The Handover Rule: When a Support Becomes a Bribe

Many garages hold two supports for the same goal side by side, a pair of training wheels and a balance bike without pedals, and both disappear sooner or later, because no child is meant to grow up with them. In 2022 Blommenstein and van der Kamp surveyed parents in retrospect, and according to those reports children who had practised on a balance bike cycled independently at a younger age than children with training wheels. The difference lies neither in how long the support lasts nor in what it costs, but in what it takes off the child: the training wheel takes over balance, which is exactly what the child has to learn, and the balance bike takes over pedalling, which the child does not need yet.

Companies build supports like these all the time, they just call them bonuses, points, leaderboards, seminars, coaches or learning platforms, and nobody expects them to stay forever either. Unfortunately you cannot tell from the outside which of the two kinds a support is, because at the start both do the same thing and make a behavior possible or attractive. You only notice the difference on the day it ends and the behavior either carries on as if nothing had happened or collapses that same afternoon.

In July I proposed a touchstone for reward systems here, the switch-off test from the article on bribery design: you take the reward away, in your head or for real, and look at what remains of the behavior. The test is right, it just comes too late, because when a behavior collapses at switch-off the support has long been paid for and has often cost more than no support at all. Uri Gneezy and Aldo Rustichini introduced a small fine for late pick-ups at day-care centres in Haifa, after which more parents arrived late, and the number did not fall once the fine was removed. That is why I moved the question forward, into a rule I call the Handover Rule, which has been available as a working paper since this month.

Where the expectation points

So the rule asks neither whether a reward is given nor whether it ends at some point, but where a person's expectation points while the support is still present. The image of the scaffold comes from building and carries exactly this logic, because a scaffold stands beside the wall and not in it, and a wall that leans on it while it is being built does not progress when the scaffold comes down, it falls. The word bribe is a term of the rule and is not meant in a legal sense: it describes how a support is built, not what the person who introduced it had in mind.

One possible account of why the direction of expectation matters so much comes from the work of Wolfram Schultz, Peter Dayan and Read Montague. According to their finding, certain dopamine neurons signal not how good an outcome was, but whether it turned out better or worse than expected, and once a cue reliably announces the reward, they fire at the cue while the reward itself triggers little. If the reward fails to arrive after the cue, their activity drops below baseline at exactly the moment it was due. On this reading, for someone who is reliably rewarded for an activity the activity itself becomes the cue, and when the reward ends it keeps announcing something that no longer comes, so the disappointment sticks to the activity. The three findings are well established, but nobody has yet measured the whole chain down to the collapse of behavior in a single experiment.

The same support can also teach something quite different, because while it is present the activity often already gives something back on its own: you notice you are getting better, a colleague nods, the customer calls back. If that happens reliably, people learn to expect these outcomes too, and when the support goes they are still met. The rule calls the moment from which the activity or a stable habit carries the behavior the handover, and you recognise a successful handover by the fact that almost nothing happens when the support ends.

Two receivers, and sometimes none

A behavior can be handed over to two receivers, and the first is the activity itself, which at some level of skill gives something back, the balance on the bike or the song you can finally play. This receiver answers slowly because it requires skill, but it travels with you, since someone who enjoys playing the piano also plays in a new flat. The second receiver is a habit that a fixed cue starts in a stable setting, and it needs repetition rather than skill: among 39 participants whose data fitted the model well, the median time to near-maximal automaticity was 66 days, with a range of 18 to 254 days. Habits also depend on their place, and among students who moved to a new university they survived only where the new circumstances resembled the old ones.

Some activities, however, have no receiver, because they give nothing back at any level of skill and occur too rarely to become a habit, and the expense report is the friendliest example I can think of. Then a reward is not a scaffold but open pay, and it may stay as long as it is named as pay. It only becomes a bribe when it is dressed up as enthusiasm or loyalty, and the better investment is usually to reduce friction so that the activity needs less pay. That fits a meta-analysis of 128 experiments by Edward Deci, Richard Koestner and Richard Ryan, which found the damage of expected tangible rewards mainly for tasks people had enjoyed beforehand, because where there is nothing, a reward has nothing to crowd out.

The preliminary question: who else reads the bonus

At an industrial laundry company, Timothy Gubler, Ian Larkin and Lamar Pierce studied what an attendance award does, and the result is, unfortunately, more instructive than the award itself. Among workers with attendance problems punctuality improved, while workers with excellent attendance before the award became less punctual in periods when they were not eligible, and the authors attribute this to perceived inequity. The award was meant for those who came late, but everyone read it, and the punctual ones took from it a message about their own effort, which had gone unrewarded until then.

