ2 = 0.17; 2y: μ1 = 0.48, σ1 = 0.07, μ2 = 1.26, σ2 = 0.27; 1y: μ1 = 0.62, σ1 = 0.09, μ2 = 1.49, σ2 = 0.17; 3g: μ1 = 0.48, σ1 = 0.12, μ2 = 1.3, σ2 = 0.17. The horizontal gray line separates the bimodal distributions at a downstroke / upstroke ratio of 0.94 (average midpoint between bimodal distribution peaks among birds). The vertical gray line separates the bimodal distribution at a flap frequency of 13.3 Hz (average among birds); n = 697 wing beats, N = 5 birds. Due to the 2000 fps sample frequency, and the fact that wingbeat, downstroke, and upstroke time are all integer values measured in number of frames, the data appear in a raster and can overlap precisely among wings beats, flights and birds. (B) The normalized saccade distributions illustrate when a saccade was started and ended during the downstroke vs. the upstroke phase. Shown is the average across birds (solid lines) and the standard deviation (shaded area). Binning: 0:10:100; n = 72 saccades, N = 5 birds.
Fig 4.
Head saccades occur predominantly during the turn and their speed compares to insects.
(A) Flight traces of all birds depicting the position of saccade initiations during a U-turn. Turning flight traces are shown as grey lines, saccades as red colored symbols: triangles depict saccades made before or after the turning phase, cycles illustrate saccades during the turning phase. The dashed red line in the color bar represents the saccade detection threshold of 400°/s. The perch position is represented by the vertical black bar. A rare right turn flight of bird 2y is illustrated for representative reasons by the thicker line, n = 16 flights, N = 5 birds. Panels (B-F) illustrate differing head kinematics during intersaccades and saccades as well as the extraordinarily fast nature of lovebird head saccades. The shaded areas illustrate the standard deviation between birds. (B) Amplitudes of horizontal head saccades. Shown are the normalized absolute saccade amplitude distributions relative to a horizontal axis through the flight arena and the normalized relative saccade amplitude relative to the body yaw orientation. Binning = 0:10:60. (C) Normalized average of saccade duration across birds. Binning = 0:10:50. (D) Normalized average head yaw rotation speed in space for saccades and phases between saccades (intersaccades). Binning = 0:100:2000. (E) Normalized average head yaw rotation speed relative to the turning body, plotted for both saccades and intersaccades. Binning = 0:100:2000 (see text for definition). (F) Peak saccadic rotation speeds of the head. To compare the performance of lovebirds to other flying animals showing similar saccadic gaze behaviors, we inserted reported saccadic peak rotation speeds of the honey bee [28], zebra finch [3], the blowfly [27] and the fruitfly [29]. Binning = 0:300:3000. n = 72 saccades, N = 5. Saccade analysis is based on video data resolved at 2000 Hz.
Fig 5.
Lovebird head rotation is predominantly saccadic at the highest average speed recorded amongst vertebrates.
(A) Comparison of horizontal saccade amplitude as a function of its duration. Shown are saccade amplitudes and durations measured in this study (blue triangles, n = 72, N = 5) and data for other species extracted from earlier publications. Gray triangle markers represent the combined eye-head gaze shifts of rhesus macaques (n = 544, N = 2) [30]. Red, green and violet circles illustrate horizontal eye saccades in humans (n = 187, N = 3) [31], rabbits (n = 191, N = 2) [33] and cats (n = 34, N = 2) [32]. We coarsely approximate average head rotation velocities by fitting the data with a linear regression. Line equations: lovebird: y = 1.5*x-14, slope = 1500°/s, R2 = 0.78; rhesus macaque y = 0.4*x+3.4, slope = 400°/s, R2 = 0.69; human: 0.38*x+3.4, slope = 380°/s, R2 = 0.93; rabbit: y = 0.33*x-10, slope = 260°/s, R2 = 0.67, cat: y = 0.11*x-4.2, slope = 110°/s, R2 = 0.81. (B) Proportion of saccadic turns on the whole U-turn maneuver. Shown is the average cumulative saccade amplitude (and standard deviation) as a percentage of the whole turn amplitude, with the red line showing the average across birds (n = 72 saccades, N = 5 birds).
Fig 6.
Distributions of intersaccadic azimuthal feature positions reveal that maneuvering birds stabilize arena features in their frontal visual field.
(A) Schematic top view into the arena. Azimuthal positions of wall corners, the gray square and the perch (colored lines) were obtained relative to the bird’s head yaw orientation (red line). (B) Relative azimuthal angles of arena features for the example flight shown in Fig 2. The thick green line depicts the angle of the perch center. The gray shaded area represents feasible horizontal eye motions of ±10° relative to the horizontal head orientation. Positive azimuthal angles represent the left visual hemisphere, negative values the right visual hemisphere (see bird illustration above legend). Note that the diverging azimuthal angles of the perch edges (dark green and blue lines) are caused by the bird getting closer to the perch. Approaching the perch causes the retinal size of the perch to expand in the bird’s frontal visual field. (C-F) Averaged relative azimuthal distributions of arena features that were stabilized in the frontal visual field in the intersaccadic phases during a turning on a dime maneuver; before the turn (C: fine dashed line), during the turn (D & E: solid line) and after turning (F: coarse dashed line). Averaged distributions illustrate left turn flights (n = 92, N = 5). Standard deviations across birds are illustrated by the colored areas. (G-K) Intersaccadic azimuthal distributions for arena features that were not stabilized in the frontal visual field. By stabilizing the perch center frontally after the turn (F), the right and left edge distributions are positioned more laterally, are broader and have lower peaks than the center (E & K). The vertical bar extending ±10° illustrates feasible horizontal eye motions relative to the head orientation in unrestricted birds (review: [14]). Perch position for 92 flights is approximated by using the position of a static perch (thick lines in panel E-F, n = 92 flights, N = 5). For 15 left-turn-flights tracked at 2000 Hz, the position of the swinging perch relative to the birds was tracked as well (gray dashed line in panel E, F & K n = 15 flights, N = 5). For normalization, we divided each distribution by the cumulative sum of all other feature distributions. Binning = -90:10:90.
Fig 7.
Low variation of retinal size and relative expansion velocity (RREV) of the approached perch suggests these cues matter for controlled landings on a swinging perch.
We defined the tail pitch as the behavioral indicator for landing initiation (time = 0 ms). Negative time values represent the time before and positive time values the time after the downward pitch of the tail feathers. Shaded areas ranging from -30 ms to 0 ms mark the minimal time period of visuomotor delay during which visual flight control is unlikely. Absolute horizontal flight speed (A) has less variation across flights than relative horizontal flight speed (B) with respect to the moving perch. (C) The most parsimonious landing parameters are indicated by a minimum in the coefficient of variation (c.v.) across flights and birds. The retinal size (orange) and RREV (green) for the approached perch varied less that the parameter tau and the retinal expansion. Tail pitch timing was extracted individually from high-speed flight videos. n = 16, N = 5 birds.
Fig 8.
Lovebirds improve visual flight control by coordinating super-fast gaze shifts with the end of their downstroke.
(A) Top view schematic of a typical recording that shows that the wings occlude the lateral visual field at the end of the downstroke. The lovebird’s azimuthal visual field is approximated from ophthalmologic measurements at the visual equator of Senegal parrots [57]. (B) For demonstrative purposes a lovebird was filmed from the side during a turning on a dime maneuver with the side panels of the arena removed (this video sequence is not part of the data analysis). (C) Most saccades were initiated at 75% of the downstroke (see Fig 3B), when the wings occlude more than half of the lateral visual field.