First published online August 18, 2005
Journal of Experimental Biology 208, 3233-3247 (2005)
Published by The Company of Biologists 2005
doi: 10.1242/jeb.01790
Generalization of convex shapes by bees: what are shapes made of?
Miriam Lehrer1,* and
Raymond Campan2
1 Department of Neurobiology, Institute of Zoology, University of Zurich,
Winterthurerstrasse 190, 8057 Zurich, Switzerland
2 Laboratoire d'Ethologie et Cognition Animale, Université Paul
Sabatier, 118 route de Narbonne, F-31062 Toulouse, France
View larger version (12K):
[in a new window]
|
Fig. 1. The experimental apparatus. (A) Plan view; (B) frontal view. Two shapes,
one rewarding (termed positive), the other not (negative), are presented
simultaneously, each mounted on a 50 mm-long Plexiglas tube that penetrates
its centre and connects the front wall of the apparatus to the reward box
fixed behind it. The tube associated with the positive shape is open (A,
right) whereas the other ends blindly. After training bees with a particular
pair of shapes, they are tested by offering them a choice between the positive
shape and each of several novel shapes (discrimination tests) or between a
novel shape and other novel shapes (transfer tests), as explained in the text.
The first choice of each bee on each of its visits is recorded. Choice
frequency (CF), i.e. the percentage of choices for a given shape, is
calculated from the sum of choices (N) obtained in each type of
test.
|
|
View larger version (27K):
[in a new window]
|
Fig. 2. Results of Expt 1, colour discrimination. Bees were trained with a blue
square (positive) versus a yellow square (negative) presented on a
white background. The trained bees are then tested with various pairs of
coloured triangles as specified along the abscissa; bl, blue; yel, yellow; gr,
green; viol, violet. The bars showing the choice frequencies (Cfs) obtained in
favour of each of the two triangles are coloured accordingly. The total number
of choices (N) recorded with each pair of triangles is given above
each pair of bars.
|
|
View larger version (30K):
[in a new window]
|
Fig. 3. Results of Expt 1, shape discrimination and generalization between colours.
The same bees as in Fig. 2,
trained with a blue square vs a yellow square. The inset shows the
choice frequency (CF) obtained during the training and the number of choices
(N) on which it is based. The trained bees were then offered a choice
between a square and each of four novel shapes (abscissa). In the tests, both
test shapes were either blue, as was the positive training shape, yellow, as
was the negative training shape, or black, which was novel to the bees. The
bars depicting the CF obtained in favour of the square are coloured according
to the colour of the two shapes used in each test. The number above each bar
denotes the total number of choices recorded in the test.
|
|
View larger version (42K):
[in a new window]
|
Fig. 4. Results of Expt 2, shape discrimination and generalization from luminance
contrast to colour contrast and motion contrast. Bees were trained with a
black diamond (positive) versus a black disc (negative) placed
against a white background (inset). (A) The black-and-white bar on the left
shows the choice frequency (CF) obtained during the training. Bees were then
given transfer tests with a diamond vs a disc, which were both either
blue or yellow, again on the white background (blue and yellow bar,
respectively, with a white margin), and with black, blue and yellow shapes
placed on a patterned background (bars with patterned margins). (B) Bees
choosing between the learned diamond and a novel shape, a square, placed on a
white background. The pair of test shapes is black, blue or yellow. (C)
Control tests showing that the bees' choice behaviour constitutes a
generalization performance, as explained in the text. In all three panels, the
total number of choices (N) on which the CFs are based is given above
each bar.
|
|
View larger version (54K):
[in a new window]
|
Fig. 5. Results of Expt 2, shape discrimination based on motion contrast, and
generalization from motion contrast to luminance contrast and colour contrast.
Bees were trained with a patterned disc (positive) versus a patterned
triangle (negative) placed against a similarly patterned background. (A) Inset
shows the choice frequency (CF) obtained during the training and the number of
choices(N) on which it is based. The trained bees were then offered a
choice between the trained patterned disk and four different novel patterned
shapes, all placed on the patterned background. (B) The trained bees were
given 12 different transfer tests, presenting black or coloured shapes either
on a white background or on a patterned background. The bars denoting the CFs
obtained in each test are coloured or patterned accordingly to the colour or
pattern of the shapes used in the test, and the type of background is shown in
the margins of each bar. In these tests, the disk was tested either
versus a triangle, which was the negative training shape, or
versus a square, which was novel to the bees (abscissa). The number
above each bar denotes the total number of choices recorded in the test.
|
|
View larger version (41K):
[in a new window]
|
Fig. 6. Results of (A) Expt 4 and (B) Expt 5. Two pairs of randomly patterned
convex shapes were used in two separate training experiments (A and B),
presenting the shapes 50 mm in front of a patterned background (i, top). The
shapes are shown separately for better visibility (i, middle). In this
situation, bees could use only motion contrast for recognizing the shapes.
When the distance between shapes and background is eliminated (i, bottom),
bees cannot tell the two shapes apart. The trained bees are then tested using
the outlines or fragments of the outlines of the two training shapes
(ii–viii). Choice frequency (CF) and N-values are given for
each test. Results of reciprocal training experiments (not illustrated) were
very similar.
|
|
© The Company of Biologists Ltd 2005
