The principles of visual guidance for flight have been explored extensively in many insects. This research has revealed that insects rely heavily on optic flow—the pattern and speed of the motion of the image of the environment in their eyes that they experience during flight—to orchestrate a number of important behaviors (rev. Srinivasan, 2011a). A straight course and a stable attitude are maintained through the so-called “optomotor response,” which acts in such a way as to counteract rotations of the image of the environment in the eyes. The study of this optomotor behavior, pioneered by Hassenstein and Reichardt (1956) in the beetle Chlorophanus, has since been pursued in a wide range of insects, including the housefly Musca, the fruitfly Drosophila, (rev. Borst, 2009) and honeybees (rev. Srinivasan, 2011a). Honeybees (e.g., Srinivasan et al., 1996; Baird et al., 2005) and Drosophila (e.g., David, 1982; Fry et al., 2009) regulate the speed of their flight by monitoring and holding constant the optic flow generated by the surrounding environment. This ensures that the insect flies at a high speed in a safe, open environment (such as a field), and automatically slows down when it enters a cluttered environment such as a forest. Honeybees avoid collisions with objects by steering away from regions of the visual field that induce strong optic flow: rapid image motion signifies the presence of an object that is dangerously close (Kirchner and Srinivasan, 1989; Srinivasan et al., 1991; Srinivasan and Zhang, 1997). Honeybees (Kirchner and Srinivasan, 1989; Srinivasan et al., 1991) and bumblebees (Dyhr and Higgins, 2010) navigate safely through narrow gaps, avoiding collisions with the edges, by moving along a trajectory in which the two eyes experience the same magnitude of optic flow. Drosophila veer away from objects that generate rapidly expanding images that herald an imminent collision (e.g., Muijres et al., 2014). Honeybees orchestrate smooth landings by holding constant the magnitude of the optic flow that is generated in the vicinity of the landing target, as the target is approached (Srinivasan et al., 2000; Baird et al., 2013). This strategy ensures that the insect slows down progressively as the target is approached, reaching a speed that is close to zero at touch down. There is evidence that Drosophila control their deceleration while landing by measuring the rate of expansion of the image of the surface that is being approached (e.g., Van Breugel and Dickinson, 2012) and/or computing the time to contact from the image expansion (Wagner, 1982), and that the extension of the legs in preparation for the final touchdown (in houseflies and Drosophila) is trigged by the size and/or rate of expansion of the image of the surface that is being approached (e.g., Eckert, 1983; Borst and Bahde, 1986; Tammero and Dickinson, 2002; Van Breugel and Dickinson, 2012).

Recent work indicates that birds, like insects, have strongly developed brain areas that are tuned for processing optic flow (revs. Frost, 2010; Wylie, 2013). Does this provide birds with equally adept visual flight control? What visual strategies do birds use while navigating through the environment, and how do these strategies compare with those known to be key for visual flight control in insects (Figure 1)? Until the past decade, relatively little research has been carried out to address these questions. Zebra finches hold their heads at a constant orientation -interspersed by brief saccadic rotations—while flying past an obstacle (Eckmeier et al., 2008). This suggests that optic flow could provide a reliable estimate of obstacle distance. How the information derived from optic flow is used to control the various phases of bird flight—such as takeoff, cruise, obstacle avoidance, and landing—remains to be uncovered, although some clues have begun to emerge, as will be described in this review.

Interestingly, the perception of movement appears to be “color-blind” in honeybees (Lehrer, 1987; Srinivasan, 2011a) as well as humans (Zeki, 1993), even though both creatures possess excellent trichromatic color vison. In the context of this article, it is of interest to enquire whether the movement-detecting pathways of birds—many of which possess tetrachromatic color vision—are also color-blind.

Schiffner and Srinivasan (2015, 2016) investigated the control and regulation of flight speed by flying Budgerigars in tunnels, and filming and reconstructing their flights in 3D using high-speed stereo video cameras. In their first study, the tunnel was of constant width, and displayed grating patterns that were projected on the side walls. The flight speeds of the birds were measured when the gratings on both walls were stationary, as well as in conditions when they were moved at a number of different speeds, both in the direction of flight and against it (Schiffner and Srinivasan, 2015). When the gratings were stationary, the birds flew through the tunnel at an average speed of about 6.3 m/s. When the gratings were moved in the direction of the birds' flight, at various speeds, the birds showed an increase in flight speed that was proportional to the increase in grating speed, but was not large enough to match it (Figure 2). When the gratings were moved in the opposite direction, the birds did not change their speed at all. Thus, Budgerigars do respond, to some extent, to changes in the motion of the image of the environment. However, unlike bees, these birds do not appear to rely solely on image motion to regulate the speed of their flight; other factors seem to play a role as well, as we shall discuss below.

