My grandfather, Irvine Robbins, co-founded Baskin-Robbins. Yes, that Baskin-Robbins — 31 flavors of ice cream, iconic pink spoons: an empire that helped make frozen dessert a fixture of American childhood.
My dad, John Robbins, grew up in a home with an ice cream cone-shaped swimming pool, gave input on the invention of new flavors, and ate more than his share of the family product. His personal favorite was Jamoca Almond Fudge.
But my dad chose a very different path (what he jokingly called his own “rocky road”), devoting much of his life to reforming the food system into something healthier, more ethical, and more sustainable. His first book, the 1987 best seller Diet for a New America, makes the case that what Americans eat (including, as fate would have it, dairy and sugar) makes them sick and shortens their lives.
Eventually I joined him in that work, and for decades now, our family has been on something of an anti-ice-cream crusade — hoping, perhaps, to clean up some chilly karma.
So you can imagine how I felt when I came across a stack of Harvard studies suggesting that ice cream might actually be good for you.
Had we been wrong all this time? Was Grandpa Irv a public health visionary and nobody told us?
I looked into it. The answer is no; Grandpa Irv was not vindicated, at least not fully. But what I found as I dug deep into the research has made me a much better consumer and interpreter of nutritional science.
And it leads somewhere surprising. So let’s dive in.
The Stack of Studies
I first encountered this vein of research when I stumbled across a 2018 doctoral dissertation on dairy products and cardiometabolic health. It was written by Andres Ardisson Korat, a student at the Harvard T.H. Chan School of Public Health. His advisor was Frank Hu, PhD, one of the most cited nutrition researchers in the world.
One chapter looked at nearly 16,000 people who already had type 2 diabetes, followed for years in 2 of the longest-running health studies ever conducted: the Nurses’ Health Study and the Health Professionals Follow-Up Study. The question was whether any dairy food affected their risk of heart disease, and for almost everything, the answer was no. Total dairy, high-fat dairy, low-fat dairy, milk, cheese, yogurt, butter; none of them materially moved the needle one way or the other.
Except ice cream. Among people living with diabetes, those who ate ice cream at least twice a week had about 12% lower risk of cardiovascular disease than those who almost never touched it.
That would have been a curiosity on its own. What made it truly puzzling is that it wasn’t the first time. Four years earlier, Hu co-authored a study tracking the dairy consumption of nearly 200,000 American adults, more than 15,000 of whom developed type 2 diabetes over the follow-up period.
Whole milk and cheese were associated with slightly higher risk, while two dairy products were associated with lower diabetes risk. Yogurt was one, at about 17% lower risk per daily serving. Ice cream was the other, at about 22% lower risk.
Between the dissertation and the 2014 study, a meta-analysis of 22 cohorts covering more than half a million people found more or less the same thing: An average of roughly 10 grams of ice cream a day — about a tablespoon (a truly small amount to any serious fan of the frozen dessert) — was associated with 19% lower risk of type 2 diabetes.
The Scientific Outcry and the Hidden Oops
The Atlantic ran a widely quoted article by David Merritt Johns, with the subheadline: “Studies show a mysterious health benefit to ice cream. Scientists don’t want to talk about it.” The narrative that spread across dozens of aggregator sites was that Harvard researchers ran every possible analysis to make the result go away, without success.
That’s a fun story, but it’s not quite right.
Buried in the supplementary tables is a series of statistical calculations in which the researchers pushed back the starting line to see whether the association between ice cream and reduced risk of heart disease still held. With a four-year delay, it did; ice cream still looked protective. With an eight-year delay, it didn’t. In Ardisson Korat’s own words, the association “was attenuated,” and the numbers no longer cleared the bar for statistical significance.
A second test asked whether people who ate more ice cream over time fared better than people who ate less, and found no association at all.
Neither result made the news. The puzzling ice cream-prevents-diabetes finding had already become the story.
The coverage missed one more thing. Other chapters of Ardisson Korat’s dissertation went on to appear in peer-reviewed journals. This one, as far as I can tell, did not. The ice cream analysis that got all the attention exists in Harvard’s institutional repository and nowhere else.
