What's in a Face? Are Composite Faces More Attractive than Real Faces?

Note: the website faceresearch.org, on which this project was originally based, is no longer available. However, artificial intelligence (AI) image generators are now widely available. AI image generators can generate lifelike images from large amounts of training data consisting of real pictures, so they can serve a similar purpose and generate "average" faces. If you want to do this project, search online for an AI image generator and try using it to make your composite faces. You may want to try out several websites to see which one gives the best results.

Creating the Individual Images

1. Go to the Face Research website at www.faceresearch.org/demos/average.
   1. Note: You do not need to sign up for an account on the Face Research website to do this science project.
2. Use all female or all male faces for this procedure. Click on one individual face to select it. The selected face will have a highlighted red square around it.
3. Click View Average.
4. A larger picture of the individual selected face will appear.
5. Take a screenshot of your screen using the print screen feature. Open the image in a photo editing program and crop it so that the larger picture on the left (of the individual selected face) is the only part of the image showing. Save the new image (as a .jpg file) in a folder on your computer where you can locate it.
   1. If you are using a PC, you can use a photo editing program such as Paint to open the screenshot and crop it (by using the crop feature).
   2. If you are using a Mac, you can use a photo editing program such as Paintbrush to open the screenshot and crop it (by changing the canvas size).
6. You will now need to import the image from your folder into Microsoft Word or other word processing software.
   1. To maximize space, you might want to set your paper setting to "landscape" (instead of "portrait"), which you can do in your program's page setup margins menu. In Word, you can find this by clicking "File," then "Page Setup," and selecting "Landscape" in the "Margins" tab. For help doing this in another program, find a help page online or ask someone familiar with your word processing software for help.
   2. To import in Word, click on "Insert" at the top of the program, scroll down to "Picture," then select "From file." Find the image on you computer and click "Insert." The image should now appear on the page. For help doing this in another program, find a help page online or ask someone familiar with your word processing software for help.
7. Repeat steps 2–6 with two more individual images. Remember to use all male or all female images. In the end, you should have a total of three different female images or three different male images.
8. Import the images and position them next to each other on a single page in the word processor document. Note: If the images are too big to all fit on one page, you can click on each image, grab one of the tiny boxes in one of the corners with your mouse arrow, and drag it to the size you want.

Making Composite Faces

1. Within the Face Research website, click on all three of the faces you chose from above. They should all be highlighted. See Figure 2, below.

Image Credit: Face Research Org / Copyright University of Aberdeen

The "Make an Average Face" webpage includes a large headshot of a person on the left side of the screen, and a 7 by 7 grid of smaller headshots directly to the right. Headshots on the right can be selected, and the headshot on the left will combine each selected headshot into a composite image.

Figure 2. This is a screenshot from the "Make an average face" website. Users select faces from the thumbnails on the right, then click View Average to make a composite image.

2. Click View Average.
3. A composite face will be generated by the Face Research software. Right-click on the composite face.
4. Make and save the image of the composite face, as you did with the individual faces (by taking a screenshot and cropping it).
5. Import the image of the composite face into the same word processor page as the other three individual faces are on.
6. Label the faces 1A, 1B, 1C, and 1 average on the page by typing underneath each one.
7. Repeat the previous section and steps 1–6 of this section to make three more composite faces. Each composite should be made from three individual faces. The faces should be different from any of the ones used previously.
8. Paste the images into new pages in your word processing document.
9. Label the second set of faces 2A, 2B, 2C, and 2 average.
10. Label the third set of faces 3A, 3B, 3C, and 3 average.
11. Label the fourth set of faces 4A, 4B, 4C, and 4 average.

Creating Pages for Testing

1. Open a new word processing document, which will be the one you print out for your test subjects. For each set of faces, make one new page with just one of the individual component faces, as well as the composite face.
2. Label the pages Page 1, Page 2, Page 3, and Page 4 by typing the text at the top of the page. Be sure to type as near to the top of the page as you can so the frame you will be making does not hide the text.
3. Label the two pictures on each page A and B. Again, be sure the labels are below the images enough that the frame you will make does not hide the text. See the final result in Figure 3, below. Mix up the order so that the composite is not always A or B. Note: The black frame you see here will be explained in the next section.

