Skip to content

Blast from the past

Paul Alper points us to this news article, The Secret Tricks Hidden Inside Restaurant Menus, which is full of fun bits:

There is now an entire industry known as “menu engineering”, dedicated to designing menus that convey certain messages to customers, encouraging them to spend more and make them want to come back for a second helping.
“Even the binding around the menu is passing us important messages about the kind of experience we are about to have,” explains Charles Spence [whose recent book Gastrophysics: the New Science of Eating], a professor in experimental psychology and multisensory perception at the University of Oxford.
“For a large chain that might have a million people a day coming into their restaurants around the world, it can take up to 18 months to put out a menu as we test everything on it three times,” says Gregg Rapp, a menu engineer based in Palm Springs, California
Perhaps the first thing a customer will notice about a menu when the waiter hands it to them is its weight. Heavier menus have been shown to suggest to the customer that they are in a more upscale establishment where they might expect high levels of service.
A study conducted by researchers in Switzerland found that a wine labelled with a difficult-to-read script was liked more by drinkers than the same wine carrying a simpler typeface. Spence’s own research has also found that consumers often associate rounder typefaces with sweeter tastes, while angular fonts tend to convey a salty, sour or bitter experience.
“Naming the farmer who grew the vegetables or the breed of a pig can help to add authenticity to a product,” says Spence.
A study from the University of Cologne in Germany last year showed that by cleverly naming dishes with words that mimic the mouth movements when eating, restaurants could increase the palatability of the food. They found words that move from the front to the back of the mouth were more effective – such as the made up word “bodok”.
Dan Jurafsky, a professor of computational linguistics at Stanford University, performed a study that analysed the words and prices of 650,000 dishes on 6,500 menus. He found that if longer words were used to describe a dish, it tended to cost more. For every letter longer the average word length was, the price of the dish it was describing went up by 18 cents (14p).
“When we[Rapp] do eye tracking on a customer with a menu in their hand, we typically see hotspots in the upper right hand side,” he says. “The first item on the menu is also the best real estate.”

But filling a menu with too many items can actually hamper choice, according to menu design experts. They say offering any more than seven items can overwhelm diners. To overcome this, they tell restaurants to break down their menus into sections of between five and seven dishes.

“More than seven is too many, five is optimal and three is magical,” says Rapp. There is some research to back this up – a study from Bournemouth University found that in fast food restaurants, customers wanted to pick from six items per category. In fine dining establishments, they preferred a little more choice – between seven and 10 items.

“The problem with pictures is that the brain will also taste the food a little bit when it sees a picture, so when the food comes it may not be quite as good as they imagined,” warns Rapp.
In recent years, Pizza Hut began testing eye-tracking technology to predict what diners might want as they scan through 20 different toppings before offering a likely combination to the customer.
But the article is outdated
This article was originally published on November 20, 2017, by BBC Future, and is republished here with permission.
Putting brand names into dish titles is also an effective strategy for many chain restaurants, as are nostalgic labels like “handmade” or “ye olde” according to Brian Wansink from the Food and Brand Lab at Cornell University. A dose of patriotism and family can also boost sales.
I guess we can apply some partial pooling.  If this news article reports the work of several different research groups, and Wansink’s is one of them.  Then, given other things we’ve learned about Wansink’s work, we can make some inference about the distribution of studies of this type . . .
One can also consider this from the reporting standpoint.  100% of the quotations come from people with a direct incentive to promote this work.
Really sad to see this coming from the BBC.  They’re supposed to be a legitimate news organization, no?  I can’t really fault them for citing Wansink—back then, there were still lots of people who hadn’t heard about what was up with his lab—but even in 2017 weren’t they teaching journalists to interview some non-interested parties when preparing their stories?
P.S. The most extreme bit is this quote:
More than seven is too many, five is optimal and three is magical . . .
But that just gives away the game.  Now we’re talking about magic, huh?