Before the four test questions there is therefore a preliminary question: who already shows the behavior without the support, what carries it, and what does the support tell these people? What carries a behavior without support is called a carrier in the rule, and it can be enjoyment, as with the children in Lepper, Greene and Nisbett's study who liked drawing anyway, a sense of duty, as with the punctual parents in Haifa, or a habit. From this follows a proposal that has not been tested yet: a support that makes a behavior possible, such as coaching or less friction, can be aimed at those who lack a carrier, while a support that makes a behavior attractive is read by everyone. The direction of that signal is open, by the way, because in a field experiment by Bradler and colleagues recognition for the best performers raised later performance mainly among those who had not received it.

A sales team, checked once through

A sales team introduces a new customer system, and management plans points for every entry, a ranking in the monthly meeting and a bonus for the top three, which is pretty much what people usually plan in such cases. The preliminary question already shows the first problem, because some sales people have long kept careful notes, in their own files and out of habit, and a bonus per entry tells them that notes are something one writes for points. The rebuilt plan therefore pays nothing for entries and gives personal help with entering data to those who ask for it.

The first test question can be answered with yes, provided the system really helps, because a well-kept system reminds the sales person to call back in time, and a timely call-back brings orders. If it does not help in practice, there is no activity to hand over to, and then the problem is the tool and not motivation. The plan fails the second question, because points per entry reward volume and not use, so whoever creates many empty entries wins. In the rebuilt plan every reminder therefore shows which note it came from, and the sales person sees how yesterday's note saved today's call. In a seminar the same question asks whether participants practise what is hard in daily work, and a role play with a friendly colleague who puts up no resistance is the training wheel among learning formats.

The plan fails the third question as well, because the ranking comes at the end of the month, stands apart from the customer call and mainly says that one complied. In the rebuilt plan feedback arrives at the moment of use, and if there is recognition, it names a concrete order that depended on a good note. For the fourth question the extra support, a weekly reminder round and personal help with entries, runs for two to three months and is then withdrawn in steps. After six weeks a pause week without the reminder round follows, and one records the share of customer calls that still lead to a note. The figures in this example are chosen for illustration.

A note field that opens by itself after every call is not a scaffold, by the way, but a permanent part of the environment that gives the habit its fixed cue and may therefore stay. A reward cannot be made permanent in the same way, because once it is reliably expected it triggers little, and to keep the same pull it would have to grow. If it stays the same, on this reading it turns into pay: its presence adds little, and its absence is felt as a loss. The budget in the rebuilt plan has stayed roughly the same, and the only thing that changed is the expectation it trains, because before only the ranking could fulfil it and afterwards the next call-back does.

The residual share on Monday morning

The rule's measure is deliberately plain and is called the announced pause: the support stops for a short period, and one records in what share of the real occasions the behavior still occurs. This residual share should rise from pause to pause, and in learning the most honest pause is the first Monday morning without the trainer. Every pause is itself a small removal, though, and can cause the disappointment it is meant to measure, which is why it should be short and announced, and even four yes answers to the test questions guarantee nothing, because they are design checks after which the behavior still has to be observed.

The Handover Rule grew out of consulting practice, and its working paper is conceptual: its components rest on established research on reward prediction, habit formation and transfer of training, nobody has tested the rule as a whole yet, and the paper itself calls one of its six propositions the most exposed. The core of the rule is surprisingly old, since Trevor Stokes and Donald Baer already found in 1977 that the most common strategy for carrying learning into daily life was to train and hope, and they recommended moving control of behavior from contrived consequences to stable natural ones. Of all schools, the one of B. F. Skinner, which points and leaderboards like to cite, has held this rule for half a century, so the problem of many reward systems is not too much Skinner but that they ignore what the behavior analysts themselves learned.

Whoever introduces a bonus, a points system or a seminar is putting up a scaffold, and in this picture is the site manager who already knows, while building it, the day it will come down again. The preliminary question can be put to any support that is running right now, and it needs no study, only a calm look at the people who already show the behavior today: what has the support told them over the past weeks about their own effort?

Source: Rackwitz, R. (2026). The Handover Rule: A test for when a support becomes a bribe (Working Paper, Version 1.0). Zenodo. https://doi.org/10.5281/zenodo.24196843