Always worth attempting a randomised trial but sometimes you need a bit of creativity. So I admire this trial, which tests the effect of menu labelling. The basic idea is simple - randomise diners to receive menus with or without calorie labelling. But there were various nice touches, like giving the volunteers their meal free of charge in a university classroom rather than a restaurant, and following them up next day to see if people who picked low calorie meals made up for later with a late night snack.
The abstract of the paper is rather hard to understand. The basic finding was that calorie labelling did indeed lead people to choose and eat less (about 250 calories): but they make up this deficit with a late night snack unless you also put on the menu information on daily calorie requirements (2000 calories for an average adult).
Here are a couple of other papers to add perspective: first an enjoyable but speculative estimate of what would happen in Los Angeles if all fast-food restaurants labelled their menus, and second a dose of reality from the Netherlands.