In a second study (Schiffner and Srinivasan, 2016), the tunnel carried stationary patterns on the walls, but the width of the tunnel varied gradually from one end to the other. Birds flying in this tunnel displayed only two, distinct flight speeds: a high speed of ~10 m/s in the wide section, and a constant low speed of ~5 m/s in the narrow section (Figure 3). Thus, like bees, Budgerigars fly faster in open environments and slower in narrow environments. However, while bees continuously adjust their speed during flight in a tapered tunnel—so as to hold the optic flow constant—Budgerigars seem to use primarily two distinct speeds.

These findings do not necessarily imply that optic flow is the visual cue that is used by the Budgerigars to gauge the width of the tunnel. Further investigation is needed to explore other possibilities. Given that the eyes of these birds possess laterally oriented fields of view with relatively little binocular overlap, stereo-based ranging is unlikely, and it is difficult to conceive of other plausible visual cues that could be used to assess tunnel width.

Why do Budgerigars behave differently from bees? We propose that, for Budgerigars (and possibly for other birds as well), energy requirements may play an important role in the choice of flight speed. The high speed measured in the tapered tunnel closely approximates the preferred (energy-optimum) cruising speed of Budgerigars, as measured in wind tunnel studies (~9.75 m/s; Tucker, 1968). The low speed, on the other hand, appears to be a speed that these birds adopt when they fly in relatively dark or cluttered environments. While it is not yet clear whether the low speed represents another local minimum of energy consumption (possibly a flight mode that involves flap-bounding), we find that our Budgerigars consistently fly at a low speed of ~5 m/s when they fly in environments that could be perceived by them to be unsafe. We therefore propose that Budgerigars use two flight speeds—a “cruising” speed of ~10 m/s in open environments, and a safe “maneuvering” speed of ~5 m/s during flight in dark or cluttered environments, switching between the two in response to changes that they encounter in the environment. Our results also reveal that when the birds fly through the tapered tunnel in the narrowing direction they switch to the low speed further into the tunnel, compared to the location at which they switch to the high speed during flight in the widening direction (Figure 3). The separation of ~2 m between these points suggests that the switching behavior is “anticipatory”: the birds are setting their speed by gauging the environment that is approximately 1 m ahead of their current position; that is, by sensing the magnitude of the optic flow in the regions of the visual field that are located about 20° lateral to the flight direction (Schiffner and Srinivasan, 2016).

During flight in dense forests, birds often need to fly through extremely narrow passages without hurting themselves. Guidance of flight through a two dimensional array of obstacles (vertical poles), akin to a forest of trees, has been investigated in the laboratory in pigeons (Lin et al., 2014). The birds had the option of taking several different routes of flying through the constellation of obstacles. This study revealed that the birds tended to choose routes in which the successive gaps that were encountered were (a) as wide as possible, and (b) consistent with the desired flight direction. Similar results were obtained by Ros et al. (2017), in which pigeons were again trained to fly through a two-dimensional array of obstacles, but where the obstacles were oriented horizontally, rather than vertically. This required the birds to fly through the obstacle constellation by rapidly varying their flight height, rather than changing their flight direction. Here again, the birds tended to choose routes in which the gaps that were encountered were as wide as possible, and consistent with the desired flight direction.

Schiffner et al. (2014) investigated the kinematics of flight through a single aperture by training Budgerigars to fly through a corridor which presented a single narrow, vertically oriented aperture (a slit), mid-flight. The birds' flights were video-filmed as they flew through the aperture for various aperture widths - ranging from several times the wingspan, down to values just marginally greater than the width of the thorax. The results revealed that, during the passage through the aperture, they close their wings if (and only if) the wingspan exceeds the aperture width. Figure 5A illustrates a comparison of a flight through a wide aperture, with a flight through an aperture that is narrower than the wingspan. Analysis of the wing-closing behavior of Budgerigars flying through apertures of various widths reveals that these birds are aware of their wingspan to a precision of ± 1 cm, which is about 6% of the average wingspan (Schiffner et al., 2014). Furthermore, analysis of the behavior of individual birds (with wingspans that vary over a range of 29–33 cm across the group of birds tested) reveals that wing closure is not triggered at a fixed aperture width. The critical aperture width varies from bird to bird, and depends upon its individual wingspan. Thus, each bird is precisely aware of its personal wingspan (Schiffner et al., 2014).

Wing closure, when it occurs, occurs well ahead of the aperture—at a modal distance of about 22 cm. Thus, the cues that are used to determine whether wing closure is necessary—and if so, to determine the distance at which it occurs—are most likely based on vision, and not on tactile or aerodynamic interactions with the aperture.

