Debunking a Shoddy
How can one measure the average angle of gaze of stylized cartoon eyes to be 9.67° with such precision? The conventional representation of cartoon eyes appears in the cereal box characters, in movie cartoon characters, and newspaper comic strips. A careful study of all of these would, I suspect, show that the cereal box characters have about the same distribution of apparent gaze direction as all the others, and most are not intended to sell cereals to children.
|The Lucky Charms
leprechaun looks up.
|The Trix rabbit looks up.|
|The cartoonist didn't get the message.|
Besides, from whatever height you observe eyes in pictures, the eye gaze direction appears the same. It doesn't matter whether you look at the picture from above, below, level, or off to the side. It's the visual illusion of "eyes that follow you". See below.
Not all cereal box characters have eyes drawn the same way. The statistical results of this study included the "fact" that the average direction of gaze of the cartoon characters on boxes of cereal marketed for children was 9.67° downward. Adult cereals, when they pictured people's eyes, had eyes aimed 0.43° upward. Such precise results! How this might be determined for flat drawings and photos on a flat surface was unclear to me until I read the original paper. Then I was astounded at the creative ingenuity of the authors, and their ability to concoct a bogus method that looks "scientific" to the casual reader, but is seriously flawed. 
Here's how the eye angle was determined. The cartoon eye depiction was assumed to be an ellipse with its long axis vertical. Not all cartoon eyes are of this form, but it is most common. Then a side view was constructed to calculate the eye angle. This models the eye as having an assumed depth equal to the height of the ellipse of the cartoon eye. Human eyes are nearly spherical. How can we infer the depth of cartoon eyes that are drawn as ellipses, not circles? Obviously cartoon characters' eyes are not like human eyes.
This is clearly a flawed construction for determining the angle of gaze. They begin with a cartoon eye of elliptical shape and then cavalierly model it as a three-dimensional sphere for the purpose of calculation. What possible justification does this have? You won't find out by reading their paper.
I'm not making this up. Fig. 1 above is from the original paper.
Taking their method seriously for a moment (very difficult to do) I have amended their figure to reveal the error in the calculation from their already discredited eye model. The blue line indicates the optic axis of the eye. This corrected drawing yields a value of the angle of inflection nearly twice as large as their faulty calculation gave. That average value that they calculated as 9.67° should have been more like 18°. But maybe cartoon eyes obey different optical principles from human eyes. Clearly more research is needed to resolve this issue—a fine opportunity for students seeking Ph.D. thesis topics in "food psychology". 
I find it incredible that, so far as I can determine, no one else has picked up on this incredibly flawed model and calculation. (As of June 18, 2014.) Don't people think about what they read?
And do children really stand 4 feet away from the cereal boxes? The ones I observe run all over the aisles and right up to the cereal boxes. Some ride in the shopping cart.
This figure shows their calculations. (It is obviously not drawn to scale.) Using this higher mathematics, they determine that the eyes on "adult cereal" boxes are (on average) aimed at adult eyes 53.99 inches above the floor, and eyes on "children's cereal" boxes are aimed 20.21 inches above the floor. There's that remarkable precision again—four significant figures! Notice how they are careful to include the height of the cartoon eyes above the bottom of the box, adding it to the height of the shelf. Such attention to small details! Especially since their eye angle calculations for the cartoon eyes was wrong by nearly a factor of 2. 
Not that a factor of 2 matters, considering that the fundamental flaws cited above have already doomed this futile exercise. The technical term for such false precision is "gilding the garbage."
Any other errors of methodology seem trivial in light of the whoppers above. But I'll mention one anyway. What about the cereals displayed on the very bottom shelves? The cartoon eyes there will be less than 20 inches from the floor. If the hypothesis of eyes "looking down" had any merit, they would be making eye contact—with a child's feet. 
