VOL. XVI, No. i JANUARY, 1939

NEW EXPERIMENTS ON COLOUR VISION IN BEES

BY MATHILDE HERTZFrom the Subdepartment of Entomology, Zoological Department, Cambridge

(Received J April 1938)

(With Two Text-figures')

I. INTRODUCTION

A METHOD to ascertain whether a given material appears coloured for the hive bee,or appears merely neutral grey or white, has been described in a previous paper(Hertz, 1937 a). The procedure adopted may be summarized as follows. The beeswere first trained to visit a feeding table placed in the open air in normal daylight.The table was covered over with a deep black paper. The material which it wasdesired to test was cut into the shape of a simple square or circle, about 6 cm. indiameter, and placed on the black background. The whole.surface was then coveredwith a sheet of glass which was known to transmit a high proportion (100-80%)of the longer ultra-violet rays, besides the visible spectrum. By placing a largedrop of sugar solution on the glass over the material to be tested, attempts weremade to condition the bees to visit the samples and to continue to do so when nosugar solution was present.1 The result of such experiments was that the beesreacted to samples of white of whitish paper of the ordinary kind made from woodpulp in the same way as when trained to visit blue or yellow papers; they easilylearned to visit these samples exclusively. If other kinds of white or whitish paperwere used, as for instance good writing paper or barytes paper, the bees continuedto distribute themselves indiscriminately over the table; attempts to conditionthem gave no result at all. When the optical qualities of the two kinds of paperwere investigated, that kind which, on a black background, proved to be completelyunattractive, was found to reflect the whole of the spectrum visible for bees(6500-3000 A.). Such papers apparently are white for the vision of the bee justas they appear white for man (reflecting the red rays as well). The kind of whitepaper that proved to be attractive was found to reflect only the visible part of thespectrum and comparatively very little ultra-violet rays. I assume such papers tobe coloured for the bee for the same reason that white light becomes coloured forman if red or any other of the visible rays are excluded from the mixed light.These white papers should appear for the bee identical in colour to a definitesection of the visible spectrum, namely, that one which might represent the com-

1 It is impossible to repeat here all the precautions usually taken in experiments with bees toavoid erroneous conclusions, they are to be found in earlier papers. A description of the techniqueand the methods of the present experiments is given in Hertz ( ^ )

JEB-XVli

2 MATHILDE HERTZ

plementary colour to ultra-violet. Against my expectations I did not succeed,when I tried to get the bees to confuse the training colour "white devoid of ultra-violet", with any hue of the twenty-four full-coloured papers of the Ostwald naseries. Thus a final explanation regarding the complementary colours for beescould not be given at that time.

II. EXPERIMENTS WITH COLOURED PAPERSI began a new series of experiments, when the following explanation of the

results just described occurred to me: It is impossible for the eye of man to controlthe amount of ultra-violet, which may or may not be reflected by any given surface;and it may be that some of the colour papers used in the earlier experiments reflectultra-violet in addition to their range of the visible spectrum. If it is true thatultra-violet and blue-green are complementary colours for the bee—as has beensuggested by Kiihn (1927)—a blue-green paper, reflecting ultra-violet also, mustappear neutral grey or whitish for them. Whether this is actually so in the Ostwaldpapers can be ascertained by the method already described. In the full-colouredna series of Ostwald the blue-green papers range between No. 16 (oxide blue) andNo. 22 (emerald green). Samples of each of these seven colours were put, oneafter the other, on the black background and attempts were made to conditionthe bees to visit them. As a cover to keep the papers clean a glass platethat transmits the whole of the solar spectrum was used.1 There was nodifficulty in attaining regular visits to Nos. 16 and 22, representing blue andgreen. Papers 17 and 21 required a much longer training and the visits wereless regular and reliable. In the case of the blue-green papers, Nos. 18, 19 and20, prolonged attempts to condition the bees gave no result at all. UndoubtedlyNos. 16 and 22 are brightly coloured for the bee, Nos. 17 and 21 very much less so,and Nos. 18-20 appear for them only greyish or whitish. The most probableexplanation is that the three last-mentioned blue-green papers reflect not onlyvisible light but also a large proportion of ultra-violet, and that Kuhn was rightin assuming that blue-green and ultra-violet are complementary colours for thebee. If this is actually the case, the behaviour of the bees must change completelyif the ultra-violet reflexion be cut off by means of filters. I had three slightlydifferent filters at my disposal (square glass sheets accurately fitting the papersamples).2 When any of the blue-green colour samples hitherto neglected by thebees were covered by a filter glass which absorbs ultra-violet, the behaviour changedaccording to expectations. There was no longer any difficulty to get the beesconditioned to pay regular visits to the coloured surfaces.8 The main and con-clusive experiment was now made. When the bees had been trained to visit the full-coloured blue-green paper na 20 covered by a filter, a sample of a whitish paperreflecting very little ultra-violet (made from wood pulp) was placed in competition

