Note: The following is a guest post by G. Adrian Horridge. Dr. Horridge is probably best known as co-author of the massive two volume book, Structure and Function in the Nervous System of Invertebrates, published in 1965.
Enthusiasm and activity in the young, and scientific investigation by experiment, not to mention the publicity, are all commendable. There are some irrelevant mentions of dogs, human puzzles and helping us to understand ourselves better, that add to the interest but not to the science. Without them, the matter would be boring. However, science is improved when the conclusions agree with the data.
Three general remarks
We have abundant information about the vision of honeybees, and for the time being must assume that bumblebees are similar.
1. It is impossible to show that bees detect THE pattern, because they may have recognized only a small part of the pattern (a cue), as shown in 1938 by K. S. Lashley for the rat.
2. Honey bees detect cues, not patterns as we do, as shown when trained bees treat as equal a learned pattern and a different pattern that displays the same cues. They recognize the place of the reward by means of landmarks (combinations of cues).
3. Bees can be forced to see blue and yellow apparently at the same time, as shown by Baumgärtner in 1928 with small areas close to the reward hole, but they can do it only if they use one eye for each colour, because colour processing is done by the two eyes separately. Apart from this, bees cannot be trained on two alternative different colours at the same time at the same place, because what they learn on one colour, they unlearn when presented with the other.
Internal evidence in the Blackawton paper.
The bees were first trained with all the discs white and rewarded. Previous work tells us that if only the central four discs were rewarded at this stage, the bees would locate the reward by triangulation on the outer rows of discs. Whatever is then done to alter the outside edge in the training regime would spoil their piloting towards the centre. They used the outer edges because these were separated by large angles on the eye, making their triangulation more accurate as they approached.
The reward was then restricted to the central four discs and the blue and yellow colours were applied. As shown by the test results, the bees did not solve the problem by ignoring the colours and simply going to the centres, as a colour blind person would, so they did not use the simplest solution. Bees prefer blue to yellow inn the learning situation, so they learned to respond to blue when they arrived on the scene.
Two bees clearly learned to go to the blue centres, probably using the edges to locate the centre, and I expect they would never have learned anything about yellow. For them, yellow was not rewarded. They accepted green at the centre in Test 2.
To locate the position of the reward, two bees learned to avoid the blue at the periphery, and they accepted a yellow centre. For them, yellow was rewarded. These two bees rejected the green at the centre in Test 2. For related data, see figs 11.1 and 11.2 in the book referred to below.
Table 1c shows these two classes of bees. For bees O and B/Y the total score was yellow centres 7, blue 54. For bees B/O and B the total was yellow centres 64, blue 1. These two classes persist to some extent in tests 2 and 3; notably bees B/O and B accept green at the centre and yellow in test 3.
There were therefore two classes of bees, those that preferred blue at the centres and those that avoided blue at the periphery. All the bees used the position of the periphery as their cue to locate the centre. The bees that used a blue periphery need not have learned the colour at the centre at all; those that simply went to the blue centre triangulated on the edge of the square and expected blue at the reward.
To keep the data homogeneous, and to make inferences from it, results from these two classes of bees should have been kept separate.
The authors say that the salt was added in the outer ring of discs "so that they would learn to go not just to the colours, but had to learn the pattern". I suspect that they did not learn the pattern, only the cues mentioned above.
Both of the possible training patterns were presented, each twice, presumably so that the squares could be interchanged by rotation of the whole.
In tests, honey bees have to be refreshed with a short spell of more learning after making a few choices, because they do not persist in making hit after hit without a reward.
1. In test 1, the authors say they allow each forager to make 30 or so choices and each bee was tested three times, total 90 tests, and there were five bees, so was that a total of 450 choices? There is a small discrepancy with the total number of choices in tests 1, 2 and 3, which was 394.
2. Test 2. The trained bees were tested with squares in which the four central discs were green "to see whether the bees learned to go to the colours or the location". The authors say that the bees apparently chose at random. One explanation might be that all the bees piloted by using the periphery and none of them expected to find green at the centre, so they started to relearn the place of the reward by trial and error and appeared to choose at random. However, bees B/O and B that had learned to avoid blue at the periphery still went to the centre, and why should they not? However, bees O and B/Y that had learned to expect blue at the centre now went to the periphery, so the randomness was an artifact of adding the two classes together. The authors inferred that "that the bees did not solve test 1 by only going to the middle flowers", but no explanation was offered, and the bees had indeed solved the original problem by going to the middle flowers, but the authors hinted at something more.
