you eat with your eyes


welcome! this website was inspired by my senior thesis, submitted to the statistics and folklore and mythology departments at harvard. the original thesis is written, but to be honest, i prefer this interactive abbreviated version. because, at the end of the day, don't we all eat with our eyes?

a heartfelt thank you to my advisors ben berman, mark glickman, joseph nagy, lowell brower and many many more for your feedback and support. and like many of my projects, this one is a constant work-in-progress that will continue to grow with more input, interactives and writing.

you can also view more of my writing and projects at or follow my food adventures on instagram and twitter @goudatalks. because i was bored, i've inserted a few easter eggs in this maze of a website. you might have to switch to desktop or change the browser size to find them all.

i'm always down for a coding, design or food collaboration. as you can probably tell, i love meeting new people and cooking both elaborate and unfussy meals. bon appétit!

1: history


  • To date, the world's oldest bowl was found in China’s Hunan province in 2009 by Israeli researchers.
  • Bowls predate plates by several thousands of years. The lack of flatware suggests that people ate from bowls, other receptacles, or with their hands.
  • In the early European medieval period, the standard food vessel was known as a trencher and made from metal or for the poor, thick stale pieces of wood or bread.
  • In the late 16th century, metal trenchers were replaced by pewterware, earthenware and finally porcelain, which closely resembles the rimmed white plates used today.
  • One of the earliest portioned plates, referred to as a divided serving dish, can be traced to Iran, Nishapur in the 10th century.


Materials used to make eating vessels don't stray too far from home. Shaped by the natural landscape and historic traditions, people craft plates to reflect their space. Here are a few:

  • seashells in norway
  • banana leaves for filipino feast kamayan
  • pineapples for bowls in jamaican cuisine

Or assembled without plates

  • injera, fermented teff flour, in eritrean food
  • khobez flatbread to scoop hummus
  • fufu to soak up stews in west africa
  • tortillas substitute for plates in mexico

plating in art

  • Ancient Egyptian tombs carried visual recipes in them. One of the earliest recipe paintings from the 15th century BC showed workers forming cone shaped pastries out of tiger nuts.
  • Closeby in Ancient Rome, artists decorated their walls and floors with mosaics dedicated to food. Depicting fish bones and scattered nuts, Asàrotos ikos or “unswept floor” mosaics were constructed to look like the aftermath of a banquet.
  • After a dry spell of food artwork, due to its association with hedonistic pleasures, the Renaissance revitalized this genre. In 1596, Carravaggio painted a simple basket of fruit. Little did he know this would set off an entire movement to capture the delicate beauty of food.
  • In 1822, Nicéphore Niépce captured what may just be the first photographed dining table. Then in 1845, William Henry Fox Talbot took a photograph of a basket of peaches and a pineapple. The massive influence of still life paintings are felt strongly.


  • In the 20th century, photographers became fed up with still life photographs reminiscent of age-old paintings techniques.
  • In 1927, Edward Steichen projected a cross-hatched shadow onto a wall by experimenting with neatly arranged sugar cubes. Nearly twenty years later, Irving Penn arranged frozen blocks of raspberries, carrots, and corn into a colorful cover for Vogue. (Read more at Feast for the Eyes: The Story of Food in Photography)
  • Cookbooks were usually littered with text, but 1930s cookbooks incorporated more visual elements including illustrations and technicolor photography.
  • In the 1990s, food photography further divorced itself from its painting roots and sought to capture food’s bright and natural qualities for a commercial audience. Food magazines in the late 1900s and early 2000s — Bon Appétit, Gourmet, and Elle — hired still photographers to create realistic and vibrant food scenes and curated a list of stylists specializing in food.
  • In food stylist Victoria Granof’s perspective, food photography’s radical shift happened during the early 2000s. At that point, food photography had not yet reached the masses. Blogs were slowly popping up and phone cameras were in their nascent stages. Recalling a Father’s Day food assignment for Bon Appétit, Granof was surprised that prop stylists were given more creative freedom than the food stylist. “We had to make [the food] perfect and untouched and absolutely symmetrical,” she says. “And the prop stylist came in and had boxes with neck ties and clocks and books.” Food was not the star, but merely a prop to insert next to other props.
  • Everyone is a food plating expert these days. On Instagram, TikTok and other social media, food becomes a dominant, pulsing force that can be sculpted at will.
  • There are even cult followings. Popular Instagram account The Art of Plating (@theartofplating) posts weekly aesthetically pleasing dishes from famous chefs around the globe. According to their website, they commit “to the exhibition of gastronomy as a form of high art – utilizing form, texture and color to tell a story and evoke emotions.”

2: quotes

"If it's familiar, you want to eat it. If it's something you recognize, you want to eat it. I think the first rule is, is it something that I would eat myself?"

- Sue Li,
Food Stylist

"Overhead is not necessarily the best way to shoot it. You know, it's just different angles and different proportions and scales. And it's exciting. [Instagram] was very boring for a while."

- Victoria Granof,
Food Stylist

"Food photography right now is the perfectly imperfect plate of food. The first plate is the plate that everybody likes the most because it's the one that you think about the least."

- Sue Li,
Food Stylist

"We showed the food at that moment where it has the most emotional impact, where it makes you uncomfortable. That's the moment where I said I would rather make pictures that are compelling because they evoke this really strong, emotional response from people than something that's easy to look at and everybody can copy it and it's not special."