I don’t know why. Dissertation chapters go unpublished for plenty of ordinary reasons, and I’d be making things up if I told you which one applied here. But it’s worth knowing that the most talked-about ice cream finding in nutrition science never went through peer review.
None of that was a scandal. No one cheated, buried data, or lied. It was an ordinary week in nutrition science, which is precisely why it’s concerning. If this can happen to a finding this scrutinized, in cohorts this large, at an institution this careful, it can happen to any story about food and health.
So one useful thing to take away from all this is a set of questions that could be wise to ask before believing any headline about food.
How Not to be Misled by Nutrition Headlines
Let’s swing our attention from the actual research — the stuff buried in doctoral theses and nutritional journals — to the headlines that the public saw.
Here’s a representative sample of the top results of a Google search for “ice cream prevents diabetes”:
- Ice Cream is a Superfood
- How ice cream can impact your health (hint: it might be good for you)
- Eat your ice cream for a long, healthy life? This doctor says so
I don’t know about you, but when I see a credible source giving me good news about my bad habits, I have an urge to celebrate by indulging that habit. So I’m guessing that more than a few people who saw these stories immediately put ice cream on their “nice, not naughty” list.
I read doctoral dissertations (or, in this case, “dessertations”?) on nutrition for a living, so I can cut through the hype to the actual studies, data, statistical adjustments, and asterisks.
The ice cream-prevents-diabetes story is a good one to study, because it highlights four research artifacts: ways data can be misinterpreted, either innocently or to serve someone’s agenda. In this case, three of them made ice cream look healthier than it really is, while one of them tilted the interpretation in the opposite direction. Let’s look at them.
Artifact #1: Reverse Causation
Did you know that umbrellas cause rain? I mean, just look at the statistics: the more umbrellas per capita in a country, the higher the annual rainfall.
If you’re shaking your head at my foolishness, you understand the trap of reverse causation. The rain causes people to buy umbrellas, not the other way around.
So how does that apply to the correlation of ice cream consumption and better health? People who show early signs of metabolic trouble often get told by doctors to clean up their diet. One step they may take is to stop eating ice cream. Meanwhile, those who haven’t received that warning are still happily enjoying sweet scoops in a cup or cone.
So when researchers ask people about their diets, the ice cream eaters might look healthier. Not because ice cream is protecting them, but because being told they’re at metabolic risk leads some people to give up ice cream.
Reverse causation has a close cousin worth knowing about. Suppose the people who eat ice cream a couple of times a week differ from the people who never touch it in some way the researchers didn’t measure. Maybe they’re wealthier. Maybe they’re less anxious about food. Maybe they’re the kind of people who eat dessert at a table with other people rather than alone. Any of those could affect diabetes risk on its own, and none of them are ice cream.
Researchers adjust for everything they can think of, and this study adjusted for a lot: age, weight, smoking, exercise, alcohol, family history, and a long list of other foods. But you can only adjust for what you measured, and what’s left over is called residual confounding. It’s the reason observational nutrition findings so often shrink or vanish when someone runs an actual trial.
The Harvard team saw the reverse causation problem coming. In their three cohorts, ice cream consumption dropped after people were diagnosed with high blood pressure or high cholesterol. So they reran the analysis to account for those diet changes, and the ice cream effect shrank by half, from a 22% lower risk of type 2 diabetes to 11%. But it was still there. Cut in half, but still statistically significant. The researchers described it as attenuated, “although still significant.” Which means reverse causation might explain some of what was going on, but not all of it. Something else is in the data.
It’s worth noting that the same adjustment wiped out negative findings about whole milk and cheese. After the reanalysis, both went null.
Artifact #2: Underreporting Bias
istock.com/Daniele Mezzadri
Most studies of diet and health suffer from a really big flaw: They don’t measure what people actually eat. Instead, they rely on self-reporting, often through food questionnaires. And people are notoriously unreliable narrators when it comes to both remembering and honestly sharing this information.
There’s a systematic bias at play here: People are much more likely to underreport consumption of foods they think they shouldn’t be eating, and the effect is strongest for the indulgent, “sinful” foods that carry the most stigma. Ice cream sits pretty close to the top of that list.