One could set aside all considerations of eye angle and simply test the hypothesis that cereal companies use cartoon characters on children's cereals because such characters usually have oversize eyes of vertical oval shape. Then do a study of the child appeal of such eyes compared with less exaggerated eye size and shape. But then there's still the nagging fact that the oversize vertical oval eyes are a stylistic convention in most cartoon characters from Krazy Kat to Mickey Mouse, even those that were not intended for, or ever used for, marketing children's cereals.
|Houdini's eyes follow you.|
The whole eye angle hypothesis collapses into absurdity when you realize that the apparent direction that cartoon or photographed eyes seem to be looking is independent of whether you are viewing the picture head on, or from right, left or above. This is the "eye tracking illusion", for which there is a vast literature of psychological studies. The authors of this paper don't mention or allow for this effect. Apparently "food psychology" is a sub-discipline isolated from the body of genuine psychology.
Then there's the influence of the head tilt angle of the cartoon characters. A character with head tilted back will have its eye pupils below the center of the eye orb when looking forward. Shouldn't these researchers have calculated the head angle to correct their values for eye angle? They didn't. A web search of cartoon eye art shows that most cartoon eyes (whether used to market cereals or not) have eye pupils below the center of the eye orb (on average). The researchers didn't mention this fact in their paper.
|Is Mickey looking down,|
or directly at you?
© Walt Disney Company.
Look at this picture of Mickey Mouse, who isn't trying to sell you anything. Are these eyes looking directly at you, or looking down? According to the screwy methodology of the "Eyes in the Aisles" paper (Fig. 1 above) they are looking down.
|Donald Duck, looking forward.|
© Walt Disney Company.
Or consider this cartoon picture of Donald Duck ©. This rare image goes a long way toward answering the "scientific" question about how cartoon eyes work and how (not) to calculate the angle of gaze. Donald is clearly looking nearly straight forward, through the level telescope. Yet the eye pupils are in the lower part of the eye oval, which would, using the above flawed calculation method (Fig. 1), yield a value of negative (downward) angle of gaze, to three significant figures, of course.
But how do children perceive eye angles of cartoon characters? You won't find out by reading the "Eyes in the Aisles" paper. Inquiring minds want to know. Remember that the paper was titled "Why Is Cap'n Crunch Looking Down at My Child?" The words down and child clearly indicate that the eye angle effect on children was the reason for this study.
In any case, if cereal marketers wanted the cartoon characters to make eye contact with children (and with adults as well) they'd be drawn to look directly forward. Then no matter where the boxes were shelved, or where you were standing, the characters would be perceived as looking directly at you.
In this "brand appeal" experiment they find that their subjects reacted more positively to eyes with pupils more centered in the eye orb. These results contradict and invalidate all their fussing about calculated eye angles and shelf height. Never mind that this "experiment" was done with adults, not children. So what conclusion can we draw from this about cartoon characters' effect on children?
This study is singularly lacking one feature required of good scientific research. Where are the control groups? We've mentioned that no study was done of eye angle of cartoon characters in contexts totally removed from marketing products: movies, newspaper cartoon strips etc. No study was done of preferences for cereal boxes at a child's eye level and also on shelves well below eye level. No study was done of children's preferences for differently designed boxes of the same cereal—only adults were included in that study. In fact, I'd have loved to see a study of how children judge where a cartoon character is looking, as a function of various shapes and styles of cartoon eyes and pupil placement.
The folks who wrote this nonsense could have done a real study by placing cereal boxes on shelves with childrens' cereals displayed at a whole range of heights, then ask children suitable questions about the appeal of the characters as a function of box height. They could have even studied the effect of the cartoon character's gaze, up, down, or level. Weren't they clever enough to do such simple studies? Or were they afraid of getting null results? Just speculating.
I'd like to see a "clean" study of how people (children and adults) judge the direction that eyes of cartoon and human images are looking, without the results being muddied by subjective psychological components such as "character appeal". I predict the results would not correlate with the bogus method of eye angle calculations used in the "Eyes in the Aisles" study.
Inquiring minds want to know answers to these burning questions. Of course, as always, "More research is needed."
