In 1957, Charles Ferster and B.F. Skinner published a book cataloguing what happens when you vary how a reward arrives while holding constant what it is.
The finding that outlived everything else: the schedule matters more than the reward.
The four basic shapes
Reinforcement can be scheduled on time or on actions, and predictably or unpredictably. That gives four combinations.
Fixed interval — reward available after a set time. Produces bursts of activity right before the deadline and very little otherwise. Studying for an exam.
Fixed ratio — reward after a set number of actions. Steady work, a pause after each payoff. Piecework.
Variable interval — reward after an unpredictable amount of time. Slow, steady, persistent checking. Waiting for a reply.
Variable ratio — reward after an unpredictable number of actions. And this one is different in kind.
Variable ratio produces the highest response rates and the strongest resistance to extinction of any schedule. When the reward stops entirely, VR-trained behaviour keeps going longest, because "nothing yet" is indistinguishable from a normal dry stretch. There's no signal that says stop.
This is the slot machine. It is not a coincidence that it's also the feed.
What actually runs on VR
Almost everything designed to be checked.
Scroll a feed: most items are nothing, occasionally something lands. Unpredictable payoff per action. Refresh an inbox, a notification tray, a comment section, a dashboard, a marketplace — same structure. Each action might pay. Usually doesn't.
The design didn't have to be deliberate to have the effect. Any system where interesting content arrives irregularly and you can check at will is a variable-ratio schedule, whether or not anyone intended it.
Which produces the thing people find hardest to explain about their own behaviour: the checking that isn't enjoyable and isn't decided. VR trains persistence, not pleasure. Those are separate outputs.
The counterintuitive part
The instinct is to reduce total consumption. Cut the hours, delete the app, set a limit.
Schedule logic suggests a different lever. If unpredictability is what trains hardest, then converting one frequent unpredictable reward into a predictable one does more than removing several predictable ones — even though the second option looks like more discipline.
Checking messages twice at set times is a fixed-interval schedule. Same messages, same total minutes, radically different training signal. The behaviour it builds is different because the schedule it runs on is different.
That's why "screen time" as a metric misses. Two hours on VR and two hours on a predictable schedule are not the same two hours, and a number that counts only duration can't tell them apart.
How Myo weights it
Every logged reward in the model carries a schedule, and schedule multiplies the effective cost. Fixed sits at the baseline. Variable interval above it. Variable ratio highest — and VR events take a second penalty in the peak-hygiene score, because frequency of unpredictable reward is the pattern that predicts a sliding baseline best.
Two honest notes. The ordering — variable ratio above variable interval above fixed — is textbook. The specific multipliers are our parameterization, not measured constants, and we publish that distinction openly.
What to do with this
Don't audit your hours. Audit your unpredictability.
Which of your daily rewards pay off on a schedule you can't predict? Which of those do you check most often? Take the single worst one and give it a shape — a time, a cadence, anything that makes the payoff predictable.
One schedule change beats three deletions. That's not a productivity trick; it's what the 1957 data says about which variable is doing the work.