This matters enormously for the ice cream research. If people in larger bodies or those with metabolic diagnoses underreport their ice cream more than lean, healthy people do, then the sickest eaters will appear to eat the least ice cream — not because they do, but because they say they do. The dessert gets statistically sorted toward the healthy. Ice cream ends up looking protective when all that’s really happened is that the people most likely to be harmed by it have left some of their consumption out of the questionnaires.
Mark Pereira, leader of the CARDIA study (Coronary Artery Risk Development in Young Adults, for those who love their acronyms spelled out), put it bluntly: “I don’t believe that the heavier people consume less desserts. I believe they underreport more.”
While underreporting bias is a big problem for ice cream and other “indulgent” foods, the issue affects almost all nutrition research. From broccoli and quinoa to soda and alcohol, pretty much every food and beverage choice has a moral component, which points to something uncomfortable about our society that’s arguably more alarming than nutritional uncertainty.
The reason ice cream gets underreported and broccoli doesn’t is that our society moralizes food, categorizing it as virtuous or shameful or somewhere in between. And we’ve taught people, especially people in larger bodies, to feel watched when they eat. That shame doesn’t just distort the data; it may be its own health exposure, and the fear we’ve wrapped around foods like ice cream may do as much or more damage than the food itself (more on this below).
Artifact #3: The Displacement Effect
Every nutrition finding has a hidden question attached: Compared to what? When a study says people who ate ice cream had less diabetes, the comparison is not ice cream versus nothing. It’s ice cream versus whatever those people would have eaten instead. And in the modern industrialized diet, sadly, plenty of options are worse than ice cream for blood sugar control.
That’s because ice cream, with its high fat content slowing glucose absorption, has a lower glycemic index than many common ultra-processed foods, including white bread, rice cakes, crackers, cookies, and sugar-sweetened breakfast cereals.
Someone who eats a bowl of ice cream instead of a large serving of refined carbs may, in that specific moment, be making a marginally better metabolic choice. Not because ice cream is good for you, but because some of the foods it displaces deliver an even sharper blood glucose spike.
That’s the displacement effect, in which a food can be described as beneficial not because it’s intrinsically good for you, but because it replaced something that causes a bigger glycemic response. And notice how narrow that comparison is. Glycemic index measures one thing: how fast blood sugar rises after a single food, eaten by itself, over a couple of hours. It says nothing about what that food does to insulin sensitivity over months and years. On that longer timescale, ice cream’s fat content stops looking like an advantage. A substantial body of research links saturated fat to insulin resistance through mechanisms in which fat accumulates inside muscle and liver cells and interferes with insulin signaling. Add a heavy sugar load, and you have two separate routes to the same destination.
So a food can win on the measure being used and still lose on the outcome that matters. That’s the deeper lesson of the displacement effect. Better than the alternative is always relative to some metric, and the metric may not be measuring the thing you care about.
This artifact has implications that go far beyond ice cream. First, in a nutritional environment dominated by ultra-processed carbohydrates, the displacement effect can produce false positives for many foods that don’t do your body any favors but are simply less damaging than even more harmful options.
Second, it’s a way for industry-funded researchers to produce research showing that their products are not unhealthy. Displacement could be an honest accident, as may have happened here. But food companies can engineer that accident on purpose, running the study so the funded food beats something worse.
Or they could stack the deck in the other direction, studying people so unhealthy that nothing moves the needle: “Eating rusty nails slathered in hot sauce was not associated with a significant increase in ulcerative colitis.”
That last example is ridiculous, granted, but the underlying tactic isn’t hypothetical. In 2007, a team at Children’s Hospital Boston gathered every study they could find on the health effects of soft drinks, juice, and milk over 5 years. One group of researchers classified each article’s conclusion. A separate group, working with no knowledge of those conclusions, classified who had paid for it. Among the interventional studies, not a single article funded entirely by industry reached an unfavorable conclusion about the product it examined. Among the independently funded studies, 37% did.
A later analysis of reviews on artificially sweetened drinks turned up the same pattern, along with something more unsettling. The industry-funded reviews weren’t sloppier than the others. Measured against standard criteria for how a review should be conducted, the two groups looked about the same. The bias didn’t show up as bad methodology. But it did show up in the conclusions.
