5: Milkshakes and dietary restraint
In the first part of this series, we explored the general idea of linking the “more than you need” structure of an all-inclusive vacation with the “more than you’ve always convinced yourself you need” structure of successful recovery from a restrictive eating disorder. I also described the anorexic version of such a vacation (think nocturnal routines and lots of food hoarding), the pseudo-recovered version (full of “healthy eating/exercise” rules and plenty of body comparison) and the fully recovered blissed-out version that was my reality a few months ago.
You can do the thought experiment yourself, if you like: If you were to book a couple of Caribbean weeks for yourself tomorrow, what would your reality be like? What does that tell you about what you could be prioritizing right now?
Meanwhile, let’s take the next step in the self-regulation part of the argument. All-inclusive resorts give us one angle on what it means to make your own decisions: to self-regulate in ways not constrained by blindly applied rules. Some interesting experiments carried out in the 1970s gave us another. They investigated what “dietary restraint” (DR: using self-control to try to limit one’s food intake) does to people’s eating habits when they’re presented with something they like eating (ice cream) after they’ve already consumed a milkshake they wouldn’t normally.
The way DR survives—and is proliferating—as an approach to eating is that it promises you’ll eat less, or make better food choices, and so end up with a slimmer (read “better”) body. Anorexia nervosa (AN) is DR on steroids. AN gets you thin. And it keeps you not eating very much—until it doesn’t.
Many people diagnosed with AN progress to a bulimia nervosa at some point (and/or progressed to AN from something else), and transitions between diagnostic categories are common in many directions and at many phases of illness and life (Schaumberg et al., 2019).
Just as missing a meal often leads to later eating more than what you missed, so chronic restriction often leads to chronic over-eating—specifically in a way that feels out of control and may culminate in deliberate vomiting or other negation/compensation attempts. Of course, this leaves the others: the people who don’t shift from anorexia to any other eating disorder. They either have lifelong anorexia (which is not a victory for the human host), or they recover fully and permanently—a victory for the human and not for DR (or AN).
In the non-clinical context of people who exert higher or lower levels of dietary restraint, the same kind of picture emerges: restraint can not just be fragile, but create fragility. In Herman and Mack’s (1975) experiment, participants were asked to consume either no milkshakes, one chocolate milkshake, or one chocolate and one vanilla milkshake in a fake taste test acting as a pre-load.
Then everyone was presented with three tubs of ice cream (chocolate, vanilla, and strawberry) with another taste survey and invited to “taste” as much of each as they wanted in ten minutes, supposedly to provide accurate taste ratings. The researchers’ predictions were as follows:
subjects required to consume two milkshakes in addition to their daily quota of calories would be in a position of having exceeded the “permissible” limits of consumption for a restraint-governed daily intake. Normally restrained subjects might be expected temporarily to give up the attempt at restraint, once they had come to perceive themselves as having already “overeaten”.
If such subjects had not consumed a milkshake, their normal restraint would remain intact. Highly restrained subjects, then, were expected to consume more ice cream in the two-milkshake condition than in the zero-milkshake condition.
By contrast, subjects who are normally not restrained would not be “triggered” by the excessive milkshakes. Such subjects should behave internally, eating less ice cream after a larger milkshake preload. For both types of subjects, the one-milkshake preload was expected to have an intermediate effect. (p. 650)
The data were a strikingly good fit for the predictions. The researchers found that the low-restraint eaters would simply eat until they were full, whereas the high-restraint eaters who had already had either one or two milkshakes ate more ice cream than any of the others:
Quite clearly, the data conform strongly to the predicted interaction. High restraint subjects consume more ice cream after the milkshake preload than after no preload at all. Low restraint subjects consume decreasing amounts of ice cream as a function of the size of the preload. (p. 654)
One milkshake seemed to be enough to eliminate restraint in the high-DR participants, presumably through the “what the hell” disinhibition effect that leads to counter-regulation. The “dieters” who drank one or two milkshakes kept eating because they’d already failed by breaking the rule, so they might as well fail big. (This “what the hell” effect is the diet-specific version of the all-or-nothing fallacy so pervasive in eating-disordered forms of thought.)
All results controlled for “acute deprivation” by assessing time since the last eating and rough calories eaten then. And it’s worth noting that there was no difference between the “normal weight” and “obese” (>15 percent overweight) participants’ behaviour: The difference that makes a difference is the one between low “restraint” and high, or being a non-dieter or a dieter. So what we have is oscillation between extremes: from long (or not so long) periods of eating less than one would like to shorter periods of eating as much as one would like—which is a lot because the appetite is typically denied. The oscillation, by definition, precludes getting practice at existing in the middle ground of sensing and responding and adjusting without hard and fast top-down rules.