1 "Uviolglas" supplied by Schott & Gen, Jena.1 Supplied by Schott & Gen, Jena.• It is necessary in experiments of this kind to cover carefully the edges of the little filter glasses,

where their own colour, blue-green, visible only for bees, must show in a high degree of saturation.The arrangement of filters and papers is kept clean by means of an Uviolglas.

New Experiments on Colour Vision in Bees 3

with the training colour on the black background. Since the bees visited bothsurfaces with approximately the same eagerness, they gave ample proof that whitedaylight, devoid of ultra-violet or, to put it differently, the mixed light of the"visible " spectrum, is identical in colour for them with a more or less homogeneousblue-green. The same confusion took place when white papers, reflecting a highproportion of ultra-violet, were placed in competition with blue-green na 20,both samples being covered with filter glasses. It is evident, therefore, that inthe vision of the bee ultra-violet and blue-green must be complementary colours.

Bees which are well trained to visit the blue-green and white papers—bothdevoid of ultra-violet—will not visit competing samples of yellow and blue(Ostwald na Nos. 1-4, 13-16 and 22-24).1 The light reflected by the ultra-violet

3000 4000 5000Wave-length (A)

6000

Fig. 1. Spectrum reflected by the blue-green Ostwald paper na 20. I, II, III, IV are thefour regions of the spectrum separately discriminated by the bee. Wave-lengths in A.

absorbing glasses themselves must add a tinge of blue-green to the colour of anyunderlying surface. It is therefore important to know that the bees readily dis-criminate blue-green and visit it exclusively when the competing colour samples arealso covered with filters.

In order to know the accurate range of the wave-lengths effective in the blue-green Ostwald paper na 20 spectrographic measurements were carried out on thatpaper.2 The curve representing the percentage of reflexion for the different wave-lengths is reproduced in Fig. 1. Kiihn, when using monochromatic light in trainingexperiments, found that the blue-green which represents for the bee a principalcolour, apart from blue and green (yellow), ranges exclusively between 4800 and5000 A. The fact that the two maxima of reflexion represented in Fig. 1 appear

1 The percentages of reflexion in the visible light is known for some of these colours (Hertz &Imms, 1937).

1 By Messrs Adam Hilger Ltd., London.

1-a

4 MATHILDE HERTZ

exactly at 4900 and 3600 A., confirms as well as possible the previous results andthe conclusions so far made.1

In order to know the apparent darkness or lightness of the blue-green paperna 20—which represents a mixture of two pairs of complementary colours—thefollowing experiment was made. For the purpose of training the bees a sample ofthat paper was put on a white background. Since not only colour but also relativedarkness attracts their attention, the bees will learn more easily to pay regularvisits to the sample on a white background the darker it is to their vision (Hertz, 1931).Apparently the paper was rather light for the bee, because prolonged training didnot give a positive result. In order to improve the training conditions by making thepattern more attractive the square paper sample was cut into strips 6 mm. in widthand the strips were arranged in such a way on the white background that a patternof cross-bars resulted. The increase in outlines had the effect that was to be expected.The bees began to visit the pattern regularly, but the reactions continued to be veryslow. It is evident, therefore, that the paper na 20 is just dark enough to be dis-tinguished by the bees from the white background; it represents a very light grey.

