Pigeons trade efficiency for stability in response to level of challenge during confined flight

Abstract

Individuals traversing challenging obstacles are faced with a decision: they can adopt traversal strategies that minimally disrupt their normal locomotion patterns or they can adopt strategies that substantially alter their gait, conferring new advantages and disadvantages. We flew pigeons (Columba livia) through an array of vertical obstacles in a flight arena, presenting them with this choice. The pigeons selected either a strategy involving only a slight pause in the normal wing beat cycle, or a wings-folded posture granting reduced efficiency but greater stability should a misjudgment lead to collision. The more stable but less efficient flight strategy was not used to traverse easy obstacles with wide gaps for passage but came to dominate the postures used as obstacle challenge increased with narrower gaps and there was a greater chance of a collision. These results indicate that birds weigh potential obstacle negotiation strategies and estimate task difficulty during locomotor pattern selection.

Keywords: Columba livia; cluttered flight; collision; efficiency; stability.

Fig. 1.

Two traversal postures: wings paused in upstroke and wings folded. Two stereotyped groups of postures were adopted in traversing the gaps between obstacles: (A) the “wings-paused” case in which the wings were held at the top of upstroke, with downstroke resuming on the far side of the obstacle array, and (B) the “wings-folded” case in which the wing stroke was interrupted and the wings tucked along the body and unfolded after passing between obstacles. These postures were quantitatively classified into paused and folded clusters (C) by applying a normal mixtures model to wrist elevation and wing pitch. Each posture class exhibits a stereotyped or typical posture near the centroid of their respective clusters, within a distribution of similar poses. In D, wrist elevation, the height difference between the mean of the torso-mounted markers shown in Fig. S3A and the wrist marker, is directly compared between postures. Similarly, in E, wing pitch, the pitch angle between wrist–wingtip line and upper-lower torso markers line, is directly compared between postures. In both cases, one-way ANOVA tests show significant differences between each posture’s mean pitch and elevation. Additionally, standard least-squares linear models of pitch and elevation as they depend on individual pigeon, traversal method (pause or fold), gap size, and the interaction between traversal method and gap size show traversal method as having a significant effect (P_effect < 0.001 in both cases).

Fig. 2.

Fraction of trials using each posture as gap size changes. As the openings narrow, the strategy used shifts from predominantly pausing the wings to folding the wings in a majority of trials. A least-squares linear fit to the fraction folded or fraction paused, means across individuals, shows a significant dependence on gap size (P = 0.035 in both cases, as each is the inverse of the other).

Fig. 3.

Effects of traversal method and gap size on flight metrics. Flight metrics are affected by individual pigeon used, traversal method, and gap size. The interactions between traversal method, gap size, and flight metrics are shown here as quantile plots (with outlier whiskers), whereas their significance is judged by ANOVA. Traversing the obstacles with the paused posture (A) takes less time than doing so with the folded posture and (B) loses less height than doing so with the folded posture. These advantages of the paused posture are not a result of either (C) faster traversal of the obstacle array when the paused posture is used nor of (D) faster downstroke speed when the paused posture is used; neither of these flight metrics are significantly affected by traversal method. Likewise, maneuver duration and height lost are not correlated with gap size. This rules out preference for a given traversal method at a given gap size giving a false correlation between the traversal method and the respective flight metric. There is (E) no significant interaction between where in the wing beat cycle a bird would be at pass through and what traversal method it uses during pass through. Here, downstroke starts at 1 and upstroke is completed at 0, giving position within the wing beat as the fraction of a complete stroke remaining. This, along with the lack of correlation between gap size and stroke fraction, suggests that stroke fraction contributes little to the choice of traversal method.

Fig. 4.

Stability comparison between postures. Positions of the body, the head, and along the wings (as in Fig. S3A) determine the locations of point masses in a mass distribution model (9). Spheres whose volume is proportional to their relative mass depict both posture models, with beak and eye markings for orientation. In the paused posture (A), the mass of the wings is extended above and outward from the body. In the wings-folded posture (B), the wings are tucked lower and closer to the body. The posture models are used to calculate (C) the response to a perturbing torque generated by a simulated forward flight collision between the environment and the hand wing (where all observed obstacle collisions were observed to impact). This response is given as the angular acceleration (dω/dt) resulting from a perturbing torque, normalized to the maximal acceleration, providing a measurement of destabilizing rotation to be countered following wing strikes. The folded posture suffers less rotational perturbation across 90% of the wingspan, reflecting a more stable posture.