The dietary restraint metric is hard to measure. Subsequent studies have suggested that the original DR scale may have been tapping a particular combination of high restraint and high susceptibility to disinhibition (Westenhoefer et al., 1994) or of restraint plus negative mood, which has been found to be associated with bingeing, purging, and generally more “disturbed” eating habits (Penas-Lledo et al., 2008).
In general, however, the restraint theory literature speaks to a recurring structural feature of what makes recovery from restrictive eating disorders hard: that withdrawal of rule-based restraint inevitably creates a self-regulation vacuum (at least temporarily), because that responsive middle ground is so unknown.
The middle ground was what used to appall me most about the idea of not being ill. Every day was that radical oscillation from nothing to a lot, but intentionally: fast all day, eat a large meal in the dead of night. I didn’t know how anyone could cope with the dullness of just messing around in the lowlands of neither very hungry nor very full, let alone tell me I ought to. I thought the long deep trough and the ecstatic peak were the best kinds of happiness on offer—or rather, I increasingly knew I hated it but didn’t believe anything else could be less bad.
Being well again is basically about existing in the dietary middle ground—which in turn, as I failed to realize back then, makes exploring the much more interesting extremes of other territories feasible (the highs and lows of love, sex, intellectual curiosity, professional ambition, aesthetic creativity, etc.). But crucially, recovery as the process that gets you here is not initially about inhabiting the middle ground.
I’ve written a bit before about how normality can be a slippery concept in recovery (and the more I think about it, the more misleading a guide it seems). One of the worst mistakes you can make at the start of the recovery process is to imagine that your task is to switch straight from an anorexic way of eating to a “normal” one.
Even if by normal you mean not statistically common but good and sustainable—the happily pragmatic way of eating that will sustain you for the rest of your life—you can’t get there directly because your body is incapable of self-regulating. After all, it’s had zero opportunity to do so as long as you’ve been ill.
In the next section, we’ll go on to explore what this mysterious ideal called self-regulation really is.

6: Self-Regulation, Plus or Minus Feedback
In the previous section of this self-regulation series, we took a look at some evidence suggesting that if you (1) have strict limits on eating and (2) are temporarily disinhibited for some reason (maybe you flouted a dietary rule in a small way, or you were in an altered state thanks to alcohol or strong emotion), you may well break your food rules by a wide margin. A range of recent studies adds further support to the general idea that dietary restriction impairs behavioural self-control: for example, a study from last year found that 10 days of calorie deprivation reduces people’s food-related but not other types of self-control (Standen and Mann, 2021).
Rules and self-regulation
One way to interpret this type of evidence is that dietary rules prevent effective self-regulation. What exactly do I mean by self-regulation in this context? I guess I mean stopping and starting eating because of a sensitive, adaptable set of instincts, not a rigid and arbitrary set of requirements.
Think of the milkshake drinkers we met in Part 2, who normally exercise high dietary restraint: All they’re doing is applying an input rule, and, once it’s broken, they’re lost. Once some researchers have induced them to have the milkshake, but their rule says not to, then the rule is useless to them because it’s already broken. And so they often end up compounding their rule-breaking by eating significantly more than they would have if the rule to eat less had never existed and (maybe most paradoxical of all) more than if they hadn’t eaten anything beforehand. They do so because (1) they have an awful lot of unfed hunger generated by following the rule, and (2) what the hell, there’s no difference between a rule broken by a millimetre or a mile; it’s still a failure. As soon as the boundary of the permitted is transgressed, there’s nothing left to support the supposedly desirable behaviours and resulting experiences. You’re at sea with your standard compass useless to you.
Closed loop versus open loop
In structural terms, this neatly illustrates the difference between a closed loop and an open loop. In an open loop, you have a rule (say, eat 2,000 calories a day, or run 2 x 20 minutes a day) and you apply it regardless of changes in state. This may have the advantage of simplicity: You don’t need to make adjustments based on a range of measurements. But it also means that uncertainty about the things that matter gets amplified as time passes. You can’t correct for mistakes, either because you have no idea where you are (because you’re taking no measurements of anything meaningful—e.g., you’re blind to how well or badly your life and health are panning out) or because you’re not acting properly on the measurements you are taking (e.g., because you’re afraid to do anything differently, even though you see how badly what you’re doing is serving you). You have no robustness to perturbations, in your environment or in yourself, because you have no way of even gathering the relevant information effectively, let alone acting on it to reliably counteract instability.
