With friends, yes, and by a lot. A 2019 meta-analysis of 42 studies found strong evidence that people select and eat more when eating with friends than when eating alone, with a standardised mean difference of 0.76. With strangers or acquaintances, the same analysis found no evidence of the effect at all.
That split is the useful part. It is not company that raises intake. It is familiarity.
What the meta-analysis pooled
A 2019 systematic review and meta-analysis in the American Journal of Clinical Nutrition searched four databases for studies examining food intake and food choice as a function of the number of co-eaters, then handled the two study types differently: naturalistic studies were synthesised narratively, and experimental studies were pooled statistically.
Across the friend comparisons the result was clear, with Z equal to 5.32 and a p value below 0.001, giving a standardised mean difference of 0.76 and a 95 percent confidence interval from 0.48 to 1.03. In a field where effects of 0.2 are common, that is substantial.
The comparison that found nothing
The studies comparing eating alone with eating among strangers or acquaintances produced a standardised mean difference of 0.21, with a confidence interval running from -0.10 to 0.51 and a p value of 0.19. In plain terms, the data cannot distinguish that from no effect.
This matters because the folk version of social eating is that being watched changes what you do. If that were the mechanism, strangers should produce the strongest effect, since strangers are who you would be performing for. They produce the weakest. Whatever drives the extra eating is something friends supply that strangers do not.
The two mechanisms with evidence behind them
The review identifies candidates and is careful about how much weight they carry. In two studies there was evidence that the effect of social context on eating may be partly mediated by longer meal durations and by the perceived appropriateness of eating.
Both are plausible and neither is established. A meal with friends runs longer, and time at the table is one of the more reliable predictors of how much gets eaten. And in familiar company, continuing to eat reads as normal rather than as something to be justified, which removes a brake that operates quietly among strangers. Two studies is a thin base, and the authors say directly that further research is needed to establish mediating factors.
Their broader proposal is more interesting than a warning. They suggest social facilitation of eating evolved as an efficient evolutionary adaptation, which frames it as a feature of how humans eat rather than a lapse.
There is a methodological reason to trust the direction here more than a single striking study. The review deliberately kept the two kinds of evidence apart, synthesising naturalistic studies narratively and pooling only the experimental ones, rather than blending observational and experimental findings into one number. That is the same distinction that decides the sweetener argument, and it is the reason this meta-analysis can say people eat more with friends rather than only that the two tend to occur together.
What to do with this, which is less than it sounds
The instinct is to turn a measured effect into a rule, and this is a case where that goes badly. Eating alone to avoid a standardised mean difference of 0.76 is a genuinely poor trade against what shared meals do for people.
The moderator findings argue for restraint too. There was some evidence that gender, weight status and food type change the size of the effect, but the review flags that this evidence was limited by a lack of experimental research on those factors among friends. So the headline holds and the personalisation does not.
What is actually actionable is expectation setting. If Friday dinners with the same four people reliably run long, that is a predictable pattern with a known direction, not a weekly surprise to be disappointed by. Eating out in maintenance covers the mechanics, and what a planned bigger meal actually does deals with the guilt that usually follows.
The familiarity split has one genuinely practical implication, though it is a modest one. The occasions where this effect is largest are the recurring ones, the standing dinner and the regular group, not the one-off work lunch with people you barely know. Recurring occasions are also the ones you can see coming, which means the situation where the effect is strongest is the same situation where planning is easiest. That is an unusually convenient overlap and it is most of what this research is good for.
Why OffRamp will not flag your Friday
Two design rules in OffRamp exist to keep single days from turning into feedback, and a long dinner with friends is the clearest case they were built for.
The protein hit rate is counted over the days you actually logged, not over the calendar, so a day you did not log is absent from the denominator rather than counted as a zero. And the Regain Radar needs a 7 day rolling average to sit at least 2 percent above your baseline for seven consecutive days before it flags drift, with a 4 percent threshold for an alert. One meal cannot move that. Neither can one weekend.
The reason for those thresholds is that a system which reacts to a single evening teaches you to treat every social meal as a risk, which is both wrong on the evidence and corrosive over a year of maintenance. A 0.76 effect at a dinner table is real and it is also not the thing that decides how maintenance goes. A healthy relationship with food has to survive Friday, and an app that panics on Saturday morning is working against that.
None of this is medical advice. If eating around other people is a source of real distress rather than a logistical question, that is worth raising with a clinician.