The same analysis found it running the other way too: Every review funded by a competitor of the sweetener industry concluded that artificial sweeteners were bad news.
It seems that funding shapes conclusions, regardless of whose funding it is.
A similar pattern emerged when scientists analyzed red meat research: Clinical trials funded by the meat industry were nearly four times more likely to show favorable or neutral cardiovascular outcomes than those conducted independently.
For what it’s worth, the Harvard dairy study at the center of this article was funded by the National Institutes of Health, and the paper states the funders had no role in the design, the analysis, or the decision to publish. That’s the disclosure I’d want to see, and it’s why I’ve been willing to take these findings seriously enough to spend 5,000 words on them.
Artifact #4: Confirmation Bias
The first three artifacts all make ice cream look healthier than it really is. This one pushes the other way.
Confirmation bias is the tendency to notice, pursue, and promote evidence that fits what we already believe, and to work hard to explain away the evidence that doesn’t. Scientists are trained to resist it. They are also human, and confirmation bias is one of the most powerful mental habits ever studied.
Here’s how confirmation bias played out in the ice cream and diabetes research. That 2014 Harvard paper followed nearly 200,000 people and examined each dairy product separately. In the raw numbers, ice cream came out ahead: about 22% lower risk of type 2 diabetes, against yogurt’s 17%. But raw numbers are the weakest version of this argument, so let’s use the researchers’ own strictest analysis instead, the one that accounts for people changing their diets after a diagnosis. Run that way, most of the dairy foods land on nothing, including skim milk, cheese, and whole milk. Two survived. Yogurt, at 14% lower risk, and ice cream, at 11%.
In other words, unlike skim milk and whole milk, ice cream appeared, along with yogurt, to have anti-diabetes effects.
But the result that got the press release, the news coverage, and a decade of follow-up research was that yogurt lowers your risk of type 2 diabetes. The ice cream finding was footnoted, if mentioned at all, and set aside.
The paper’s stated conclusion was that higher yogurt intake was associated with reduced diabetes risk, while other dairy foods were not appreciably associated with it. Ice cream is a dairy food. It had just cleared the same bar yogurt did, in the same table, under the same adjustment. Years later, Dariush Mozaffarian, MD, one of the paper’s co-authors, acknowledged that the conclusions weren’t accurately written, telling The Atlantic that “ice cream was associated.”
Sit with that for a second. A co-author of one of the most cited nutrition papers of its year, saying the summary line at the top of his own study didn’t match what the study found. Not because anyone lied. Because the result didn’t fit the story, so it didn’t make it into the sentence that nearly everyone reads instead of the paper.
Why did yogurt get the spotlight? Probably because yogurt already had a reputation as a health food and a plausible biological mechanism of probiotics supporting the gut microbiome, while ice cream was the thing you eat in the dark by the tub after a breakup. When a result fits the story we already have, we call it a finding. When it doesn’t, we call it an artifact and go looking for the flaw.
I’m not standing above the fray from some position of ethical superiority. When I came across a stack of studies suggesting ice cream might be good for you, my first reaction was to go looking for the flaws (which I did find and have just shared with you).
But I have to sit with the possibility that I found them so easily because I went in looking for them: my own case of confirmation bias.
So What Actually Survived?
Two different findings have been in play this whole time, and they deserve separate verdicts. The cardiovascular claim, the one from the dissertation, is dead. It faded when researchers pushed back the starting line; a second test found no beneficial association, and it never made it through peer review. The diabetes association is a different story. It was published, and it survived the researchers’ own strictest adjustment. It came out at half strength, but it was statistically significant, and it was one of only two dairy foods that were.
Sit with what that means. Ice cream is essentially sugar and saturated fat, chilled. Nearly everything I understand about nutrition predicts it should have led to negative outcomes. It’s a food my family has spent decades arguing against. And in cohorts of hundreds of thousands of people followed for decades, it came out associated with lower risk.