The fact that cereals aimed at children make incidental eye contact with children has practical implications to the extent that eye contact indeed fosters more positive attitudes towards a product and increases choice. However, though such effects make theoretical sense, there is currently no evidence that we know of to support its occurrence.So they are claiming only "incidental" eye contact, whatever that means. I wonder what "sense" is implied by "theoretical sense". Certainly nothing in this paper supports the conclusion that cereal makers deliberately use cartoon characters drawn to make eye contact with children (or adults, either). Are they claiming that some established theory of psychology (which one?) supports the very conclusion that they are saying they found no evidence for? But in spite of this wishy-washy disclaimer the message being promoted by the many media citations of this paper is that cartoon eyes are intentionally and cleverly designed looking downward, to influence children's preference for cereals. Those who cite this study favorably don't seem to have functioning humbug detectors.
Then there's the trivial and obvious observation that children don't buy cereals, adults do. 
In fact, it may be that few commentators read past the paper's title: Eyes in the Aisles: Why Is Cap'n Crunch Looking Down at My Child? That title is certainly "sensational" in its implications, in contrast to the closing disclaimer of the paper. If the title had been something like "An informal and inconclusive study of the effect of cereal box cartoon artistic styles on consumer choices" would anyone have noticed?
One can't help noticing that the majority of web media accounts of this paper refer to cartoon characters on cereal boxes as "creepy". (Do a web search for "eyes in the aisles" AND "creepy" if you doubt me.) That word is seldom seen describing other cartoon characters such as Mickey Mouse or Donald Duck. (It was not used in the original paper, nor was it used in the Cornell news release.) Not only do media sources copy each other's news, they mindlessly copy each other's language. They delight in colorful and catchy expression rather than hard facts and critical thinking. A Bloomberg Business Week headline reads "Is Cap'n Crunch Staring Straight Into Your Child's Soul?"
|Beware cartoon eyes looking downward.|
© Warner Bros.
If this "study" isn't a deliberate hoax, it certainly is indistinguishable from one. A web search shows that it has been taken seriously by many journalists and commentators who don't hesitate to suggest that it supports the idea that the eyes of cereal box cartoon characters are deliberately styled by unscrupulous cereal marketers to influence children to want certain unhealthy cereals. They may be unscrupulous marketers, but they aren't dumb enough to fall for this parody of "research".
I wonder what would happen if similar cartoons with seductively drawn eyes were on boxes of a supposedly healthful adult cereal, like Brocoli Chex, Liver Bytes, or Spinach Puffs. Would that persuade kids to like these cereals? The authors of this study make the same suggestion in their "conclusion" section:
Importantly, this insight should be utilized by healthier brands to promote healthier choices and potentially encourage healthier food consumption (Wansink, Shimizu, & Brumberg, 2013). Moreover, using spokes-characters on healthy packaged goods targeted toward adults might be a useful way to introduce adults to healthier categories or brands they would have otherwise overlooked during a busy shopping trip (Wansink, 2005).Personally, I doubt that even the most appealing cartoon art is likely to persuade children to like any food that doesn't taste good to them. But that's only opinion; a study is needed.
The real story here is the gullibility of those who repeat and comment favorably on this parody of scientific research, giving it widespread exposure. Do a web search for "cereal cartoon eyes" and you will get thousands of hits, sensationalizing this fatally flawed study, trumpeting it as if it were good science. Most indicate they didn't read the entire paper, but only skimmed the Cornell University press release. They shamelessly elaborated on it, to suit their own preconceived prejudices. Someone should do a proper sociological study of this phenomenon, and also include study of the pathetic comments from clueless readers of these websites.  This is more evidence that a large proportion of the public cannot tell the difference between science, pseudoscience, and outright humbug. Critical thinking seems to be a lost art.
It is shameful how the media swallowed this shabby pseudoscience unquestioningly, as if it were real and important news. They must consider it important, for it came from Cornell University, and the "researchers" are from Cornell and Yale. News reports of this don't hesitate to go beyond the published paper by suggesting that this represents deliberate unscrupulous marketing tactics by those dastardly cereal companies. Where are the critics who ought to be debunking and denouncing this sordid affair?