  1. Matt Skaggs says:

    The abstract from the mouth movement paper begins:

    “We explored the impact of consonantal articulation direction of names for foods on expected palatability for these foods (total N = 256). Dishes (Experiments 1–2) and food items (Experiment 3) were labeled with names whose consonants either wandered from the front to the back of the mouth (inward, e.g., PASOKI) or from the back to the front of the mouth (outward; e.g., KASOPI). Because inward (outward) wandering consonant sequences trigger eating-like (expectoration-like) mouth movements, dishes and foods were rated higher in palatability when they bore an inward compared to an outward wandering name. This effect occurred already under silent reading and for hungry and satiated participants alike. As a boundary condition, this articulation effect did occur when also additional visual information on the product was given (Experiment 3), but vanished when this visual information was too vivid and rich in competing palatability cues (Experiment 2).”

    I dunno, I always order the kasopi because most restaurants put too much salt in their pasoki.

    • Andrew says:


      The good news is that researchers in psychology don’t do this sort of thing anymore. They’ve heard about the replication crisis and they’ve learned their lessons, and now they know not to tell stories based on random numbers. It was a bad time during the 2000-2020 period but now we’re over it. People will have to find something different to give Ted talks about.

      • Adede says:

        I thought the bad times ended in 2014. Now you say they ended in 2020? Is this a joking way of saying that crapppy papers continue to get published?

        • Andrew says:


          Your question made me curious so I went to the webpage of Psychological Science to see what’s in their latest issue. Here are the titles of the articles:

          The Dramatic Impact of Explicit Instruction on Learning to Read in a New Writing System
          Patterns of Genital Sexual Arousal in Transgender Men
          Challenging the Link Between Early Childhood Television Exposure and Later Attention Problems: A Multiverse
          How Collective-Action Failure Shapes Group Heterogeneity and Engagement in Conventional
          and Radical Action Over Time
          Effects of Time-Varying Parent Input on Children’s Language Outcomes Differ for Vocabulary and Syntax
          Unconscious Touch Perception After Disruption of the Primary Somatosensory Cortex
          Titrating the Smell of Fear: Initial Evidence for Dose-Invariant Behavioral, Physiological, and
          Neural Responses
          The Shape of Space: Evidence for Spontaneous but Flexible Use of Polar Coordinates TC in Visuospatial Representations
          Forgetting the Future: Emotion Improves Memory for Imagined Future Events in Healthy Individuals
          but Not Individuals With Anxiety
          Harder Than You Think: How Outside Assistance Leads to Overconfidence
          Asking People to Explain Complex Policies Does Not Increase Political Moderation: Three Preregistered
          Failures to Closely Replicate Fernbach, Rogers, Fox,
          and Sloman’s (2013) Findings
          Functional MRI Can Be Highly Reliable, but It Depends on What You Measure: A Commentary
          on Elliott et al. (2020)
          Need for Psychometric Theory in Neuroscience Research and Training: Reply to Kragel et al. (2021)
          Corrigendum: Using Machine Learning to Generate Novel Hypotheses: Increasing Optimism About COVID-19
          Makes People Less Willing to Justify Unethical Behaviors
          Corrigendum: Pain as Social Glue: Shared Pain Increases Cooperation

          There seem to have been some changes since the glory days a decade ago. Several of the articles are critical reanalyses or replications of earlier published work; that seems like a good sign already.

          I remain concerned about noise mining in some of the studies. For example, the article on unconscious touch perception draws strong conclusions from the finding that “12 participants . . .reliably performed at above-chance levels on a localization task,” and the article on forgetting the future reported several interactions in a study of 53 people. I’m not saying the particular claims made in these articles are wrong—I just don’t know—I’m just expressing some generic concern. In any case, there seems to be less of this sort of thing than before.

    • Lahvak says:

      I ordered pasoki the other day. I was really looking forward to it, but it was the biggest disappointment! The dish they brought to me had absolutely nothing in common with the great classic pasoki my Ukrainian grandmother used to make! I then tried to order okapi, but it turns out these are endangered and I am not rich enough.

  2. Joshua says:

    > Heavier menus have been shown to suggest to the customer that they are in a more upscale establishment where they might expect high levels of service.

    I went to a greasy spoon back before the pandemic. They had 10 pound menu.