- Victoria Granof,
Food Stylist

“I can cut like a machine and I can make every single thing look the exact same shape, but that's not how things are supposed to be now. They need to be more random. They don't want little cubes that are all identical…Now everybody wants crumbs. Everybody wants drips.”

- Lisa Cherkasky,
Food Stylist

“Everything is a reflection of where we are as a society. When it comes to photographs and advertising and everything and food is, you know, they want it to look like you could just take it.”

- Lisa Cherkasky,
Food Stylist

"I've always had it in my blood line. I was in ceramics. I did jewelry when I was a kid. I had all these artist friends from my family, so I just decided I had to start doing that for myself again as an adult. In the past five to six years, I've just kind of been intertwining, both of them and I think it's helped both [skills] thrive at the end of the day"

- Alexandra Motz,
Pastry Chef and Artist

"Because it's not just about what you're seeing, but how those flavors are touching [each other]. It can look great, but flavors can be missed because they're not close to one another.""

- Rachael Collins,
Chef de Cuisine at Juliet

"The importance of critical empty space is true in all art forms, even plating. The difference between the distance between God's finger and Adams's finger in the Sistine Chapel ceiling is pivotal. It can't be any more, it can't be any less. That's an empty space. "

- Jonathan Zearfoss,
Professor at the CIA

"We eat first with our eyes."

- Apicius

"If, with the wolf at the door, there is not very much to eat, the child should know it, but not oppressively. Rather, he should be encouraged to savor every possible bite with one eye on its agreeable nourishment and the other on its fleeting but valuable esthetic meaning."

- M. F. K. Fischer,

3: stats

So how does statistics figure into plating? To see the connection, we must first break down plating, and the anatomy of a perfect plate, into quantifiable chunks.

Let us imagine a plate of dumplings at a dinner table. We've tossed a handful of dumplings randomly onto a plate, like darts on a bullseye. The presentation is unremarkable, but we can only switch the arrangement once. A logical person might spend a few minutes mapping out the ideal plate before rearranging the dumplings once.

But there's a faster way. Turns out, a computer measure and modify the arrangement into a well-balanced and aesthetically pleasing plate. And it will only take a few seconds.

This is an example of an optimization problem. An optimization problem begins when we want to find a *better* result and asking ourselves a few questions. What is the most desirable result? What is the starting point? And how much time, or computing power, do I have?

You'll find people crunching numbers in fields like biomedicine, finance and technology. But optimization problems are everywhere and you can apply similar solutions to any industry (yes, cooking and plating especially).

To solve an optimization problem, we select the best result from a set of alternatives based on some criteria. In the case of plating, we create a bunch of different plating scenarios, measure certain characterics of the final plates, and elimate ones that don't fit the criteria. In more technical terms we must maximize a cost function related to pleasing food arrangements. The final plate is the solution to this optimization problem.

The cost function, which evaluates how well our algorithm works with the given data, includes several criteria. Here's an example of one: We do not want dumplings to overlap or stick together on a plate. To prevent this, we create a general criteria to ensure that the dumplings are evenly spaced out. We compute every combination of dumplings arrangements on the plate, then find the distance between the two, which is then added to the criteria. The program will attempt to avoid situations where the criteria is large.

Let Cd be the sum of distances between all combinations of dumpling objects, n be the number of dumplings, D (d0, 0) be the coordinates for the first dumpling with radius r0, and (d1, 1) be the coordinates for the second dumpling with radius r1.

Even if the dumplings are evenly spaced out, the dumplings could fall off the edge of the plate. If the dumpling is in danger of falling off the plate, an “if” statement ensures that the algorithm will terminate.

Let D be the cost for the dumpling’s distance from the edge of the plate, R be the radius of the plate, n be the number of dumplings on the plate, r be the radii of the dumpling, and m be the magnitude of the position in the polar coordinate plane. To capture the distance from the edge of the plate for all dumplings, we sum their individual values n times.

Given the above criteria, and a few more elements, we can define the overall cost function. The w coefficients help define how much importance we should assign to each term or criteria in the cost function. If the cost function failed to include an edge case or a particular criteria, we would run into negative plating examples as seen below.

If the system attempted to maximize the distance between dumplings, but not maximize the distance of dumplings from the edge of the plate, the dumplings would move off the plate (wa = 0). On the other hand, if the system attempted to maximize the distance of the dumplings from the edge of the plate, but not the distances from one another, the dumplings would not move (wd = 0).

To obtain a plating arrangement, we look to an algorithm called simulated annealing. This is an iterative local search optimization algorithm that uses Markov chain Monte Carlo.

Why this one? A run of the mill optimization algorithm may find an ideal plating arrangement that satisfies the cost function and immediately stop looking for more plating arrangements. Simulated annealing is different: When cycling through possibilities for plating arrangements, it won't settle for best first it finds -- it'll keep searching, ie: an iteration can escape or overshoot the local minimum to explore more options. On the other hand, simulated annealing takes longer to run and involves more complexity than gradient descent.

Here's a second way to imagine a computer generating ideas for plating food. I took nearly a hundred images of dumplings from the internet and measured the distances between dumplings and plate lengths. Then, I fed all this data into a python program to procedurally generate data points about the arrangement: the number of dumplings, the size of the plate, the appearance of soy sauce, the distances between dumplings, etc. And then it places the items on the plate rapidly so that it covers as many examples as possible. For reference, the number of dumplings based on a Binomial distribution and distance measurements use a Log Normal Distribution.

your plate served by a machine!
a folk & myth + statistics thesis by jess eng