So what do we learn from this "study". What is new here? The child appeal of cartoon characters is already well established. We also knew that supermarkets often display adult cereals on high shelves and children's cereals on low shelves. This is likely deliberate, but does it increase sales of children's cereals? This study doesn't say. The only new "contribution to science" is a bogus method of calculating eye angles of cartoon characters, which uses faulty geometry and contains a factor of two blunder, making further inferences from that calculation invalid. Then they study 68 adults, concluding that brand appeal is influenced by the "eye contact" of the cartoon characters. They ignore the fact that this has nothing at all to do with the shelf height of the cereal box. And what does this study of adults say about cartoon characters' appeal to children?
The authors state their objectives:
...the current investigation aims to examine whether eye contact by characters may enhance their effects, and whether this effect of eye contact is used in companies' choice of the direction of gaze of cereals targeted at different audiences (adults vs. children).
They fail to make their case for any of their objectives. Why was this paper even published? Why has it received so much attention by the media? I suspect it is because the study seems to vaguely reinforce widespread public prejudices about cereal marketing, and presses the hot buttons of those who look for anything that seems to discredit cereal companies. 
I still hold out the hope that this study will be revealed to be an ingenious hoax. Dr. Wansink has quite a reputation as a clever deceiver in the way he sets up studies of food psychology in the laboratory.  This could be an elaborate way to study people's reaction to a bogus research study tricked up to give the superficial appearance of real research.
You may have noticed that the lead author only has a B.A., while the other two have Ph.D.s. Could those two have actually read the paper, or did they only casually supervise it and then add their names to get publication credit and citations? Just wondering.
 One might also quibble with the issue of just how one defines "adult" vs. "children's" cereals. This study determined that by looking at how the cereal was marketed on television, and judged a cereal "adult" of there was nothing in the marketing that indicated it would appeal to children. In short, children's cereals have cartoon characters and adult cereals do not. Or could it have something to do with sugar content? Or both? And what is the age range that defines "children"? Do teenagers fall in that group? But, I digress.
I recall when I was a child my favorite cereals were Shreddies, Shredded Ralston and Wheat Chex. For a while one of these (I cannot remember which) was marketed as an adult cereal, "not for children", which, I suspect, encouraged kids to sneak a bowl of the forbidden stuff, and many found they liked it. The advertising suggested the taste was too hearty for children. This would have been in the early 1940s. Can any older reader confirm this, or supply an image of the "not for children" advertising copy?
 I only did a rough estimate here, and even that is better than their foolish calculation deserves. The reader is invited to do it more precisely. It only requires high school level trigonometry.
 These three and four significant figure values suggest that the authors are implying false precision for these numbers. Their paper gives these values in data tables in its appendix. The conclusions of the paper, based on these numbers, would not require such precision in any case. And in light of the serious factor of two blunder in their ad-hoc calculation of the eye angle, the whole "logic", and its results, fall like a house of cards.
To give this more analysis than it deserves: What is in that sample of 65? Is it 65 brands, or 65 distinctly different cereals? A brand like General Mills makes many different cereals. Are these 65 cereals marketed for children? What sugar content qualifies a cereal as "for children"? Cereals like Granola and Cracklin Oat Bran have high sugar content, but don't have the cartoon characters or any signs of marketing for children.
The authors do tell us that of the 65 cereals 29 were "adult" cereals and 45 were "children's cereals". (Nine of them must have been for adults and children.) Of the 45, there were 85 spokes-characters on the boxes, but only 57 of these had eyes directed downward. Did they average 85 character's eye angles (as they should have), or only 57? They don't tell us.
Finally, they calculate that the average eye inflection angle was:
Children's cereals: -9.67 degrees with a standard deviation of 8.22. Adult cereals: +0.43 degrees with a standard deviation of 5.27.Whoah! The standard deviations of the averages are comparable to and even greater than the averages. Something's fishy here. That's very bad data, bad mathematics or a badly presented data table. Are they expressing the standard deviations in degrees, or in some other measure? They don't say.
