A tightly controlled experiment found that people ate more and gained weight on one ultra-processed menu than on an unprocessed menu. That finding matters: it came from an intervention, rather than a survey asking people to remember their meals. Turning it into a rule for every packaged food, however, skips the most useful details.
This guide separates what the experiment measured from the decisions it can help inform. The practical question is how to build meals you can afford, prepare and enjoy while considering nutritional quality. A processing label contributes information; it cannot make that whole decision for you.
The experiment: the same people tried both menus
In the 2019 Hall trial, 20 adults lived at the NIH Clinical Center. Random assignment determined which diet they received first. Each condition lasted two weeks, followed immediately by the other. Participants could eat as much or as little as they wanted.
| Question | Finding | Boundary |
|---|---|---|
| How much did intake differ? | 508 kcal/day higher on the ultra-processed diet; reported standard error 106 kcal/day | A mean difference between these dietary conditions, not an amount to subtract from your meals |
| What happened to weight? | Mean gain of 0.9 kg during the ultra-processed condition and loss of 0.9 kg during the unprocessed condition | Each condition lasted two weeks; these are separate changes, not annual forecasts |
| What was matched? | Menus were designed to match presented calories, energy density, macronutrients, sugar, sodium and fiber | Matching food offered does not mean participants consumed equal amounts |
Figures come from the original report’s abstract. A standard error describes uncertainty around a mean; it does not describe the range of every participant’s response. Weight change also should not be renamed fat change without the relevant body-composition analysis.
Why the design strengthens one conclusion
Randomizing the order helps reduce the chance that people with different starting habits simply selected different diets. Having each person try both conditions also makes the comparison different from a study comparing unrelated groups. The observed difference supports a causal conclusion about the tested dietary conditions in this setting.
The distinction between food offered and food eaten is central. Imagine two tables with similar total nutrients available. If people choose different amounts from them, their actual intakes can differ. A menu description alone therefore cannot establish what was consumed. This trial measured the latter question rather than assuming that matched menus would lead to matched eating.
That makes the experiment relevant to a familiar claim: that only deliberate calorie counting matters and the food environment is irrelevant. It shows that the dietary condition can affect intake when people are free to choose the amount. It does not tell us which single feature produced the difference.
What a processing label cannot settle
The intervention changed whole menus. It was not a separate randomized test of each additive, package, cooking technique or ingredient. If someone names one ingredient as the explanation, ask which experiment isolated that ingredient. The menu comparison cannot supply that missing test.
Nor does it rank every supermarket item. A trial menu is a defined collection of foods, while a grocery aisle contains many products, recipes and uses. A food eaten alone may play a different role from the same food used as one component of a meal. Price, storage, taste and preparation time also change which alternatives are realistic.
The small sample, short periods and inpatient setting limit generalization. Living in a research center is different from shopping, working and cooking at home. The trial cannot establish the long-term effect of replacing one usual product, or tell a household what its food bill would become.
Use nutritional quality alongside convenience
WHO’s healthy-diet guidance emphasizes adequacy, balance, moderation and variety. It also recognizes frozen and canned produce as options when added sugars or excess sodium are avoided. Processing and nutritional usefulness should therefore be examined together rather than treated as interchangeable judgments.
Start with the meal you actually need. A lunch that requires equipment you cannot access is a poor practical substitute, even if its description looks attractive. A workable comparison asks what can be added or changed while keeping the meal available. Our protein food examples can help identify components without setting a personal intake target.
A grocery comparison you can use
The following is an editorial planning exercise, not a tested intervention or a scoring system. Choose two realistic versions of one frequent meal. Leave the decision open until you have compared both.
| Write down | Meal A | Meal B |
|---|---|---|
| Main components | List the foods and drinks you normally include | List the proposed replacement and additions |
| Nutrition information | Check relevant labels and what the whole meal supplies | Use the same comparison basis, including portion size |
| Practical cost | Note price, preparation, equipment and storage | Include extra shopping or cooking required |
| Enjoyment and fit | What makes this meal useful? | What would make you willing to repeat it? |
For an illustrative comparison, someone might consider their usual packaged lunch beside a meal assembled from canned beans, vegetables and bread. Those descriptions do not establish which has less sodium, enough food for that person, or a lower cost. Check the actual products and the whole meal. The example is a way to gather information, not a verdict about either lunch.
If an alternative is unaffordable or repeatedly goes uneaten, revise the proposal. You might change one component rather than replace everything. Medical dietary restrictions require more specific advice than this worksheet provides; bringing actual labels to a dietitian can make that discussion concrete.
Read the next headline with the menu attached
Before sharing a claim, ask what was compared, whether intake was measured, how long follow-up lasted and whether the conclusion concerns whole diets or individual foods. These four questions preserve the strength of the experiment without expanding it into an explanation for every person’s weight.
A useful takeaway can remain specific: these tested menus produced different intake and weight changes over two weeks. Your grocery decision needs another layer of information about the actual meal. Keeping both levels visible is more helpful than treating a processing category as a complete meal plan.