This is anorexia nervosa. The main form of measurement going on that’s allowed to have any appreciable effect on anything is the measuring of the input variable to which the rule attaches (e.g., you measure calories consumed so far today to decide whether to eat any more this evening). Other measuring (e.g., of bodyweight or calories in or out), even if it’s as obsessive-compulsive as the behaviour-guiding kind, probably makes no difference to the application of the rule (you eat/exercise the same regardless of your weight, or you eat the same regardless of your exercise, say). If these secondary forms of explicit measurement do have effects, it’s by supplanting any other information you might have gathered about your current state (e.g., feeling tired or unwell, being injured, being hungry).
Thus, returning to the milkshake and ice cream experiment, things get eaten or not eaten depending on (say) a daily calorie rule, and if this rule is prevented from being applied (e.g., some calorie value is concealed from you) or is broken (e.g., overshot because some nasty experimenter induces you to have something all sugary and fatty and unnecessary), the whole thing collapses, because there was never anything else to turn to. As with cruise control that isn’t actually measuring vehicle speed and/or is assuming no disturbances (by road surface, gradient, etc.), the system is fundamentally fragile. It may be stable for a while by accident, but it won’t stay that way for long in the real world.
By contrast, in a closed-loop system, the information gathered is used to determine the next action. For example, say in the open-loop context you have a fixed exercise routine that proceeds every single morning regardless of physical or mental state, location, busyness, etc. The rule (e.g., x reps by y sets @ intensity z) is always obeyed and never changes. In a closed loop, the type, length, intensity, and sheer presence or absence of the routine adjust to all the relevant factors in the organism and its environment. The things being measured are dynamic; the measurements are about assessing the current state in order to be able to act in line with it.
Most of the closed-loop measurements don’t even feel like measurements in the numerical sense that explicit rule following requires, because they’re automated by evolved biological mechanisms that support performance within acceptable homeostatic bounds (e.g., via ghrelin and leptin secretion, metabolic modulation, force of muscular contraction under load), as well as by automatically introspected inclinations and sensations (anything from musculoskeletal mobility to lethargy to personal/professional priorities for today). All of this is compromised if top-down exercise rules are imposed without acknowledgment of the relevant signals that could prompt self-correction.
In reality, of course, the distinction isn’t absolute. The open-loop version doesn’t manage to entirely override all the equilibrium-geared control systems of the human body. It may also incorporate limited feedback, in the form of compensation for some episodes of rule-breaking, though this is still irrespective of the wider causes and effects. Meanwhile, the closed-loop version has rules of thumb, input defaults that mean, in absence of any off-the-charts measurements of other kinds (e.g. serious DOMS [delayed onset muscle soreness], coming down with a cold), the habitual behaviour will occur, within standard bounds. This reduces the cognitive load of having to make every decision from scratch every time. But all kinds of micro-adjustments (e.g., in amount of warmup, rest time between sets, whether to go for your personal best today), are happening more effectively for the fact that the open-loop rules aren’t constantly trying to override everything else.
If you’re operating open-loop, or attempting to, in the realm of food (or any other life domain I can think of), you’re probably not living very happily, even when you’re not transgressing. So, in the penultimate part of the series, we’ll take a look at what your options are if this open-loop state sounds like you, and you’d like it not to be.

Read on to the final instalment here.
References
Herman, C. P., & Mack, D. (1975). Restrained and unrestrained eating. Journal of Personality, 43(4), 647-660. Paywall-protected journal record here.
Peñas-Lledó, E. M., Loeb, K. L., Puerto, R., Hildebrandt, T. B., & Llerena, A. (2008). Subtyping undergraduate women along dietary restraint and negative affect. Appetite, 51(3), 727-730. Full PDF download here.
Schaumberg, K., Jangmo, A., Thornton, L. M., Birgegård, A., Almqvist, C., Norring, C., … & Bulik, C. M. (2019). Patterns of diagnostic transition in eating disorders: A longitudinal population study in Sweden. Psychological Medicine, 49(5), 819-827. Open-access full text here.
Standen, E. C., & Mann, T. (2021). Calorie deprivation impairs the self-control of eating, but not of other behaviors. Psychology & Health. Paywall-protected journal record here.
Westenhoefer, J., Broeckmann, P., Münch, A. K., & Pudel, V. (1994). Cognitive control of eating behavior and the disinhibition effect. Appetite, 23(1), 27-41. Paywall-protected journal record here.
2 thoughts on “All-Inclusive Resorts and Dietary Self-Regulation (Part 2)”