Of course, if people underreport the ice cream they eat, and if those most likely to be harmed underreport the most, then it’s possible that the data might not be as cheery as it appears. But even if we set aside the idea that ice cream actively protects against diabetes and chalk that up to survey quirks, we’re still left with a huge puzzle. Where is the damage? By every rule in the modern nutrition textbook, a food combining heavy saturated fat and refined sugar ought to trigger a clear red flag in metabolic data. Yet across massive cohorts over decades, it completely failed to show the harm we expect. Even a null where you confidently expected a negative would still be worth sitting with. And it’s what sent me looking somewhere the questionnaires never went.
Food, Pleasure, and Connection: The Missing Nutrients?
A food frequency questionnaire asks how often you ate ice cream over the past year. It does not ask whether you ate it standing in front of the freezer at 11 pm because you had a miserable day, or whether you were sitting on a porch on a balmy July afternoon surrounded by loved ones. It doesn’t ask whether you felt disgust or whether you experienced delight.
In other words, the studies never looked at at least two variables: pleasure and connection.
So here’s the question I want to explore in the rest of this article: What if some of what we’ve spent a century attributing to the chemical composition of food actually derives from the circumstances in which we eat it?
My grandfather understood something about the emotional power of ice cream that no food frequency questionnaire has ever captured. Think about when people love to eat ice cream. Birthday parties. Summer evenings. Little league victories. First dates. A pink spoon extended across a counter to a child quivering with excitement.
We don’t yet know whether joyful eating improves the nutritional effects of food. I hope that research is coming. But before we shrug and give up, let’s go in the opposite direction. It turns out there’s a strong vein of research showing that stress and disconnection can change how you metabolize food in ways that make it worse for you.
Same Meal, Different Lives
In 2015, a research team at Ohio State led by Janice Kiecolt-Glaser did something novel. They brought women into a hospital research unit, fed them a specific meal, and watched what their bodies did.
The study brought in 58 women for 2 separate visits each. On one visit they ate a meal high in saturated fat. On the other, a meal high in sunflower oil. Neither the women nor the researchers serving the plates knew which was which. And before each meal, the team asked what stressful things happened to them yesterday.
The common answers: An argument with a family member. A deadline. A kid home sick. Ordinary life occurrences.
Then they measured what happened over the following hours — how much energy the women burned, how much fat they oxidized, what their insulin did.
The women who’d had more stressors the day before burned fewer calories after eating. They oxidized less fat, and their insulin ran higher.
The researchers put a number on the gap. The 6-hour difference between having 1 stressful thing happen the day before and having none came to about 435 kilojoules, roughly 104 calories. If that pattern repeated, they estimated, it would add up to almost 11 pounds in a year. Lower energy expenditure, less fat burned, and more insulin are key markers on the road to type 2 diabetes.
The food didn’t change. What changed was what the women had lived through in the previous 24 hours. And their bodies handled identical food differently because of those experiences.
Ice Cream and Social Connection
Ice cream is a much more social food than yogurt or whole milk. And there’s a body of research suggesting that the feeling of belonging can support healthy metabolic regulation. We know the inverse is true: Chronic loneliness and social isolation have measurable negative effects on metabolic health and inflammation.
The parasympathetic nervous system (also known as the rest-and-digest system) activates in response to safety, pleasure, and connection. And there’s emerging evidence that psychosocial context — who you’re with and how you feel about them — affects how food is metabolized.
Ice Cream and Pleasure
Beyond the social connections that often accompany ice cream consumption, there’s just the pure pleasure of eating a cold, sweet dessert that’s associated with celebrations and treats. Which raises the question: Could the positive health effects that appear to be associated with ice cream be, at least a bit, a signal about the healing power of delight?
Research on what some nutritional psychiatrists have called “vitamin P” (pleasure) suggests that the act of genuinely enjoying food, independent of who you’re eating with, produces measurable physiological benefits.
When we eat with full presence and real enjoyment, we activate the same parasympathetic state that supports healthy digestion and metabolic regulation. The key distinction is between eating for pleasure — present, connected, savoring — and emotional eating, which often involves dissociation from the food itself.
As Sarah Gold Anzlovar, MS, RDN, puts it: “When eating for pleasure, there’s usually a true connection and enjoyment that you are getting from the food.”
Emotional eating tends to bypass that connection entirely. One is nourishing in ways that go beyond calories. The other is using food as a substitute for something else.
