A Behavioral Test to Examine the Evolution of Color Vision in Vertebrates

Part of the Young Naturalist Awards Curriculum Collection.

by Lena, Grade 9, Maine - 2009 YNA Winner

Our perception of color enables us to see the fire-red of autumn leaves and the deep purple of violets on a rich green bank. It is hard to imagine that some species may not share this ability, while others may see a wider spectrum of colors than we do. It is even harder to imagine that some species perceive a world of color with richer detail than we do. Yet evolution has created these variations among species. Because of modern advances, scientists can analyze the molecular structure of animals’ eyes and use electrodes to probe their color vision. These studies have shown that birds and mammals generally have very different color vision. To understand color vision fully, though, it is essential to learn not only the structure of the eye, but also how the visual system functions. Therefore, the purpose of my experimental project is to learn about the color vision of birds and mammals using a behavioral test of color perception.

It is important to learn about the effects of evolution on the many variants of color vision in different species, not only because it deepens our understanding of human color vision and color vision disorders, but also because knowledge of animals’ color vision can help us in many ways. Learning about animals’ color vision can help explain animal behavior and perhaps solve problems between wild or domesticated animals and people. For example, scientists could wear a color that is hard for a wild species to see and more easily study them in their natural habitat. Fences containing livestock or excluding wildlife could be painted a certain color that the animals could clearly see. Perhaps painting markers alongside roads in a color that wild animals could easily see would decrease accidents. Deer ticks and Lyme disease could be controlled by having wild deer come to feeders where they are sprayed for ticks. Painting feeders with colors easily seen by deer would increase their effectiveness. These feeders could also hold pills to reduce pregnancy so the population might not have to be controlled by hunting.


Figure 1: Evolutionary Tree. Different species evolved to have different numbers of cone cells and different color vision abilities. The common ancestor of mammals and birds had four cone cells, but nocturnal mammals lost two of these. Primates regained another cone cell, and humans now have three cone cells.

The brain perceives color based on two major light detectors in the eye: Cone cells detect color and rod cells detect shade. The number of cone cell types creates the range and detail of color an eye can see. Each type of cone cell absorbs specific colors (wavelengths) of light (Goldsmith 2006). Rod cells allow the eye to distinguish the shades of a color (either light or dark) and help an eye see well when there is little light (Jacobs et al. 1998). Throughout evolution, species have evolved different numbers of cone cell types for their different needs and to aid in food gathering. The earliest vertebrate ancestor of both birds and mammals had four cone cell pigments. Perhaps because they were nocturnal, early mammals did not have a strong need for so many cone cell types, and they lost two during their evolution (Goldsmith 2006). Today, most mammals are dichromats (have two types of cone cells). However, the early primate ancestors of humans regained a third cone cell (they are trichromats) through a mutation, which was probably advantageous because one of their foods was brightly colored fruit. Throughout this time, birds still had four cone cells (they are tetrachromats). Therefore, birds’ color vision is better than humans, and much better than non-primate mammals. Although non-primate mammals lost much of their color discrimination, they gained more rod cells and a greater ability to see different shades of colors (Goldsmith 2006).

For my experiment, I will compare the color vision of a dichromat mammal (sheep) and a tetrachromat bird (chicken). The two cone cells in sheep are the M/L-cone (medium/large wavelength, peak sensitivity = 552 nanometers) and the S-cone (small wavelength, peak sensitivity = 446 nm) (Jacobs et al. 1998). Sheep have many rod cells, with a rod-to-cone-cell ratio of 30–40 to 1. Ungulates such as sheep also have optical structures in their eyes called tapeta, which may help them see certain colors. Tapeta help the sheep’s eyes absorb more light in the dark by reflecting the light back into the eye so it can have another chance to be absorbed by a cone cell. Sheep have yellow or bluish-green tapeta (Jacobs et al. 1998).

This color spectrum shows the different peaks of the two cone cell pigments for sheep and the four peaks for the chickens.

This color spectrum shows the different peaks of the two cone cell pigments for sheep and the four peaks for the chickens.

