This is the research companion to The Architecture of Choice. Read the main post first.
The post argues that putting it down is hard because the choice is engineered to be hard, and that your leverage lives in friction, defaults, and environment design rather than in your child’s willpower. These five studies pressure-test the links in that chain. The first two show that restructuring the choice, not lecturing about it, is what actually moves behavior -- across hundreds of real-world food experiments, and then in a single clean test with children. The third shows the long-run stakes of self-control while reframing it as something contextual and trainable rather than a fixed virtue. The fourth grounds the “developing system” claim in the biology of executive function. And the fifth, a deliberate gut-check, shows that simply removing the product, with nothing built in its place, does surprisingly little. Together they make the post’s case sharper and more honest: design beats discipline, but only design that adds something, not design that just subtracts.
Changing the Environment Beats Explaining It -- and the Gap Grows the More a Nudge Targets Behavior Instead of Belief
Cadario, R., & Chandon, P. (2020). Which healthy eating nudges work best? A meta-analysis of field experiments. Marketing Science, 39(3), 465-486.
DOI: 10.1287/mksc.2018.1128
What they found: Pooling 299 effect sizes from field experiments run in real cafeterias, restaurants, grocery stores, and schools, the authors sorted healthy-eating nudges into three families by what each one targets: cognition (giving people information, like nutrition labels), affect (making the healthy option feel appealing), and behavior (changing the environment itself, by making the healthy option more convenient, more visible, or the default). The average nudge produced a small effect (Cohen’s d = 0.23, about 124 fewer calories a day). But the progression of nudges mattered enormously. Effects climbed steadily as nudges moved from cognition (d = 0.12, about 64 calories) to affect (d = 0.24, about 129 calories) to behavior (d = 0.39, about 209 calories). The single most common public-health move, telling people what is healthy, was the weakest of the three. And nudges were consistently better at reducing unhealthy eating than at increasing healthy eating, which tracks the post’s point that subtracting the engineered pull matters as much as adding a good option.
Why this matters for you: This is the empirical backbone of the post’s central move, measured in the exact domain the post is about. The instinct to explain, to tell a child why the snack or the screen is a bad idea, is the cognitive nudge, and it is the weakest lever in this data. What reliably moved behavior was changing the structure: making the better option the convenient one, the visible one, the default one. Friction, defaults, environment design. The gradient from belief to behavior is the same gradient the post walks from lecturing to redesigning the room.
What it doesn’t answer: This is food, not screens. The authors did not test attention or media, so the jump to devices is an argument by analogy, not a measured result. Most of the studies were brief and recorded behavior at the point of choice, not whether the change held for weeks or hardened into a habit. Effects varied widely by setting, and field experiments in cafeterias cannot capture how a parent’s relationship with a child interacts with the environment they build at home.
With Children, a Tiny Reward Plus a Reason Moves Food Choices That a Lecture Alone Cannot
List, J. A., & Samek, A. S. (2015). The behavioralist as nutritionist: Leveraging behavioral economics to improve child food choice and consumption. Journal of Health Economics, 39, 135-146.
DOI: 10.1016/j.jhealeco.2014.11.002
What they found: In a cluster-randomized field experiment across 24 after-school sites serving 1,614 low-income children (ages 7 to 18), kids chose between a dried-fruit cup and a cookie. At baseline, about 17 percent chose the fruit. Attaching a small token incentive worth 50 cents or less raised fruit selection to roughly 76 to 86 percent. A three-minute educational message by itself did nothing (11 percent chose fruit, versus 16 percent at baseline). But the educational message combined with the incentive produced both the highest selection rate (86 percent) and the only condition with significant carryover a week later (26 percent still chose fruit, versus 13 percent at baseline). Loss framing did not outperform gain framing, and there was no sign that the rewards crowded out kids’ later willingness to choose healthy food.
Why this matters for you: This is the post’s argument tested directly on children and on food. Information alone -- the lecture -- moved nothing. What moved behavior was changing the structure of the choice (a small reward tied to actually eating the food), and what made it stick was pairing that structure with a brief reason. It also models the post’s “modulate to developmental capacity” point: the durable version is not pure external control or pure explanation, but a scaffold that does some of the regulating the child cannot yet do alone.
What it doesn’t answer: The follow-up was only one week, so “habit formation” is suggestive, not established. The contrast (dried fruit versus cookie) is nutritionally modest, and consumption was eyeballed by observers, not weighed. The sample was exclusively low-income Chicago after-school programs, and the intervention bypassed families entirely -- it tells us nothing about how the same moves work at a parent’s kitchen table.
Self-Control Has Real Long-Run Stakes -- But It Behaves Like a Gradient Shaped by Environment, Not a Fixed Virtue
Moffitt, T. E., Arseneault, L., Belsky, D., Dickson, N., Hancox, R. J., Harrington, H., Houts, R., Poulton, R., Roberts, B. W., Ross, S., Sears, M. R., Thomson, W. M., & Caspi, A. (2011). A gradient of childhood self-control predicts health, wealth, and public safety. Proceedings of the National Academy of Sciences, 108(7), 2693-2698.
