D

Fig. 4.5. Discrimination among patterns of different spatial frequencies viewed at a distance. (a) The apparatus consisted of 12 arms opening into a central arena through which bees had access to any of the arms. (b) Bees were trained to six different checkerboard patterns presented, one at a time, on the back wall of one of the arms. In the tests, the trained bees had to choose between four novel patterns, each presented on the back wall of three different arms. Choices were recorded at each arm's entrance. (c) and (d) The four patterns offered in each test differ in frequency, but not in type. Percentage of choices is given under each pattern. n depicts the total number of choices. Values of p are results of x2 tests comparing the test results with the distribution of choices expected under random-choice conditions. (Modified from Lehrer et al. 1995.)

Fig. 4.5. Discrimination among patterns of different spatial frequencies viewed at a distance. (a) The apparatus consisted of 12 arms opening into a central arena through which bees had access to any of the arms. (b) Bees were trained to six different checkerboard patterns presented, one at a time, on the back wall of one of the arms. In the tests, the trained bees had to choose between four novel patterns, each presented on the back wall of three different arms. Choices were recorded at each arm's entrance. (c) and (d) The four patterns offered in each test differ in frequency, but not in type. Percentage of choices is given under each pattern. n depicts the total number of choices. Values of p are results of x2 tests comparing the test results with the distribution of choices expected under random-choice conditions. (Modified from Lehrer et al. 1995.)

the rewarding disc from each of a series of test discs rotated about their centers at various angles. The larger the angle of rotation, the better the discrimination. However, a comparison between results obtained with clockwise and counterclockwise rotation of the test patterns showed that the cue used in that discrimination task was not the spatial orientation of the edges, but rather the distribution of black and white areas in the visual field. These findings led to the formulation of the so-called "template theory" (see reviews by Srinivasan 1994; Heisenberg 1995), which proposed that bees store an eidetic ("photographic") image of the rewarding pattern in such a way that the pattern can be recognized only when its elements project on the eye in the same retinal positions that the bee viewed them during training. Discriminating a memorized pattern from a novel pattern is then based on the amount ofoverlapping and non-overlapping contrasting areas between the two patterns. In further experiments, Wehner (1972) showed that the discrepancy between the template and the currently viewed pattern is weighted more strongly in the ventral part of the frontal visual field than in other eye regions. A similar conclusion was later drawn concerning the discrimination of spatial frequencies (Lehrer 1997) and colors (Lehrer 1999) in different frontal regions of the bee's eye: best performance is observed when stimuli project onto the ventral part of the frontal visual field.

The use of a template requires the presence of a fixation point at which the various pattern elements project onto particular, constant regions of the eye. When viewed, such a point allows the flying bee to position itself in a constant way with respect to the target. The occurrence of a fixation reaction was demonstrated directly in a series of experiments using, among other techniques, cinematographic recordings (Wehner & Flatt 1977).

The patterns used by Wehner were rather large, subtending 130° at the fixation point. It is very likely that the bees were able to discriminate between the previously rewarding pattern and the novel one when they were still some distance from the target, i.e., during the approach flight. Therefore, although a bee's choice was recorded as the bee touched the hole in the center of one of the patterns (where it expected the reward), the observed preference for the rewarding disc need not be based on a space-invariant eidetic memory. More direct evidence for the use of a template would be presenting the bees with patterns that cannot be resolved before the bee has arrived at the fixation point.

In a recent study, Giurfa et al. (1995a) used four-petal model flowers

(7 cm diameter) whose shape could not be resolved before the bee approached them at very close distance (5 cm). Bees were trained to a "flower" of a constant spatial alignment that was then tested against the same flower displaying other orientations. The bees' choices confirmed Wehner's conclusion that the rewarding pattern is stored as a template in which the lower half of the visual field is weighted more than the rest of the visual field (Wehner 1972). Furthermore, even when bees can resolve the patterns at some distance, as in Wehner's experiments, their choice behavior is based mainly on the image viewed during the fixation phase (see also Wehner & Flatt 1977). Some patterns, such as radially symmetrical sectored discs, offer a fixation point even at some distance (Horridge 1999), but, in all cases, the presence of a fixation point is a prerequisite for template learning.

The role of pattern orientation

To determine whether or not pattern orientation could be learned and used by bees as an independent feature, bees must be prevented from forming a template. Hateren etal. (1990) studied the role of pattern orientation using a Y-maze similar to that shown in Fig. 4.2. The patterns were linear gratings in which the distribution of black and white areas was randomized in both the positive and the negative grating throughout the training, keeping the spatial alignment of the patterns constant. This training procedure prevented the bees from memorizing a particular distribution of black and white areas. Each bee's first decision was scored as it entered one of the two arms, i.e., decisions were evaluated at some distance from the patterns. In those and in a series of similar experiments (see review by Srinivasan 1994), bees were shown to learn the orientation of contours and to use this parameter in subsequent discrimination tasks even when they were presented with novel patterns.