Chickens have four cone cells typical of birds: the UV-cone (ultraviolet, peak sensitivity = 420 nm), S-cone (small wavelength, peak sensitivity = 470 nm), M-cone (medium wavelength, peak sensitivity = 540 nm), and the L-cone (large wavelength, peak sensitivity = 600 nm) (Osorio et al. 1999). Some birds’ eyes also have features called oil droplets that increase the quality of their color vision by filtering light (Goldsmith 2006).


The purpose of my experiment is to learn more about vertebrate color vision by testing sheep versus chickens using a behavioral assay. Sheep are dicromats (Jacobs et al. 1998) and chickens are tetrachromats (Osorio et al. 1999). I will test color vision by determining if the species can correctly pick between two color options. In Test 1, the positive control, black versus white, will be tested; in Test 2, green and purple, very different; and in Test 3, red and orange: very similar for sheep (Jacobs et al. 1998). If the animals can be trained to get more than 80% correct, it can be concluded that they can distinguish the colors (Hanggi et al. 2007).

The independent variable in my experiment is the different types of animals, chickens versus sheep. I will be able to compare the color vision of the chickens and the sheep. I will also change the colors on my hands for three separate tests. The dependent variable in my experiment is the number of times the animal selects the correct color.

One variable that I kept constant is the amount of experience the animal had with the experiment. Observational learning could have affected the results of the experiment. Also, I used a series of random numbers to switch the colors between my hands so it was not the position being learned, but the color. I put food rewards in both my hands but only gave one to the animal if the correct response was chosen because the animal might figure out which hand to pick by smell. I kept the size and shape of the colored spots constant for each species. Also, the shade of the two colors on my hands were the same to avoid allowing sheep to use their many rod cells to distinguish the shades of the circles even if they couldn’t see the color. I kept constant the amount of food the animal had before the experiment and the lighting during the experiment. Finally, since variations in color vision occur between individuals in a species—due to color blindness or other mutations—I controlled for this by testing several individuals.

Based on the research papers I found on the structure of bird and mammal eyes, I hypothesized that the chickens’ color vision would be better than the sheep, but the sheep would still be able to distinguish between certain colors. I hypothesized that chickens and sheep would both perceive a difference between black and white, green (525 nm) and purple (475 nm), but the sheep would not be able to distinguish between red (625 nm) and orange (575 nm).

Methods And Materials

Six circles of solid colors in a horizontal row from left to right: black, white, green, a dark pink or purplish color, an orangey-red color, and an orange circle.
Figure 2: Sample Colors. These are samples of the colors used in my experiment for the different tests.

The animal subjects I used for this experiment were three sheep (three ewes: 2, 4, and 6 years old) and three chickens (three hens: all 4 years old). I printed out 15 different colored circles that were created using Photoshop Elements. Six of the spots were five centimeters across: one white, one black, one green, one purple, one orange, and one red, all of the same shade (the S value in the color window was 97%) and saturation (the B value in the color

A woman leaning into a pen containing a sheep. She is holding in either hand a black disc and a white disc. The sheep is nudging the white disc.
Test 1: Control with sheep # 2.

window was 96%). I also made six more of the same colors, all eight inches across, and an additional
purple, black, and red circle, (Figure 2) also eight inches across. I got a list of the numbers 1 and 2 in a random order from the web site www.mathgoodies.com. Also, I needed a large supply of sheep grain and cooked noodles as food rewards. I gathered materials for recording data. I collected tape and cardboard to mount the large circles. I set up a pen and a large box to keep the individual subjects isolated while testing. Before beginning the testing, I removed all food from the animal’s pen two hours before testing. I performed three separate tests. In the first, the animals were trained to recognize black versus white. This was a control trial to see how many trials the two

Test 2: Green and purple with sheep #1.

species needed to learn the assay. In the next test, the animals were trained to pick green versus purple, and in the third they picked
between orange versus red.

For testing the sheep, I first isolated the sheep in the testing pen. I filled both hands with grain and held the two different colored circles (eight inches across) in each hand. I lowered my hands into the pen around a corner in the pen so the sheep needed to advance to the colored disks and choose one. If the sheep picked the right color, I fed them the grain and said “Yes.” If they picked the wrong color, I said “No” and removed my hands. This helped the animal understand when they did something right and when they did something wrong. I repeated this ten times until the

Test 1: Control with chicken #3.

sheep understood what they were supposed to do, switching the circles between my hands according to the random number list (1 was right hand, 2 was left hand). Next, I held out two circles of the same color but in a color that wasn’t being taught (for example, black for Test 1). I did this for five trials without feeding the sheep any grain so the sheep would not learn the position of the food (I discovered this to be essential for sheep, which tend to learn position easily). Then I reintroduced the circle that was being taught and recorded the results of 25 trials, with rewards for correct responses. I did this procedure for all three sheep for each of the three tests.

