Why Thinking About Food Feels Like Wanting More
Most of us assume imagining chocolate makes us want it more, because a vivid image can boost salivation and hunger within seconds, which is why ads rely on close-ups and why diet advice often says avoid thinking about tempting foods.
That story says imagery amplifies desire. More imagery means more desire. Yet perception science offers a competing mechanism where repeated exposure reduces response even when the stimulus has not changed, a process called habituation that explains why the tenth bite of cake is less wanted than the first before nutrients reach blood.
Carey Morewedge at Carnegie Mellon wondered whether imagined bites could habituate too, because imagery shares neural machinery with perception for motor simulation and emotion, suggesting that repeating a mental eating action thirty times might mimic exposure and reduce motivation.
What They Actually Did
Five experiments enrolled 292 adults who each imagined 33 actions one at a time to hold effort constant.
Experiment 1 assigned 51 adults to imagine 33 quarters, 30 quarters plus 3 M&M's, or 3 quarters plus 30 M&M's, then weighed a 40 gram bowl covertly. Experiment 2 crossed repetition with content in 51 adults, testing whether 30 repetitions of anything reduces intake or only imagined eating does.
Experiment 3 gave 68 adults M&M's to either imagine eating or merely place into a bowl, 3 or 30 times, separating habituation from priming. Experiment 4 gave 41 adults matching or mismatching imagery, 30 cheese cubes or 30 M&M's before cheddar, testing general fullness against food-specific habituation.
Experiment 5 put 81 adults through a reinforcement game, clicking up to 128 times per cheese point, to separate wanting from liking, with liking rated 1 to 7 before and after.
Imagined Eating Cut Intake in Half
Experiment 1 delivered the headline. Imagining 30 M&M's led to 2.21 grams eaten versus 4.18 grams after 3 and 4.08 grams control, ANOVA F(2,46) equals 3.61 p less than 0.05 contrasts F(1,46) equals 5.81 and 4.50, roughly 47 percent less that looks small because one M&M weighs about 0.8 grams.
Experiment 2 replicated with interaction F(1,47) equals 4.65 p less than 0.05 where 30 imagined M&M's beat 3 F(1,47) equals 4.24 p less than 0.05 while 30 quarter insertions did nothing F equals 1.05 p greater than 0.30, so repetition alone fails.
Experiment 3 distinguished mechanisms because eating 30 versus 3 led to 4.28 versus 7.59 grams F(1,64) equals 6.10 p less than 0.05 while placing 30 versus 3 led to 7.00 versus 3.87 grams F(1,64) equals 5.11 p less than 0.05 interaction F(1,64) equals 11.17 p less than 0.05, consistent with habituation and opposite priming.
Experiment 4 showed specificity because 30 cheese cubes cut cheese to 6.36 grams versus 11.25 grams after 30 M&M's F(1,37) equals 5.14 p less than 0.05 while M&M imagery did not cut cheese, arguing against general fullness.
Experiment 5 showed wanting fell while liking held steady because 30 cubes reduced work for cheese mean log 3.68 versus 4.35 F(1,66) equals 5.01 p less than 0.05 with no liking change delta 0.00 versus 0.03 F less than 1 while response rate fell linearly as clicks doubled F(5,62) equals 315.08 p less than 0.001.
For scale, saving 2 grams per snack three times weekly is 6 grams weekly, 312 grams yearly, roughly 1500 calories: a modest 5 percent cut achieved without changing availability or cost.
The Strongest Counterargument
The sharpest critique is that these are small lab experiments where a tiny gram difference is inflated into a headline and the mechanism is confounded by boredom after imagining the same action 30 times consecutively in a quiet room.
Samples of 51, 51, 68, 41, and 81 were typical for 2010 but underpowered for interactions by current standards, and the outcome is covert weighing of a single bowl eaten immediately after imagery rather than real-world satiety, total calories, or weight change over weeks.
Demand characteristics loom because participants might infer the hypothesis after 30 repetitions, and mental fatigue could reduce approach independently of habituation, a confound constant effort mitigates but does not eliminate.
Replication is where the story gets interesting. The heaviest replication artillery ever built confirmed this finding. Camerer's 2018 project re-ran 21 Nature and Science experiments with pre-registered plans reviewed by the original authors and samples averaging five times the originals. Its redo of Experiment 1, 30 imagined M&M's versus the quarter control, found a significant effect in the same direction, and co-author Joachim Vosgerau cites it on his CV as an independent replication of the study 1 result.
But the largest single test found nothing. Andersen, Byrne and Wang (2023) ran about 300 adults at a Danish food festival through 30 trials of imagined eating, actual eating, or a token-imagery control, then measured desire, enjoyment, and wanted M&M's. Imagined eating matched control on every outcome. Actual eating satiated far more, rejecting both hypotheses. Plausibly, festival participants clicked through imagined bites at a median 3.4 seconds per trial in a noisy tent with no vividness check, while the original lab studies were self-paced and immersive. That fits Missbach's moderator neatly: if imagined habituation needs cognitive resources, distracted minds should not habituate, which is what Missbach found in the lab and what a festival tent produces.
What We Didn't Prove
We did not prove a dieting tool, because all outcomes were immediate grams eaten within minutes with no data on hunger later, total intake, or weight, and habituation shows spontaneous recovery after delay, nor generality across foods or people, since the original foods were M&M's and cheddar cubes, the participants were US adults capable of vivid imagery, and aphantasia, eating disorders, and aversions were never tested.
We did not measure brains, so the claim that mental representation alone engenders habituation rests on behavioral analogy rather than imaging within these experiments, leaning on Epstein and Kosslyn theory, with Science correcting Table 1 SDs October 23, 2018 without changing conclusions.
The Bottom Line
Imagining eating 30 times cut immediate intake of that same food by roughly half versus imagining 3 times or control, across four intake experiments plus a fifth on motivation. The effect was food-specific and ran through reduced wanting, not liking, which challenges the intuition that thinking about food always feeds desire. But it worked only right away, and only for the imagined food.
What You Can Do
Try it before a snack you already plan to eat by taking 60 to 90 seconds to imagine 30 bites of that exact food, including picking it up, smelling it, chewing, and swallowing while counting silently and making each bite distinct.
Match imagery to the food you will actually eat because imagining M&M's did not reduce cheese intake and imagining moving food into a bowl increased intake, so specificity matters and you should imagine the food you expect to overeat rather than a generic alternative.
Do not use this as a license to skip meals or suppress broadly, since experiments show reduced motivation immediately afterward not increased self-control generally and liking stays intact, so use the window of lower desire to portion a smaller amount and then stop.