Image Credit: Face Research Org / Copyright University of Aberdeen

Figure 3. Make four pages, each with a composite face and an individual face that was used to make the composite. Label the pages 1–4, and label the images A or B. Note: The frame you see here will be explained in the next section.

4. You should now have four pages—a page for each individual composite, accompanied by one of the individual component faces that is part of that composite.
   1. Note: The test could be done comparing a composite face to a face that was not used as a component face, but either way, it is important to be as consistent as possible. For this procedure, compare a composite face to a face that was used as a component to make the composite.

Preparing the Pages

1. You can see on your screen that the composite face has a faint "halo" where the hair from the individual faces has been digitally combined. It would be distracting if all of the hair styles were averaged, so this part of the composite has been de-emphasized (that is, made faint in the image). You can remove this halo effect by making a frame that just shows the facial features.
2. Print out the four pages of images. Now use the scissors, glue, and four sheets of black construction paper to create frames to hide the halos.
3. Arrange the frames over the pictures so that the frames cover the halo on the composites. See Figure 4, below.

Image Credit: Face Research Org / Copyright University of Aberdeen

Figure 4. The composite image (A) has a "halo effect" derived from the hair of the component faces. The black oval frames hide the halo effect on the composite image. Make four pages like the one on the right, each with a composite face and an individual face that was used to make the composite.

4. You will now need to make copies of each page, which you can do at your local drug store, library, or your school. Make enough to either hand out to all your test subjects, or to show them and have them answer on the spot (if you do this, be sure no other test subjects are nearby as they are answering).
5. Staple the sets of pages together, in order, from page 1–page 4.

Testing

1. Have your lab notebook and pieces of paper for your test subjects to write their answers on ready.
2. However you decide to test your subjects (whether by handing out your pages to each of them, or by showing them the images and recording their answers on the spot), inform them of the following:
   1. Thank you for agreeing to take part in this human behavior science project.
   2. Look at each of the four pages, numbered 1–4.
   3. Each page has two faces. The faces are labeled A or B.
   4. Write down on a piece of paper Page 1, Page 2, Page 3, and Page 4.
   5. Next to the page number, write down the letter of the face that is more attractive, in your opinion.
   6. Give the list with your answers to me when you are done.
   7. Your final list of answers should list either A or B for pages 1–4.
   8. Please provide your answers within 24 hours.
   9. Thank you again for helping with this science project!

Analyzing Your Data

1. Count the total number of comparisons you received.
2. Count the number of comparisons in which the composite face was determined to be more attractive.
3. Count the number of comparisons in which the individual face was determined to be more attractive.
4. Make a data table in your lab notebook showing your results.
5. Is there a clear preference for the composite face or the individual face? Or is there no preference either way? If there is a difference, how can you determine if it is due to a real preference and not just chance? To confirm your conclusion, you can use a statistical test called the Chi-square test. The Chi-square test is often used to decide if the data observed during an experiment met the expected results. Categorical data is any data divided into named groups; for example "more attractive" vs. "less attractive," as opposed to represented by numerical measurements, such as a person's weight or height. Use the references in the Bibliography to familiarize yourself with the Chi-square test.
6. Use the online Chi-square calculator from GraphPad to help you calculate the result.
7. For the test, the expected number of comparisons that found the composite face more attractive is one-half of the total number of comparisons.
   1. This is what you would expect if there is no preference.
   2. The observed number of comparisons that found the composite face more attractive is the result you recorded in the data table.
8. Use the Chi-square test to determine the probability that the numbers you found in your test were due to chance. The point of the Chi-square test is to determine how likely it is that your results are due to chance. It will allow you to say something along the lines of "There is only a 1 percent probability that my results are due to chance. Therefore the results are statistically significant." Many researchers view any result with a probability of 5 percent or below as statistically significant.
9. Based on your data analysis, what can you conclude about people's responses to composite faces compared to real faces?