In other experiments, again using the Y-maze, Giger & Srinivasan (1995) trained bees to discriminate between two gratings that differed either in the orientation of bars (horizontal versus vertical) or in the distribution of black and white areas. Their results show that, when patterns contain strong directional cues, as linear gratings indeed do, bees ignore the template and use the contour orientation for accomplishing the discrimination. They use eidetic memory only when they are forced to do so, e.g., when the two patterns to be discriminated differ in the distribution of areas, but not in the spatial orientation of the contours (Giger & Srinivasan 1995).

In earlier experiments that recorded the bee's choices very close to the patterns, discrimination of orientation of linear gratings was excellent, even though the distribution of the black and the white areas was kept constant (Wehner 1971; Lehrer et al. 1985). In the light of the results obtained by Giger & Srinivasan (1995) described above, the results obtained in those earlier studies need not be based on eidetic memory. There might be no difference between close-range and long-range performance in the context of the discrimination of orientation when patterns contain strong directional cues. When directional information is weak, as was, for example, the case in the half-black, half-white patterns used by Wehner (1972), or in the four-petals patterns used by Giurfa et al. (1995a), then bees form a template and use the distribution of areas to accomplish pattern discrimination, regardless of whether decisions are made at a close range (Wehner 1972; Giurfa et al. 1995a) or at a distance (Horridge

The role of symmetry

Symmetry is a visual cue available in almost all floral patterns (Neal et al. 1998). In recent years, results of several behavioral studies revealed that bees perceive symmetry and use this parameter in pattern-discrimination tasks.

To investigate the bee's long-range appreciation of symmetry, Lehrer et al. (1995) used the 12-arm experimental set-up described above (see Fig. 4.5a), training the bees, again, to the randomized checkerboard patterns. In subsequent tests, bees were required to choose between four patterns, each pattern being repeated in three different arms. When the four patterns differed in type, bees expressed a significant preference for radially symmetrical sectored patterns over all other types of pattern (Fig. 4.6a). When the four test patterns contained a constant number of bars that differed in arrangement, bees displayed a preference for radially symmetrical arrangements over less symmetrical or asymmetrical ones (Fig. 4.6b). And when bilaterally symmetrical patterns were presented, bees preferred patterns with a vertical axis of symmetry to patterns with a horizontal axis of symmetry (Fig. 4.6c). These and further results (Lehrer et al. 1995) suggest that bees prefer patterns that resemble natural flowers.

In another study, Giurfa et al. (1996a) examined the bees' capacity to learn bilateral symmetry using very small patterns (diameter 7 cm) and evaluating the bees' decisions at a very close distance (5 cm) from the patterns. During training, three patterns were presented simultaneously on

Fig. 4.6. As in Fig. 4.5, but the cues to be discriminated are: (a) the type of pattern; (b) the degree of symmetry; and (c) the orientation of the axis of bilateral symmetry. (Data from Lehrer et al. 1995.)

Fig. 4.6. As in Fig. 4.5, but the cues to be discriminated are: (a) the type of pattern; (b) the degree of symmetry; and (c) the orientation of the axis of bilateral symmetry. (Data from Lehrer et al. 1995.)

a vertical plane, one positive (rewarding), the other two not. When the positive pattern was symmetrical, the two negative patterns were asymmetrical (Fig. 4.7a), and vice versa. Throughout the training, the patterns were randomized with respect to their shapes (Fig. 4.7a), so that the bees could not form a template of the rewarded pattern and could rely on no cue other than symmetry. In the tests, the trained bees were given a multiple choice among 12 novel patterns, six of which were symmetrical, the other six asymmetrical (Fig. 4.7b).

The results (Fig. 4.7c) show that bees learned to prefer either the symmetrical or the asymmetrical test patterns, depending on whether they had previously been trained to the former or to the latter pattern category. The trained bees generalized the parameters "symmetry" and "asymmetry", respectively, to the novel patterns presented in the tests. However, bees trained to the symmetrical patterns performed significantly better in the discrimination task than did bees trained to the asymmetrical patterns (Giurfa et al. 1996a). Such a bias may be either innate or acquired through previous experience of the insects with symmetrical flowers in

(a) Symmetric and asymmetric training stimuli

(b) Symmetric and asymmetric test stimuli

Was this article helpful?

0 0

Post a comment