The old saw of operant conditioning is: "Consistent Punishment/Random Reward".
When dealing with people, the punishment part seems to be fairly simple. If someone does something you would like to discourage, you need to make them regret it. Really though, that's not exactly the case. It's too much like vengeance. What you actually need to do is to make sure that they do no benefit from the behavior. That's a little trickier, as the benefit that someone -- child or adult -- receives from a certain behavior may not necessarily be what you would think. In order to effectively counter the behavior, you first have to determine what benefit the person receives from it.
The most immediate example I can think of is the child who is generally ignored by their parents. Positive behaviors only bring less attention. Negative behaviors, specifically dangerous or combative ones, will bring immediate and intense attention. In delivering this kind of attention (i.e. yelling in the grocery store, an extended rant, a backhand), the parent thinks they are dissuading the child from future behavior of the same type. They think that the kid is just being bad because they don't respect their authority enough/are a bad ass/whatever. In reality, the kid has been conditioned to understand that the only way they can receive the benefit of parental attention is through bad behavior. Which is what screws up kids like that so badly.
The point of the example, though, is that to deliver an effective deterrent, you need to determine what the benefit of the behavior is and remove it. It's not always easy to do.
On the other side of the equation is the reward, which is what got me thinking about this topic. I was wondering: at what frequency do rewards, even random ones, lose their effectiveness? Anything that is seen as a treat, by children or adults, can easily lose its "specialness" if had too often. Clearly, it is the law of diminishing returns in effect. Of course, it doesn't only apply to rewards and positive reinforcements. It applies to anything. And, there seems to be a kind of "refresh" button on the diminishing returns, allowing an enjoyment that has become passe to regain some of its former attraction if left alone for long enough.
This would make a great and very useful study for a post-grad student or department that deals with childhood development or adult behavior issues. The active question would be: "How often is too often?" For children at different developmental stages, what are the reward frequencies that will produce optimal outcomes and what are the frequencies that will produce a spoiled brat? The same holds true for adults: how often can we do something that feels "special" before it loses that fresh quality?
My guess is that it would fall along some kind of normal distribution, with most people in an age group clustering around a certain time value. I'm also going to guess that this will be fairly hard-wired, cutting through socioeconomic groups. Just a guess, though. I also wouldn't be surprised if a person's particular threshold for reward timing is linked to their ability to perceive patterns in data points. As soon as you are able to perceive the pattern in the reward structure, it becomes significantly less effective. Very small kids, say, two year olds, probably can't abstract the pattern even if a reward is given every other day. To them, it's still Surprise-o-Fun-Time when it happens. But kids just a few years older than that will begin abstract that to "it happens a couple times a week," and it loses it's glamor. Of course, this is going to vary with the scope of the reward, but that can be taken into account.
Really, once you get past the unintentional reward via the poorly thought out punishment, over-rewarding is the next biggest problem. The first grows adults that believe that they have to do wrong to get noticed. The second grows adults that believe that they are entitled to constant reward. Neither are really acceptable, and some concrete data on this would be a great tool for both parents and employers.