A later study (Williams and Biewener, 2015) investigated the same question in pigeons and found similar results, with the additional observation that these birds adopted one of two postures while passing through narrow gaps. When the gap was very narrow, they folded their wings back completely (like the Budgerigars). When the gap was wider (but still narrower than the wingspan), they adopted a “paused” posture in which the wings were held stationary at the top of the upstroke. Here again, the birds were clearly aware of the size of their wingspan in relation to the width of the gap.

Why do birds need such precise knowledge of their wingspan? We suggest that Budgerigars and pigeons interrupt the flapping of their wings only when it is absolutely necessary because such an interruption (a) reduces the lift force that is necessary to keep the bird aloft, and (b) seriously compromises the ability to control flight, effectively reducing the bird to a projectile. It would be interesting to ask whether this type of “body awareness” in birds is genetically pre-programmed, or learnt from experience and updated steadily as the bird grows to an adult—the latter perhaps being a more likely explanation. One approach to answering this question would be to look for changes in wingspan awareness when the wingtips are trimmed or artificially extended. One has to bear in mind, however, that such manipulations could affect the weight and aerodynamics of the wings in artificial ways, complicating the interpretation of the results.

Analysis of the trajectories of the birds as they approach the apertures reveals that the birds which fly through the aperture without closing their wings (because the aperture is wider than the wingspan) maintain a constant height throughout the approach (Figure 5B; Vo et al., 2016). On the other hand, the birds that close their wings (because the aperture is narrower than the wingspan) increase their height at a mean distance of about 140 cm from the aperture (Figure 5C; Vo et al., 2016). This increase of height occurs well ahead of the point at which the wing closure actually occurs (22 cm from the aperture; Schiffner et al., 2014). We suggest that this height increase is a manoeuver to compensate for the subsequent loss of height that occurs due to the temporary closure of the wings. If this is true, it would imply that Budgerigars pre-plan flights through narrow gaps in a sophisticated way.

Analysis of the approach trajectories also reveals that the speed at which the birds approach the aperture is independent of the width of the aperture, and is independent of whether or not the aperture requires wing closure (Vo et al., 2016). The birds approach all of the apertures at a speed of approximately 4 m/s (Figure 5B). Interestingly, this speed is close to the “maneuvering” speed that the birds display in the narrower part of the tapered tunnel (see Figure 3). This reinforces the notion that Budgerigars fly at a stable, constant speed of 4–5 m/s during flight in cluttered environments. Exactly how cluttered or “dangerous” the environment has to be to trigger a switch to the lower speed, and exactly what properties of the environment are used in making this decision, remains to be investigated.

Why do Budgerigars adopt a constant flight speed in dense environments? One potential advantage of this strategy is that moving at a constant, known speed allows the range to various obstacles in the environment to be calibrated directly in terms of the optic flow that is experienced by the visual system. This appears to be different from the behavior of most flying insects, including bees, which progressively reduce their flight speed as the density of the environment increases (Srinivasan et al., 1996). Therefore, unlike Budgerigars, flying insects face a “chicken and egg” problem: The distance to an object cannot be estimated without knowledge about the flight speed, and vice versa. Nevertheless, insects have evolved ingenious solutions to overcome this paradox in the context of many visually guided behaviors such as cruising flight, obstacle avoidance and landing (Srinivasan, 2011b).

The findings with the Budgerigars beg the question of how these birds regulate their speed at each of the two speeds at which they prefer to fly. The higher speed, which is in the vicinity of 10 m/s, may be set by the energy consumption curve, which exhibits a minimum at this speed (Tucker, 1968). The mode of regulation at the lower speed remains to be explored. Two possibilities are (a) generating a constant, calibrated thrust; and (b) sensing the airspeed (possibly via the feathers) and using this information to control thrust in a feedback loop.

During flight, birds need to avoid collisions with stationary obstacles as well as moving objects, such as other birds. Schiffner et al. (2016) investigated mid-air collision avoidance in Budgerigars by launching two birds from the opposite ends of a narrow tunnel and video-filming their flights as they approached and flew past each other. The results revealed that Budgerigars avoid imminent head-on collisions by adopting a simple, consistent rule: each bird veers to its right. Evidently this strategy for collision avoidance—which is also used by airplane pilots—has been evolved by these birds millions of years ago. It remains to be explored whether this “right hand rule” applies to other bird species as well.