A bowl of ice cream eaten slowly, with full attention and genuine enjoyment, in a state of ease, may be doing something quite different physiologically than the same bowl eaten mindlessly in front of a screen after a stressful day. The food may be identical. The experience, and possibly the metabolic outcome, is not.
Here’s a rhyme that may not get me an appointment as the U.S. Poet Laureate, but sums up my own strong hunch about the impact of mood on food:
Better health in the pie that you share with good cheer
than the greens that you swallow in worry and fear.
Now, it’s a rhyming couplet, not a PhD thesis. And it hasn’t been formally tested. But it is consistent with what we know about the gut-brain axis and the relationship between emotional state and metabolic health.
So What Does This Mean for How You Eat?
So where does this leave you, the next time you’re standing in front of the freezer case thinking about buying a cone or tub of frozen dessert? Instead of delivering a final verdict on ice cream — Grandpa Irv vs. Dad John, in Family Feud: Robbins Edition — I’d like to share three takeaways:
Takeaway 1: Five questions to ask before you believe a nutrition headline
- Is this observational or experimental? Almost everything you read about food is observational. Observational studies can show that two things travel together, but they can’t show that one caused the other.
- What’s it being compared to? Every finding has a hidden baseline. Better than what? is often the moral of the story.
- What do the people who eat this food have in common besides the food? And the people who avoid it, what do they have in common?
- Who paid for it? Industry-funded research produces sponsor-favorable results at a measurably higher rate.
- Has it been replicated, and did it survive the follow-up tests? And if it survived, did it survive at full strength, or at half? A single striking result is a hypothesis, not a conclusion. The reason “more research is needed” is one of the most common phrases in science is that, much of the time, more research is needed to say anything with confidence.
None of these questions require you to obtain a statistics degree or a doctorate supervised by Frank Hu. They require some curiosity and a willingness to be a wee bit skeptical about nutrition headlines.
Takeaway 2: Food choices aren’t moral tests
Many of us live in a culture that sorts food into virtuous and shameful, and teaches people — especially people in larger bodies — to feel watched while they eat.
That sorting has a cost. It corrupts the research, as we’ve seen. But it may also do direct damage. The evidence that stress and isolation impair how we metabolize food is a good deal stronger than the evidence that any single dessert helps or hurts. If shame is a chronic stressor, and we’ve deliberately attached it to eating, we may have built a health problem out of this moralizing.
So if you choose to, eat the occasional treat. And for the best results, savor the experience with people you like, without beating yourself up inside your head.
Takeaway 3: Whole food, plant-based eating is still the gold standard
While pleasure and social connection are important parts of life, I’m not saying that you can be healthy eating only Pralines ‘n Cream as long as you’re having fun.
Joy doesn’t cancel out chemistry. Nothing I’ve learned suggests that a diet of ultra-processed food outperforms a well-constructed one. The key, I think, is to bring more cheer to healthy foods. That gives you the best of both worlds. And more joy along the way!
The evidence on dietary patterns hasn’t changed: The foods most consistently associated with longevity and lower disease risk across the most rigorous research available are whole, minimally processed plants. That was true before the ice cream studies, and it’s true after them.
What this inquiry adds is a question about the conditions of eating rather than the contents. Eat well. Eat with presence. Eat with gratitude. Eat in good company.
And if it makes you happy, every so often include a scoop of something cold and sweet.
Here’s our article on whole food, plant-based frozen desserts.
Here’s more on mindful eating and social connection.
And if you’re looking for comfort foods that deliver genuine pleasure without the ethical and health costs of conventional dairy, we’ve got you covered. Check out our article on whole food, plant-based comfort foods here.
How did the Robbins Family Feud unfold? My grandfather eventually read my father’s book — which his own doctor endorsed — and in time he gave up sugar and ice cream, and partially reversed his type 2 diabetes.
For decades, my dad and I told that story solely as a cautionary tale. Now I think it might be more complicated than that, and infinitely more interesting.
Featured image: istock.com/GMVozd
Tell us in the comments:
- Have you ever seen a nutrition headline that surprised or confused you? How did you evaluate it?
- What role do joy and comfort play in your relationship with food?
- Do you have a food that you’ve always thought of as an indulgence that might actually be part of a healthy relationship with eating?