Standard deviations are a measure of the uncertainty in an experimental value—in this case, the uncertainty in the mean (average) angle measure of eye inflection angle. A range within one standard deviation on either side of the mean will include approximately 68% of the data values. Well, it would only if the distribution of the values is normal (gaussian). Is it? We don't know, for the individual data values are not given in this paper. With such a small sample size it is usually difficult to determine the shape of the distribution. There's also the more subtle matter of whether the angle of zero is an arbitrary, or a physically significant, reference value of angle.
Since the calculations already included a blunder of nearly a factor of two in the eye angles, fussing about these matters is a futile exercise. But it does reveal pervasive carelessness on the part of the authors.
The authors have declined or ignored several requests for the original data, which would allow independent checking of their math. Perhaps they think the data too valuable to share with the scientific community. Or perhaps there's something about the data they don't want revealed. Science is supposed to be an open and transparent enterprise, or so we thought.
 Perhaps cereals stocked on the lowest shelves are intended to appeal to mice prowling the supermarket floor after hours.
Seriously, there's no mystery why adult cereals are stocked on high shelves and children's cereals on lower shelves. It is simply easier for people of any height to look at things at their eye level, whether or not the boxes have cartoon characters. This has nothing to do with the angle of the cartoon character's gaze. This simple fact was apparantly not sufficient for those intrepid researchers.
 There are better ways to sell more cereals to children. In the 1940s, when I was growing up, some cereals had cut-out-and-assemble models printed directly on the boxes, back, sides and sometimes even on the inside of the boxes (surely at extra expense). When the model was finished, very little remained of the box (an early form of recycling). Mothers were exasperated to find the cereal box missing before the cereal was eaten. There were models of locomotives and railways, a circus, a frontier town and other nicely designed projects. Different boxes of the same cereal had different models, and if you built them all you'd have an elaborate table top display. Those were what we would now call "adult" cereals, Cheerios Corn Flakes and Kix, as I recall. I remember I'd tell the grocery clerk to fetch me a particular box that I needed to complete a project. Cereals were often placed on very high shelves (near the ceiling) at our small grocery stores back then, and clerks used a long-handled reaching tool to retrieve them for customers. Those were the days when cereals often had small toys packed in each box. Call it a marketing gimmick, but there's no denying that these paper models had educational value for children. All this was before the proliferation of cereals with high sugar content.
 There were, among the inane comments, a few indicating healthy skepticism. Most of these picked up on the fact that eyes in a picture that appear to be looking down, will appear to look down whatever shelf they are placed on and whatever height they are viewed from. Likewise eyes looking directly at you will look that way from any disparity of height, adult or child. This simple fact of psychology was ignored by the "food psychologists" who did this study.
Here's a sampling of comments by those who recognized the inherent absurdity in this study:
This article is either a hoax or these are some of the stupidest people ever to be admitted to Cornell. The perspective of 2D pictures does not change dependent on your position in the room. If the subject in a picture is looking straight at you, it will appear that way no matter where you are in the room (the man in the portrait seems to follow me wherever I go). Same with a downward gaze. If the Captain were looking straight on and you took a step to the left, it would NOT appear as if he is now looking at your right shoulder etc. If he is looking down, he will NEVER make eye contact with anyone in the room, be it a child or an adult on his/her knees. He will always appear to be looking down no matter where you are. This article/research is mind bogglingly bad. The shelf height theory is sound, but it is more likely that kids are attracted to the look of amazement, surprise or “happy” that the image creates and not that it makes eye contact with kids, who are shorter and not normal size adults. In an email to The Cornell Sun, Aner Tal said:
This "research" is complete garbage. The front of the cereal box is two-dimensional, not three. Eyes "centralled" would appear to follow you regardless of your height. Eyes lowered on the character will always appear to be looking at a point below you. To a toddler in the grocery store, the wabbit is staring at the cereal bowl (or our feet).
I learned this in grade 9 art class. The teacher told us then "If you want the subject in your picture to always look out at the viewer of the art, make them look directly at you. The 'eyes' will seem to follow you." She of course was right.