    The hamburgers cost $50 and were terrible-tasting and the wait staff were insulting and not attentive. But I thought the hamburgers were a bargain and I left a 50% tip.

    Now I know why.

    • Joshua says:

      Take off on the Woody Alan joke:

      Two elderly women are at a Catskill mountain resort, and one of ’em says, “Boy, the food at this place is really terrible.” The other one says, “Yeah, I know; and such lightweight menus.”

  3. paul alper says:

    As a counter to obsession with layouts and menu manipulations, I suggest you read Calvin Trillin’s “Tummy Trilogy.” Trillin is famous for dropping into authentic NYC Chinatown restaurants and ostentatiously refuse the printed English menu, point to the Chinese characters chalked on the wall behind the counter and declare, “I would like one of each.” He does note, that once he took an owner by surprise because the Chinese characters on the wall referred not to food but to directions for the toilet.

  4. Dzhaughn says:

    Wait, is “magical” better than “optimal?”

    OMG we’ve been settling for optimal solutions this whole time. I suppose this is what Andrew doesn’t get about Elon Musk’s tunnels in Miami. And postmodernism.

    Years ago I read this recipe that said a stew was “best the next day, and even better the day after.” I’m still mad about it.

  5. Anonymous says:

    > Heavier menus have been shown to suggest to the customer that they are in a more upscale establishment where they might expect higher levels of service.

    My first thought is that I still have the menu from a visit to Chez Panisse (years ago) tucked into a cookery book.
    The menu was standard A4 run off on a photocopier. The level of service was – of its kind – the best I have ever experienced. The food was pretty good too.

    May be that Chez Panisse doesn’t count as ‘upscale’, of course. (I can also easily imagine that Alice Waters would not be particularly happy to have her place described in such Basil Fawlty-like terms).

  6. gec says:

    Here’s a hypothesis about why some of this goofy social priming results keep coming back up: They are used so often as examples in many newer versions of intro stats textbooks. ​Does changing the font color make you vote for the Green Party candidate? Do people with dogs kiss better than those who don’t? Do yawns get transferred only between people of the same socioeconomic stratum?

    I get it. Most undergrads hate having to take intro stats, and the authors see these cutesy “relatable” studies as the peanut butter they can use to get undergrads to swallow the statistics pill. But they undermine the whole project by giving students the false impression that causal identifiability and generalization come “for free” as long as you have some combination of randomization and appropriate tests. It’s the same false impression behind this style of research, and I worry that it will persist as long as textbooks keep perpetuating it, even if their use of cutesy examples is well-intentioned.

    Of course, another reason why some textbooks use this crummy style of research as examples is because they are written by people who do that kind of work and don’t see the problem (e.g., the widely-used Aron, Coups, & Aron “Statistics for Psychology” book still uses Wansink’s work as examples).

    • Andrew says:


      Can you share some images of this psychology textbook that uses pizzagate examples? Did they keep them in even after all the problems were revealed in that work?

    • jim says:

      “cutesy “relatable” studies as the peanut butter they can use to get undergrads to swallow…”

      …whatever is on the edutainment menu for the day.

      That’s pretty much modern post-secondary education in the US. As one newfangled educator/science professor noted in a story on our local NPR station, “we know” (that with our new teaching “approach”) “they learn more of what we want them to learn” – that is, they eat the peanut butter.

    • PsychedOut says:

      You might not find the topics exciting, but since when has simple randomized experiments fallen out of fashion?

  7. Kaiser says:

    Catching up on some Gelman today. This is a particularly good one. As I read through the post, I’m reminded of a name. I clicked on those links just to find out if it’s him. Disappointed. I can’t help but think these studies are probably not any better. Then came the last bit, and the suspense was over. Thanks Andrew for the laughs.
    Now, wake me up when they do an actual statistical experiment. They could partner with a restaurant chain, design an experiment involving all of the above factors, print different versions of the menu, and randomly assign them to real diners who make real orders. As a foodie, I’d love to be part of this research team!

Leave a Reply

Where can you find the best CBD products? CBD gummies made with vegan ingredients and CBD oils that are lab tested and 100% organic? Click here.