3. In test 3, the central discs were exchanged with discs from the corners of the squares to see if the bees learned to go to the colours that were fewest or still go to the middle flowers. The result was 59 choices at the corners and 86 elsewhere, but we are not told how many chose the edges or the middle, so the test data is not comparable with the training data. Only 25% of the discs are at corners, but 40.1% of 145 choices were to corners, so there was probably a preference for corners. The individual preferences for or against blue can still be seen. However, the authors "think that the bees in test 3 selected the flowers randomly". This leads the authors to conclude "that in test 2 they must have used the larger square of blue and yellow flowers to decide to forage from the middle green flowers" even though they had previously found that choices in test 2 were random. Did the bees in test 2 forage from the middle green flowers or at random? Where did the authors get the idea that they used the outer square? Test 2 shows that two bees, B/O and B, had not learned the colour at the centre at all.
Test 3 is easily understood if all the bees piloted by use of the peripheral edges, so when these were dislocated in any way the bees started to relearn the place of the reward by trial and error, retaining some of their preference for or against blue. Totaling the score made them appear to choose at random.
The authors say that the bees had to learn to go to one colour (blue) if it was surrounded by the opposite colour (yellow) but also to go to yellow if surrounded by blue. The observed preferences of the bees show that two bees solve one pattern and two other bees solve the other. In each test, two of the bees became confused about the place of the reward because their expectations about either the centre or the edge were not fulfilled.
Apparently the authors thought that the bees see the coloured patterns and the experiment was designed to discover how the bees solved the puzzle caused by the two different arrangements of the colours. So, the experimental design was based on a human vision of the task, but the bees had a different idea about how to solve the problem, by learning only one pattern. Humans are few and have big brains that solve simultaneous tasks, but bees are very numerous and have small brains, so they take one task at a time.
When targets are fixed, and landing on the rewarded place is the criterion of success, it is very difficult to discover how bees locate the reward, because when they arrive they bring a memory of what they expect. Therefore any experimental change to the targets makes the bees start to learn afresh by trial and error, or go away.
This general topic is not new. In the last decades of the 19th century, Professor Felix Plateau (1841-1911) of Ghent, Belgium, published many papers on how bees find flowers. He hid real flowers with paper but the bees went under the paper, so he concluded that the shape and colour were not the chief relevant features. He also tried to attract the bees with carefully painted paper flowers, but the bees ignored them. Plateau concluded that bees recognized flowers by their odour, although others had shown that the bees did not need the odour.
Professor Auguste Forel (1848-1931) of Zürich, produced 50 pages of objections, supported by Plateau's own data. He accepted Plateau's data but not the false conclusions. He repeated the experiments and concluded that the flower shapes and their colours were not relevant but the bees went to the rewarded place with the help of landmarks that he could not define. He says (in translation) 'In using the experiments of M. Plateau himself to show the errors of judgement that he draws from them, I fully render homage to his scientific honesty. It is precisely this honesty which has allowed us to follow the author step by step, and to pick up, by the help of his faithful narration of facts, how the facts relate to each other and that they are mutually consistent. Thanks to this, our study will not have been a sterile polemic, for it has brought us to see more and more clearly into the very question which occupies us.' The question was how to design suitable experiments.
References, and more about bee vision can be found in a book published by ANU Epress, "What does the honeybee see?" which is now online for free personal use, at the unique URL: http://epress.anu.edu.au/honeybee_citation.html
G. Adrian Horridge, FAA, FRS.
Australian National University
ACT 0200, Australia
Blackawton, P., Airzee, S., Allen, A., Baker, S., Berrow, A., Blair, C., Churchill, M., Coles, J., Cumming, R., Fraquelli, L., Hackford, C., Hinton Mellor, A., Hutchcroft, M., Ireland, B., Jewsbury, D., Littlejohns, A., Littlejohns, G., Lotto, M., McKeown, J., O'Toole, A., Richards, H., Robbins-Davey, L., Roblyn, S., Rodwell-Lynn, H., Schenck, D., Springer, J., Wishy, A., Rodwell-Lynn, T., Strudwick, D., & Lotto, R. (2010). Blackawton bees Biology Letters DOI: 10.1098/rsbl.2010.1056