For testing the chickens, I isolated one chicken. Ten trials of initial training were needed for the chickens as well as the sheep, but the five “same color” tests were not needed. I followed the same procedure for the chickens as the sheep except with the reward of cooked noodles and with the

Test 2: Green and purple with chicken #2.

small, five-centimeter circles taped to the back of my hands. Forty trials were needed for the chickens to control for species intelligence. I repeated this for three chickens for three trials each.


My results showed that both species could respond with more than 80% accuracy for the number of trials for their species, based on my control trial (20 or more correct for sheep, and 32 or more for chickens). Both also responded with more than 80% accuracy for Test 2, green and purple. Each chicken also got more than 80% correct for Test 3, but the sheep got around 50% correct.

Comparison of Sheep and Chickens. This table shows the average percentage results for both chickens and sheep for the three tests.


Individual Sheep Results. Note: Sheep #3 originally got 6 incorrect and 19 correct (very similar to her other results). These results were likely thrown off because there was a small dirt smear from her nose on the red circle. I printed new circles, and these results are shown.

Sheep #3 originally got 19 correct of 25, or 76% correct in Test 3. I discounted this data because I believe the reason she got so many correct was because there was a small dirt spot from her nose on the red circle. I printed out new circles and she got only 11 out of 25 correct, and I used this data.

A table listing results of three trials testing three sheep for color differentiation, including a control for black and white, a trial for green and purple and a trial for red and orange. All three animals got at least 80% correct on the first two trials, and 56% or lower on the third trial.
Individual Sheep Results. Note: Sheep #3 originally got 6 incorrect and 19 correct (very similar to her other results). These results were likely thrown off because there was a small dirt smear from her nose on the red circle. I printed new circles, and these results are shown.

I used the chi-square statistical test for each trial to determine how likely it was that the animals were selecting the colors by chance. I used the average of the three sheep and three chickens’

Individual Chicken Results.

correct and incorrect choices compared to the expected number if their choices were random (50%, therefore 12.5 for sheep and 20 for chickens). If the calculated P-value is less that 0.05, statisticians conclude that the choice was not random, and if it is more, the choice was random. For Test 1, the control trial, the P-value was 2.4826*10 -5 for the chickens; For Test 2, the P-value was 3.94016*10 -5 ; For Test 3, the P-value was 5.82612*10 -6 . These values support my conclusion that their choices were not random, and that the chickens could distinguish between all these colors. For the sheep for Test 1, the P-value was 6.73859*10 -4 and for Test 2, the P-value was 6.73859*10 -4 . This supports my conclusion that the sheep could distinguish between these colors. For Test 3, I found that the P-value was 0.94685. This supports that the sheep were guessing randomly and could not distinguish between red and orange.


This shows all the averages and percentages used in the data tables and to calculate the chi-square test. (The chi-square statistical test and standard deviation were calculated on Microsoft Excel.)

Average Numbers and Percentages for Sheep:

  • Test 1: 
    Correct: 22 + 20 + 21= 63/3 = 21.0 
    % Correct: 21.0*100 = 2100/25 = 84.0 +/- 4% 
    Incorrect: 3 + 5 + 4 = 12/3 = 4.0
  • Test 2: 
    Correct: 22 + 21 + 20 = 63/3 = 21.0 
    % Correct: 21.0*100 = 2100/25 = 84.0 +/- 4% 
    Incorrect: 3 + 4 + 5 = 12/3 = 4.0
  • Test 3: 
    Correct: 13 + 14 + 11 = 38/3 = 12.7 
    % Correct: 12.7*100 = 1270/25 = 50.8 +/- 1.5% 
    Incorrect: 12 + 11 + 14 = 37/3 = 12.3

Average Numbers and Percentages for Chickens:

  • Test 1: 
    Correct: 34 + 33 + 33 = 100/ 3 = 33.3 
    % Correct: 33.3*100 = 3330/40 = 83.3 +/- 0.6% 
    Incorrect: 6 +7 + 7 = 6.7
  • Test 2: 
    Correct: 35 + 32 + 32 =  99/3 = 33.0 
    % Correct: 33.0*100 = 3300/40 = 82.5 +/- 1.7% 
    Incorrect: 5 + 8 + 8 = 7.0
  • Test 3: 
    Correct: 34 + 34 + 35 = 103/3 = 34.3 
    % Correct: 34.3*100 = 3430/40 = 85.8 +/- 0.6% 
    Incorrect: 6 + 6 + 5 = 17/3 = 5.7

One possible error in my experiment is experimenter bias. Perhaps the animals had learned to detect some slight, unconscious movement or action I made. This variable could have been controlled if I had been unable to see the colors I was holding. Another possible error could have been that the animals may have had previous associations with the colors I tested. I noticed that the sheep tended to prefer black to white when I was originally training them. Perhaps this is because they have black grain buckets and associate black with food. Also, they seemed to prefer green to purple, maybe because they are browsing animals that eat green leaves and grass. I did not notice these associations with the chickens.


My hypothesis was supported by my data. I hypothesized that the color vision of the chickens would be better than the sheep. I thought that the chickens and sheep would be able to distinguish between the colors in Tests 1 and 2, but that the sheep would not be able to distinguish between red and orange. The results of Tests 1 and 2 show that the animals were not picking by chance and could distinguish between the colors. For Test 3, the chickens’ results showed that they could distinguish between the colors, but the sheep were choosing randomly.

I therefore conclude from my behavioral test that chickens have better color vision than sheep, but sheep can see some colors, which is consistent with the evolutionary model that most non-primate mammals like sheep have lost cone cell pigments and some color vision.

In my experiment, several other questions were raised. I noticed that the sheep might have previous color associations. It would be interesting to try to test color associations in animals and see if they connect colors with other objects or emotions like humans do. I also noticed that the animals got into the habit of the tests, and when I conducted a test the next day, they seemed to remember that I had tested them before. Testing an animal’s memory or desire for routine would also be interesting. I wonder if animals tend to categorize by shape, color, or position.

Experiments like mine are significant because evolution has affected animals’ cone cells and perhaps their color perception. It is important to learn about this evolutionary change not only because it will help us understand human vision better, but also because we will better understand vision in other animals, which may lead to inventions that create fewer problems for animals and humans.


Birgersson, Bjorn, Bjorn Forkman, and Ulrika Alm. “Color Vision in Fallow Deer: A Behavioural Study.” Animal Behavior 61 (2001): 367–371.

Blackmore, T.L., W. Temple, C.E. Sumpter, and T.M. Foster. “An Investigation of Color Discrimination with Horses.” Behavioral Processes 78 (2008): 387–396.

Bowmaker, James K. “Evolution of Vertebrate Visual Pigments.” Vision Research 48 (2008): 2022–2041.

Goldsmith, Timothy H. “What Birds See.” Scientific American 294 (2006): 68–75.

Ham, A.D., and D. Osorio. “Colour Preferences and Colour Vision in Poultry Chicks.” Proceedings of the Royal Society Part B 274 (2007): 1941–1948.

Hanggi, Evelyn B., and Jerry F. Ingersoll. “Color Vision in Horses (Equus caballus): Deficiencies Identified Using a Pseudoisochromatic Plate Test.” Journal of Comparative Psychology 121 (2007): 65–72.

Jacobs, Gerald H., Jay Neitz, and Jess F. Deegan II. “Photopigment Basis for Dichromatic Color Vision in Cows, Goats, and Sheep.” Visual Neuroscience 15 (1998): 581–584.

Mancuso, Katherine, Jay Neitz, and Maureen Neitz. “An Adaptation of the Cambridge Colour Test for Use with Animals.” Visual Neuroscience 23 (2006): 695–701.

Osorio, D., C.D. Jones, and M. Vorobyev. “Colour Vision of Domestic Chicks.” The Journal of Experimental Biology 202 (1999): 2951–2959.

Roth, Lina S., Almut Kelber, and Anna Balkenius. “Colour Perception in a Dichromat.” The Journal of Experimental Biology 210 (2007): 2795–2800.