All these experiments prove that ultra-violet light does not affect the eye of thebee independently and apart from the rest of the spectrum, but is only one normalcomponent in a system of four principal colours. Since this is a matter which hasbeen misunderstood in some of the earlier investigations it is perhaps desirable toadd another proof. On the usual black background a square sample of good whitepaper which reflects a high percentage of ultra-violet is placed. This is coveredwith a well-fitting sheet of black glass, known to transmit only ultra-violet light2

(3600 A. to 80%). After the bees have become well trained to visit the black (orultra-violet) square, the white paper is removed from under the filter and placedby the side of it. A square sample of blue-green na 20, reflecting ultra-violet asshown in Fig. 1, is also put on the black background. The bees returning to theiifeeding table show, by their behaviour, that they completely miss the colour towhich they had been conditioned. They hesitate to settle down and eventuallydistribute themselves indiscriminately over the table. The ultra-violet reflectedfrom the white paper to a percentage even higher now than during training givesno longer the sensation of that colour because it is mixed with the whole range ofvisible light. Ostwald na 20 proves to be neither ultra-violet for the bee nor blue-green. The black filter,- devoid now of a background that reflects ultra-violet,does not attract the bees because they are conditioned to visit an ultra-violet surfaceand not a black one.

III. EXPERIMENTS WITH FLOWER PETALS

No flower seems to exist of a brilliant blue-green, but white flowers are abundant.The spectrum of the light reflected by the white petals is known for a number of

1 If the examination of paper na 20 had revealed that the blue-green colour of this paper wasdue only to a complete absorption of red rays, the last experiment would have proved nothing butthat the bees respond alike under equal conditions; daylight devoid of ultra-violet and daylightdevoid of ultra-violet and red are identical for bees.

* Supplied by Schott & Gen, Jena.

New Experiments on Colour Vision in Bees 5

American and European species, so that the question whether or not they arecoloured for the bee can be answered immediately (Lutz, 1924; Lotmar, 1933). Allof them should appear brightly blue-green, because the amount of reflected ultra-violet is irrelevant compared with the high percentage of reflected visible light. Thisis easily confirmed when samples of white petals are placed on the black backgroundof the feeding table in competition with suitable papers. Bees that have been trainedto visit white papers devoid of ultra-violet will visit indiscriminately any whiteflower petals picked at random. If the number and eagerness of the bees be takenas a measure, the attractiveness of the petals as a rule exceeds the effect of the papersamples. In some cases, at least, this must be due to a very high amount of reflexionof the visible light and a nearly complete absorption of the ultra-violet rays. Oneinstance is Convolvulus septum, the spectrum of which is given by Lotmar as:5400 A. = 90 %, 4350 A. = 80 %, 4040 A. = 18-5 %, 3650 A. = 2-8 %, 3130 A. = 3 %.

But the number of flowers which, for bees, are apparently either identical, ormore or less similar, in colour to 4900 A. are not limited to species which are whitefor the human eye. Bees which visit the white petals of Convolvulus, Phlox,Campanula, Althaea and Rosa will visit other flowers of the same genera which,for the human eye, are faintly but distinctly coloured pink, purple, blue or yellow,along with any other specimens of faintly coloured flowers (for instance, Knautiaarvensis). The explanation is easy to give: the pale colour effects, produced in suchcases in the human eye, are made by rather slight differences in the amount ofreflexion concerning the different parts of the visible spectrum. It is obvious thatthese slight differences do not matter very much in the case of the bee as comparedwith the necessarily very strong effect of a nearly complete absence of ultra-violet.The spectrum of Lavatera trimestris (Malvaceae) may serve as an example of apink flower that is expected to be confused by bees with white ones: 6000 A. = 40 %,5460 A. = 30%, 4350 A. = 47%, 4040 A. = 49%, 3650 A'. = 6-2%, 3130 A. = o%(Lotmar, 1933).

The only flower petals which were never visited by bees that had been fed forsome time over samples of white paper lacking ultra-violet, or over white petals,were those which show a saturated orange, yellow, blue, dark purple or red.

We now select from the light, very faintly coloured ea series of Ostwald papersthose which closely resemble, for the human eye, the faint flower colours justmentioned. They are placed in competition with petals of similar colour on thefeeding table. It becomes at once apparent that there is no likeness between thepale flower colours and the pale paper colours; this must be due to a comparativelyhigh amount of ultra-violet reflected by the papers.