Classically, movement-induced responses in flying insects have been studied by tethering the insect in the middle of a drum, decorated with vertically oriented black-and-white stripes, and measuring the insect's yaw torque when the drum rotates alternately clockwise and counterclockwise. These studies, pioneered by Reichardt (1969), have since been carried out in a wide range of insects, including the housefly Musca, the fruitfly Drosophila, and honeybees (Apis), references for which are given in Srinivasan (2011a). Under these conditions the insect attempts to follow the rotation of the drum, exerting a clockwise torque when the drum rotates clockwise, and vice versa. In free flight, these turning (“optomotor”) responses serve to compensate for unintended (disturbance induced) deviations from the intended flight direction (Reichardt, 1969). Interestingly, honeybees in this experimental paradigm display strong optomotor responses when the alternating stripes are black and white, but no response at all when the colors and intensities of the alternating stripes are such that they excite the honeybee's green receptor equally strongly (Kaiser, 1975), thus providing no detectable contrast to the green receptors. This finding and a number of other observations (Lehrer, 1987; Srinivasan, 2011a) suggest that the movement detecting pathways in the bee's visual system are driven exclusively by the green receptor channel and are therefore “color blind”—although the bee's color perception pathway endows it with excellent trichromatic color vision, featuring signals from UV-, blue- and green-sensitive photoreceptors.

When bees are trained to land on a disc placed on a table to feed on a drop of sugar water positioned at the center, they touch down at the edge of the disc and walk to the reward, rather than land directly at the reward (Lehrer et al., 1990). Evidently, bees aim for the boundary at the rim of the disc, and use this high-contrast visual feature to guide their landing. Bees land consistently at the rim of the disc when the disc is black and the background (table) is white. However, if the colors of the disc and the background are such that they provide no contrast to the green receptors, the bees no longer land selectively at the rim of the disc (Lehrer et al., 1990). This demonstrates that the visual cue used to guide the landing at the target is sensed by a movement-detecting pathway that is again color blind.

In the Budgerigar, the chromatic properties of movement detection were investigated by using a similar landing paradigm in which the birds were trained to land and collect food placed at the center of a blue disc (Bhagavatula et al., 2009). Here again, the birds, like the bees, generally tended to land at the rim of the disc. However, when the color of the disc was held constant and the background was varied from black to white through various shades of gray, the birds lost their preference for the rim when the background was a particular shade of gray that stimulated the bird's red photoreceptors exactly as strongly as did the blue disc. This finding suggests that, in Budgerigars, the movement-detecting pathway that guides landings is again color blind, and is in this case driven by the red photoreceptors (Bhagavatula et al., 2009). Further work is required to investigate the chromatic properties of other movement-sensitive behaviors in birds.

Interestingly, movement perception is color blind in humans as well: it is driven by the luminance pathway, which sums signals from the red and green cones (Zeki, 1993). Why is movement detection color blind across several animal species, as the evidence so far seems to suggest? One possible explanation is that the capacity to detect and respond to movement—which is critical to many aspects of behavior—is a fundamental building block that visual systems, initially working with a single spectral class of photoreceptor, evolved as part of their basic “Bauplan,” before incorporating additional capabilities such as color vision (Srinivasan, 2011a).

There has been considerable work on the visual guidance strategies of flying insects, particularly with honeybees. Recent work with flying birds points to some similarities in the strategies used for flight control, although with substantial modification. For example, movement perception appears to be color-blind for both groups, but with different cone pigments in birds compared to insects. However, work so far indicates more differences than similarities in flight strategy, with key examples being the centering response and the control of flight speed. In both of these cases, the avian strategies seem to be less stereotyped than in insects. Centering and velocity control during forward flight in insects relies primarily on pattern velocity cues, whereas birds adjust position and speed during forward flight to a combination of pattern velocity and the rate of vertical image expansion.

Could the differences in visual guidance strategies between flying birds and insects be due simply to variation in experimental approaches? For example, in many of the insect studies, the optic flow was manipulated in the ventral as well as the lateral regions of the visual field, whereas in the case of birds, this manipulation has so far been restricted to the lateral fields. Another, broader reason may have to do with the fact that birds, given their higher visual acuity and more developed brains, possess better “scene awareness” than do insects, making it experimentally more challenging to change their behavior by manipulating the visual stimuli that they experience. It would be highly informative to repeat a larger set of the same experimental treatments for both groups.

A major driver of the recent efforts to study visual guidance in birds was the availability of tracking systems and large field stimulus presentation. At present, such systems are not well automated but it is likely that with greater availability and usability, we will know much more about avian visual guidance in the coming years. A potentially fruitful avenue of investigation would be to combine the behavioral studies with electrophysiological investigations of the neural substrates. Behavior and electrophysiology have long been integrated in research with insects, and when combined with molecular genetics, this has led to major progress over the last decade in understanding visual circuits for flight control (Rister et al., 2007; Joesch et al., 2010; Maisak et al., 2013; Behnia et al., 2014). Because molecular approaches now available in non-genetic model organisms are starting to be used with birds (Roberts et al., 2012, 2017), we are hopeful that a deep understanding of how the avian brain transforms visual information into motor output is within view.

DA and MS wrote and edited the manuscript.