The downward-looking eyes of the rabbit will ALWAYS be looking down to the viewer, whether the viewer is 50 cm tall, or 2.5 m tall. The same applies to photos ...
Everyone stop and think about this, or get a cereal box and test this. If you have Trix or whoever directly across from you, lift the box up above you until the rabbit is looking at you. It will never happen. Here is why: any picture or painting where the subject is looking at the viewer, will always be looking at the viewer, no matter where you move in the room. Likewise for any other direction or gaze. The picture is not in 3D, therefore your location to the view cannot change. The cereal box folks have the rabbit looking at the cereal, and for adults, the rabbit may be appearing to look fondly towards a youngster, but the youngster never sees the rabbit looking at them. Go test it if you don't believe it.
Junk science will always sell so long as it seems to support popular notions. Real science needs to defend its reputation from this garbage.
...the researchers in this case surely can't be serious....
Try crouching down in the supermarket and see if the cartoon characters' eyes ever swivel to make eye contact with you. They won't. Maybe this is some kind of April Fool's gag?
There is a basic flaw in the reasoning here about the angle of the eyes. Even if a head is lower than the box, the eyes that stare out STILL seem to be staring at the observer more than the others. (Try it now by putting you head down relative to your monitor, and you will see what I mean.) Not to mention that according to the article the boxes are already positioned on the bottom shelves, so it makes no sense.
I thought the article was a joke at first, since it was posted April 1. I'm surprised it passed peer review.
I would like to study the soothing effect of snail snot on sore muscles and whether or not it is tied to hallucinations of six-foot tall ducks dressed as Uncle Sam, when used in conjunction with muscle rub and horse liniment. Ya think I can get a grant for that?
Utterly stupid. No matter how you look at a photo or drawing in which the depicted person or character is looking down, it will always look "down" to you from your perspective. It will NOT look as if you're being looked AT. So when the kid looks at a box of cereal, even if it's located at or near his height it will NOT look as if the character is looking directly at him.
Excuse me but this is clearly ridiculous. The picture on the cereal box is two dimensional not three. Thus to a child looking up to the box the face still looks like it is looking down. It looks like it is looking down no matter how you look at it.
... when you're looking at a two-dimensional image, the eyes of the character will always be at the same angle relative to the viewer; the Mona Lisa is always staring at you regardless of where you are in the room. This study is bunk.
We stated in our paper that we don’t know if the eye placement and line-of-sight is intentional or coincidental, but regardless, it’s the eye contact that matters, not whether it’s intentional or not.In plain language this says that the study hasn't established that the cartoon eyes are intentionally designed to make eye contact, or even whether they do effectively make eye contact with children, but somehow that would be important, if true.
Our general mission in the lab is to help promote healthy eating, and the current research aims to encourage that within the cereal domain.So the purpose of the study wasn't a dispassionate study to see whether the style of cartoon eyes can influence children. It was done with the purpose of encouraging healthy eating. Do we smell a lack of objectivity, or even possible bias here?
 I notice that covers of two of Wansink's recent books prominently include his "Ph.D." title with his name. This is not customarily done by real scientists; it is considered ostentatious. Do you ever see the Ph.D. title on books by Stephen Hawking, Isaac Asimov, Richard Feynman or Carl Sagan? No. It used to be a joke that the only books that flouted the author's Ph.D. were popular books on slightly disreputable subjects like sex and self-help. Maybe we should include books on nutrition.
The object of the "Eyes in the Aisles" paper was to examine the effect of cartoon character's eye angle on children's cereal preferences. Much of the paper is devoted to references to other published studies that are of only marginal relevance to the object of this study. The extensive reference list at the end of the paper includes six to Wansink's own publications.
Looking at this logically, there seem to be three hypotheses about the "Eyes in the Aisles" paper.
Figures 1, 2, and 3 are from Eyes in the Aisles: Why Is Cap'n Crunch Looking Down at My Child? Environment and Behavior, April 2, 2014.
Input and suggestions are welcome at the address shown to the right. When commenting on a specific document, please reference it by name or content.