DOI: 10.1073/pnas.1010076108
What they found: Following an entire birth cohort of 1,037 New Zealanders from age 3 to 32 (with 96 percent retention), and replicating in a second cohort of 2,232 British twins, the researchers found that childhood self-control predicted adult physical health, financial standing, and criminal convictions -- even after controlling for IQ and family social class. Crucially, the effect was a gradient across the whole population, not a cliff at the bottom: removing the lowest-scoring fifth left the pattern essentially unchanged. A large share of the long-run effect ran through adolescent “snares” -- early smoking, dropping out, teen parenthood -- that low childhood self-control made more likely. In the twin study, the sibling with poorer self-control at age 5 was more likely to be smoking, struggling in school, and behaving antisocially by age 12, holding the shared family environment constant.
Why this matters for you: This study is why the post takes self-control seriously without treating it as the lever to pull. Self-control genuinely matters for life outcomes, but it behaves like a graded trait built up over the first decade, not a switch a child can flip in the moment a product is engineered against them. Tellingly, the authors’ own policy prescription is choice architecture: they call for opt-out schemes and defaults that require no effortful self-control -- precisely the post’s argument that you redesign the environment rather than demand more willpower from a developing person.
What it doesn’t answer: This is observational and predictive, not strictly causal; even the twin design strengthens but does not seal the causal case. It treats self-control as a single umbrella construct and does not isolate which components (delay, attention, persistence) carry the weight. Both cohorts are from specific countries and eras, and the study tests no intervention -- it tells us the stakes, not how much realistic effort actually shifts the gradient.
The Self-Control “Muscle” Is a Developing, Trainable System -- and the First Thing to Fail Under Stress, Poor Sleep, and Overload
Diamond, A. (2013). Executive functions. Annual Review of Psychology, 64, 135-168.
DOI: 10.1146/annurev-psych-113011-143750
What they found: This authoritative review establishes that executive functions -- inhibitory control, working memory, and cognitive flexibility -- are the top-down processes that let a person resist a tempting automatic response. Inhibitory control is disproportionately hard for young children and matures gradually through childhood and adolescence, with cognitive flexibility (which requires both other components) emerging latest. Diamond’s signature thesis is that the prefrontal systems behind executive function are a “canary in the coal mine”: they degrade first and worst under stress, sadness, loneliness, sleep deprivation, and physical unfitness -- so a child who is tired, stressed, or lonely can look like they have an attention disorder when the real problem is an unmet need. She also reviews evidence that executive functions are trainable, that the children furthest behind benefit most, and that broad programs transfer more widely than narrow computer drills.
Why this matters for you: This is the biology under the post’s “developing brain versus professional engineering department” asymmetry. Impulse control is not a virtue a child is withholding; it is a system still under construction, and it is the first capacity to buckle under load -- which is exactly why a product optimized to capture attention has such an easy target. It also reframes parental leverage in the post’s own terms: stabilizing sleep, lowering stress, and reducing the load are not soft extras, they are the conditions that determine how much self-regulation a child has available in the first place.
What it doesn’t answer: This is a narrative review, selectively curated rather than systematic, and some training claims it summarizes (notably working-memory programs) have looked weaker in later meta-analyses. It is weighted toward Western research, and its strongest causal evidence is for exercise; claims about loneliness and sadness rest on more correlational footing. It also does not give a single tidy “matures at age X” number -- the maturation it describes is gradual and component-specific.
Simply Removing the Product, With Nothing Built in Its Place, Does Surprisingly Little
Lemahieu, L., Vander Zwalmen, Y., Mennes, M., Koster, E. H. W., Vanden Abeele, M. M. P., & Poels, K. (2025). The effects of social media abstinence on affective well-being and life satisfaction: A systematic review and meta-analysis. Scientific Reports, 15, 7581.
DOI: 10.1038/s41598-025-90984-3
What they found: This preregistered meta-analysis pooled 10 studies (38 effect sizes, N = 4,674 adults) testing what happens when people completely quit social media for a stretch. Across the board, abstinence produced no significant change in positive affect (g = 0.03), negative affect (g = -0.01), or life satisfaction (g = 0.03), and longer abstinence periods did not help more than shorter ones. The authors raise a “zero net-sum” possibility -- that the relief from comparison and distraction roughly cancels the lost connection and information. They also note that compliance was poor in several studies (in some, half the participants peeked anyway), which itself says something about how hard pure removal is to sustain.
Why this matters for you: This is the honest gut-check on the post’s thesis. The intuitive parenting move -- just take the phone away -- is removal without replacement, and for adults that produced essentially nothing. It reinforces the post’s distinction between confiscation and design: the goal is not to subtract the engineered product and hope, but to build friction, defaults, and genuinely competing options so the better choice requires less willpower, not more. The poor compliance also echoes the post -- externally imposed abstinence is fragile when the underlying need it met is left unaddressed.
What it doesn’t answer: Every included study used adults, mostly Western and student-heavy, so these null findings cannot be transferred to children or adolescents, whose developing brains may respond very differently. The breaks were mostly short (around seven days), heterogeneity was high, and blinding was impossible. Critically, and consistent with the post’s discipline, a null effect of abstinence on adult mood is not evidence that the products are harmless, and it makes no claim that screens do or do not cause any specific disorder.
Coming Up
Next up how the way your family narrates its own past, the small retellings at dinner and in the car, quietly builds the scaffold a child stands on to know who they are.
Every Functional Child Development post comes with a Dive Deeper companion. Subscribe for the full picture.