onto a panel, but previously stripe patterns on photographic negatives or printed on cards were used (Figure 2.7).
FIGURE 2.7 A schematic drawing of a setup used to determine visual acuity in a Harris’s Hawk Parabuteo unicinctus. The bird is given a two-choice task. It has been trained to wait on the perch until a pair of patterns is shown on the panels, which are 10 m away. As soon as the bird leaves the perch the patterns are switched off and the bird flies to the perch of its choice. If it flies to the perch in front of the ‘correct’ panel it receives a small food reward. The bird is observed remotely so that humans cannot influence its behaviour. Many such trials are conducted and the width of the stripes on the panels is altered in a random sequence. This allows the performance of the bird with respect to the width of the stripes on the test gratings to be measured, and a threshold stripe width can be determined (see Figure 2.8).
By the time the investigation gets to this point, the person running the investigation will have got to know the birds well, but will not be in direct contact with the birds. The birds’ behaviour will be remotely observed, so that the investigator cannot inadvertently influence the birds’ choices. The way the investigator influences the bird is through the presentation of food items which are paired with the presentation of only one type of stimulus panel, for example, the one with horizontal stripes. If the bird is hungry it will usually learn rapidly to move towards, or peck at, the panel with the horizontal stripes, as long as it reliably receives a food item each time it does so.
I have done vision investigations with doves who will peck directly at the panels for grain, owls that will peck at a bar beneath a panel to be rewarded by a small piece of meat, and with Great Cormorants who will swim up to a panel underwater in return for a small fish dropped to them through the surface. Some species can be trained to make their choice of panel from a fixed distance and then fly, swim, or walk towards the panel along a runway or swim-way. Owls seem particularly amenable to running, while diurnal birds of prey will readily fly from a start perch and land on a perch placed just in front of a panel.
In instances where flying, running, or swimming is involved, the bird must not be able to change its choice once it has passed a certain point. This ensures that the bird’s decision is always made a fixed distance from the panel. The use of a fixed distance is important for calculating the visual size of the object as seen by the bird and hence for defining its best performance. Depending on the species, it might be better for the bird to make a choice between relatively widely striped gratings at a long distance and fly to the panel (for example, a hawk), rather than choose between finer stripes at a closer distance (for example, a Budgerigar Melopsittacus undulatus). It is a matter of working with the bird’s natural behaviours to enable it to perform the choice task readily and hence reveal the ultimate limits of its vision.
Once this link between a target panel and the reinforcing food has become established, it is then a matter of starting to manipulate by small increments the task that the bird faces. The left–right positions of the panels will be changed randomly between trials, the light levels of the panels will be changed randomly over a wide range of light levels, and the widths of the stripes in the grating pairs will also be varied between trials. The point of these variations is to ensure that the bird is responding exclusively to the orientation of the stripes. In other words, that the bird has really learned that it is the orientation of the stripes that it should be attending to.
Training and testing
This type of simple discrimination is likely to be learned robustly by most birds. Once trained, birds will make the correct choice at least 90% of the times that a pair of stimulus panels is presented. In many cases mistakes are never made, the birds being consistently 100% correct. By ensuring that the birds gain most of their daily food intake in these sessions, and that the sessions take place at the same time every day, the birds will be highly motivated to respond when they come to a training session. Also, it is best to give a bird a training or testing session every day, no time off at weekends.
Depending on the species and on the individual birds, this initial training may be completed in a couple of weeks, but it can take a couple of months. The investigator has to be alert to the possibility that the birds will learn to solve the problem posed by using a cue other than the one that the investigator intends. When birds are being trained to make these kinds of visual discriminations they can be particularly alert to sound cues which might also reliably indicate where to go to get food. So, for example, a click or rumble associated with changing the positions of the panels between left and right could well be learned, and the bird may then ignore the orientation of the stripes. This is perhaps not too surprising. After all, from the bird’s point of view the task is to learn how to get its food reliably, and any cue that is consistent is as good as another, and sounds are very important to birds. So if the investigator introduces more than one reliable cue the bird could latch on to any of those that are available.
Once this initial training has settled down and the investigator can be sure that the bird is responding only to stripes of the correct orientation, the actual investigation of sensory abilities can begin. The first stage is to start presenting the birds with pairs of stripes of the same width but to change stripe width between trials. On some trials stripes will be very wide and the difference between the panels’ orientations will be obvious to the bird. However, if the stripes of the grating pattern are made very narrow the task will be difficult, as the fine stripe pattern cannot be determined and the panel looks a uniform grey.
The result is that on some trials the birds may make a mistake, and errors will start to systematically creep in. If the stripes are so narrow that they are below the birds’ threshold and they truly cannot ever tell the panels apart, then random behaviour would be expected, and the birds would get the task correct on only 50% of the trials. At intermediate stripe widths, however, the birds may get the discrimination correct on 70% or 80% of the trials. Over time, by mixing up trials with different stripe widths, a bird’s motivation to keep responding even on difficult trials can be maintained.
The uncertain threshold
When enough trials are accumulated a relatively stable relationship between error rate and stripe width will emerge. This relationship is called a psychophysical function, and the point at which a bird gets 75% of the trials correct is considered the threshold of sensory performance (Figure 2.8). So, after all the training and testing, it is this 75% correct stripe width that the investigator is trying to determine, not what the bird can see 100% of the time.
FIGURE 2.8 An example of a psychophysical function of a Harris’s Hawk. Performance over a large number of trials has been accumulated on a two-choice task using the kind of setup shown in Figure 2.7. The average percentage ‘correct’ performance for panels showing stripes of different widths is accumulated. The black dots on the graph show the bird’s actual performance for different stripe widths and the red curve is the line of best fit to those data points. With wide stripes the bird is correct on nearly every trial, while with very narrow stripes its choices are random (50% correct). It is in the region between these two extremes that the threshold of the bird’s acuity lies. The region where the birds is 75% correct is usually taken as the threshold performance. In this example 75% correct would occur with stripes that are about 0.7 minutes of arc wide. Similar data are gathered for a number of birds using the same technique, and these are averaged to give an acuity estimate for the species. In the case of Harris’s Hawk, the published average acuity is 1 minute of arc (29 cycles per degree).
The important point to note is that the threshold is defined statistically. It is the stripe width at which on average the birds get the
