There are, however, a number of effects that conspire to make the identification of a colliding aircraft difficult for even the conscientious pilot, and the first of these is the problem of constant relative bearing. When two aircrafts are to collide, then each aircraft maintains a relative bearing to the other aircraft that is constant until the moment of impact. The subjective effect of this is that the colliding aircraft stays in the same place on the pilot's canopy unless he makes a head movement. This has two unfortunate consequences. The first is that no other aircraft that the pilot has ever seen (unless he has been involved in a previous midair collision or air-miss) will have possessed this characteristic, and he may not therefore have learned to use movement relative to the canopy as a cue to detection. The second is that the peripheral retina drives what is sometimes termed the 'ambient' rather than the 'focal' visual system. The ambient system is concerned with using motion of the proximal stimulus to provide a percept of orientation, and to attract the attention of the focal system (concerned with the detailed and conscious interrogation of the world) to items of interest in the peripheral retina - particularly items that change by moving or flashing. Movement is thus a very important attention-getting stimulus to the ambient system, but the colliding aircraft is the only one that fails to provide it.

As a colliding aircraft grows nearer, its retinal size naturally increases, and it may be supposed that this should act as a cue to detection. The rate at which this increase in size occurs is not, however, linear. This shows that the image of the colliding aircraft stays relatively small until very shortly before impact since, roughly speaking, the retinal size of the aircraft will double with each halving of the separation distance (separation distance changing linearly with time). It is probably for this reason that conversations with pilots who have experienced close air-misses or who survived collisions characteristically contain the assertion that a good visual look-out was being maintained, with the offending aircraft suddenly looming large, and apparently from nowhere.

Nevertheless, it may still seem surprising that colliding aircraft go unseen since they should represent targets that are well above minimal acuity levels for some time before impact. But, acuity is not distributed evenly across the retina. As the eccentricity from the fovea increases, acuity drops dramatically. Experimental work (Harris, 1973) has shown that the probabllity of detecting an aircraft target is closely related to the local retinal acuity. It is therefore true that colliding aircraft will be perceptible so long as they are acquired at or near the fovea. Many pilots, however, experience the effect of seeing an aircraft, only to look away from it and be unable to spot it again even though they search the part of the sky where they know it it to be. It may then reappear as if from nowhere when, by virtue of a chance eye movement, its proximal image is placed close to the fovea.

The efficiency of a visual search is thus likely to be governed by the way in which the proximal image of each part of the external world can be placed on the fovea and surrounding retinal area.  Visual searches may be conducted only by using saccades with rests, and its largely true to say that the world is interrogated only during the rests. Many pilots believe that they make visual searches by means of smooth and continuous eye movements, but such eye movements may be made only only when a smoothly moving target is being tracked. The percept may be of a smooth scan, but this percept is constructed from succesion of stationary images. The pilot must unconsciously decide the density with which to distribute his fixation points on the outside world. If the distribution is dense, thec probability of detecting the presence of an aircaft in the search area increases, but the overall size of the area searched decreases. If our pilot chooses to space his fixation points by  by 20 degrees, and has a visual world that subtends 200 degrees in azimuth and 60 degrees in elevsation, then a total of 30 saccade/rest cycles will be required for the search. Since each cycle occupies about one third of a second, the complete search will consume about 10 seconds. This is clearly more than enough time for an aircraft that went undetected during the first fixation to become sufficiently closer to be a hazard or even to collide.

The foregoing suggests that if pilots are to continue to fly relying only on vision to separate them from other aircraft,  then collisions will continue to occur, however diligent the pilots may be. Fortunately, the probability of a mid-air collision is low, but an understanding of the processes may enable pilots to improve their search and should prevent legislative authorities from censuring those pilots who are unfortunate enough to be confronted dramatically by the limitations of their perceptual systems.

The two main examples of visual problems discussed here - the visual approach and the mid-air collision - represent the two most critical - visual situations in flight. Nevertheless the importance of vision in  all flight must not be forgotten. It is the only sense that may be relied upon to provide reliable information on orientation, and is the source of the majority of other information that enables the pilot to build his internal model of the external world.

The essential point to understand about all perceptual processes is that our percept is a model or hypothesis about the external world which is built both from incoming  and from our expectations. Indeed, we are able to make sense of the world only because of our experience and expectations. It is, however, just these essential expectations that can sometimes lead us to make our model more in terms of the way that we would like the world to be, think that it ought to be, or in the way it always has been, than in the way it is actually is.

 

Harris JL (1973) Visual aspects of air collision. In Visual Search, pp 26 - 50. Washington:  National Academy of Sciences)