IV. DISCUSSIONThese results and arguments affect, in many instances, the interpretation of

previous experiments. Lutz, in his attractive training experiments with Trigonacressoni, was the first to notice that there are two kinds of white which appeardifferent for bees; but he is wrong when assuming that for these insects a whitesurface which reflects ultra-violet rays besides the visible spectrum looks anything

6 MATHILDE HERTZ

like ultra-violet (Lutz, 1923). An important point to consider is that, in most ofthe experiments on colour vision in hive bees, sheets of glass of unknown qualitieshave been used to. cover the training table. It therefore might be useful toanalyse more closely the limiting factors imposed by the qualities of such glasses.In my recently recorded experiments (Hertz, 1937 a) I found a good whitepaper, with the known reflexion of 60-80% of visible light and 50% of ultra-violet (3600 A.), to give no colour effect at all under a sheet of glass transmittingabout 80 % of 3600 A. It follows that a mixture of light, in which 60-80 % of thevisible rays corresponds to 32 % ultra-violet (o-8oa x 0*50), is not definitely colouredin the vision of the bee. On the other hand, the bees gave positive colour reactionsto another white paper which reflected 50-80% of the visible light and about2 5 % o r 3°°° A., and was placed under the same sheet of glass. This means that amixture of light, where 50^-80% reflexion of the visible light corresponds to 16%reflexion of ultra-violet (o-8oa x 0-25), is no longer white but coloured for the bee.In many previous experiments glasses may have been used that did not transmitmore than 50% of 3600 A. (Lotmar, 1933). By using the above-mentioned goodwhite paper under a glass of this last kind the light ultimately effective represents60-80% reflexion of visible light corresponding to 12*5 % reflexion of ultra-violet(o-5O3), a mixture that cannot be otherwise than coloured for the bee. It followsthat in many instances, when investigators employed papers which they believedto be white or grey, they were actually using coloured papers, owing to the inter-ference of the covering glass. It is perhaps well worth while to consider the funda-mental experiments of von Frisch (1914) from this point of view. His bees, afterhaving been trained to visit blue or yellow papers, never confused these colourswith many different shades of grey. But confusion between the training paper andthe grey papers took place when they had been fed on blue-green (Hering papers).He drew the conclusion that the blue-green papers were not coloured for the bee,but appeared grey. He certainly was correct if in his experiments he used a sheetof glass which transmitted a high percentage of ultra-violet light. If instead heused a sheet of glass transmitting comparatively little ultra-violet the confusionmust have been due to the fact that both the blue-green and the grey papersappeared definitely and equally coloured for the bee. The principal result ofvon Frisch's experiments and his conclusion that the bees discriminate blue andyellow colours by the wave-lengths they represent, not by their shade or darkness,is, of course, not affected by this doubt. The reasoning remains unaltered whethera certain colour is competing with many different shades of grey or of any othercolour. I am myself aware of having once used a sheet of glass transforming, forbees, grey into colour (Hertz, 1932). As others had done before me, I tried tocondition bees to visit exclusively a definite shade of grey, but noticed that afterprolonged training the bees became attracted far more to the reverse side of anypaper where the whitish wood pulp material was not covered with paint than to theactual training paper. I suspected a hidden colour effect, but I could not givethe right explanation at that time. Now it is evident that the grey training paperappeared more or less coloured to the bee. After being conditioned they gave

New Experiments on Colour Vision in Bees 1:preference to that paper in which the training colour appeared most brilliant. Itwould be wrong to suppose, on the other hand, that all grey surfaces must appearcoloured under glass plates of this kind. It might be that some surfaces, which arewithout definite colour to the human eye, reflect a surplus of ultra-violet and thata glass plate by absorbing part of the ultra-violet rays serves to destroy the ultra-violet colour effect without causing the complementary colour to appear.

Kiihn has been the only investigator to use monochromatic light in training beesand the first to ascertain that the ultra-violet and blue-green rays are discriminatedby their wave-lengths (Kiihn, 1927). One of his experiments is very suggestive.He trained bees to visit a strip of blue-green light (4900 A.), thrown on to a table;afterwards he crossed that strip with a strip of mixed white light. The bees showedthemselves well conditioned by visiting the blue-green light exclusively. In re-peating this experiment one ought to place a filter, absorbing ultra-violet rays,

Yellow Blue

" 50 A9 -18 4 5

Blue-green

Fig. 2. Provisional diagram representing colour vision in the bee. Wave-lengths in A./ioo.

into the beam of the mixed light. If I am correct, the bees would cease immediatelyto discriminate between 4900 A. and the white strip of light.

The system of colours as perceived by man is usually given in the form of atriangle, the angles of which represent red, green and blue. Lines connecting theloci of complementary colours meet in a point in the centre that represents white.We may suppose that the system of colours as perceived by the bee can be repre-sented in a similar way. The colour triangle shown in Fig. 2 represents the facts asfar as they are known to-day. How far the diagram represents the truth can beascertained only in experiments with mixtures of monochromatic light. One mostinteresting question to solve is whether there is any definite likeness between thelongest and the shortest rays in the colour system of the bee.

If future experiments would disprove a similarity between 6400 and 3200 A.,and further, if it would be impossible to get the full effect of 4900 A. by mixinghomogeneous yellow-green and greenish-blue rays, a simple triangle could nolonger serve to represent the colour system of the bee.

8 MATHILDE HERTZ

V. SUMMARY

Under the condition of full, normal daylight illumination a surface whichreflects a fair proportion of ultra-violet, as well as the visible spectrum, is neutralor white in the vision of the bee. If the percentage of reflexion of 3600 A. is lessthan one-third, or about one-quarter, of the percentage of reflexion in the visiblelight, a degree of coloration is obtained which is sufficient to be noticed by the beesin training experiments.

When both are placed under ultra-violet absorbing niters, white paper in thevision of the bee most closely resembles a blue-green paper that possesses thehighest amount of reflexion at 4900 A. Bees which have been trained to visit ablue-green surface covered by a filter glass do not do so any longer, if, by removingthe filter, ultra-violet light is added to the reflexion. The paper that possesses nowtwo peaks of maximum reflexion, one at 4900 and the other at 3600 A., appears alight grey to the bee. On the other hand, when bees have been trained to visitan ultra-violet surface—white paper under a filter that absorbs the visible lightcompletely—do not do so any longer if by lifting the filter the whole range ofvisible light is added to the reflexion of ultra-violet. The white paper at oncebecomes unattractive to the bee.

There remains no doubt, that among the four qualities of colour discriminatedby the bee (see Fig. 1) the first and the third on one side and the second and fourthon the other are complementary colours for this insect and presumably for manyothers.

In the European flowers visited by bees three principal colours are now finallyknown: (1) that colour which is perceived if the main bulk of light reflected by thepetals lies between 6500 and 4900 A. (orange-yellow for bee and man), (2) thatwhich is perceived if the reflexion'extends from 4900 to 4000 or 3500 A. (blue-violet for bee and man), and (3) that which is perceived if the reflexion extends from6500 A., or nearer to the red end of the visible spectrum, to 4000 A. (blue-greenfor the bee, but white, pink, light purple, bluish or yellowish for man).

Ultra-violet seems never to occur in European flowers in such a way as to makepetals which are brilliantly white for man to appear equally white for the bee.

I am very much indebted to Dr A. D. Imms whose help made these in-vestigations possible.

REFERENCESv. FRISCH, K. (1914). Der Farbenthm und Formensitm der Biene. Tena.HERTZ, M. (1931). Z. vergl. Physiol. 14, 629.

(1932). Biol. Zbl. 52, 436.(1937a). Z. vergl. Physiol. 24, 413.

~ (i937*)- Z. vergl. Pkytiol. 2S, 239.HERTZ, M. & IMMS, A. D. (1937). Proc. Roy. Soc. B, 122, 281.KOHN, A. (1927). Z. vergl. Physiol. 5, 762.LOTMAR, R. (1933). Z. vergl. Physiol. 19, 673.LUTZ, F. (1923). American Museum Novitatts, No. 641.

(1924). Am. N.Y. Acad. Set. 29, 181.