As Luke helpfully taught us, negative reinforcement doesn’t seem to work as well as positive. Spanking, in particular, is right out.
I’ve seen this claim all over the comments to Luke’s post, and I don’t see Luke as asserting it (or being able to). In fact, there are only two things in Luke’s post that even slightly resemble this claim:
First is a warning against punishing people in a way such that the punishment could negatively incentivize the effort instead of the failure. This is a good point, but doesn’t apply in a lot of cases—for example, punishing a child for hitting a sibling.
Second is the following quote:
When trying to maintain order in a class, ignore unruly behavior and praise good behavior (Madsen et al. 1968; McNamara 1987).
So I looked up McNamara 1987. It’s a book chapter that reports a few studies on students getting better when praised, then says they’re very weak, no one’s been able to replicate them, and then goes into how a better study McNamara clearly finds more credible shows that students do worse with the sort of hokey planned praise used in the previous studies.
Then I looked up Madsen et all 1968. It’s a study on three problem children, Cliff, Frank and Stan. The relevant conditions were Rules, where the teachers wrote some rules on the blackboard and made the children repeat them a bunch of times, and Praise, in which the teachers praised the students for following the rules. There was no punishment condition. Writing the rules on the blackboard didn’t help much, but Cliff and Frank seemed to get better after being praised a lot (Stan was equivocal).
Poor Cliff and Frank are not a sufficient basis on which to assert incredibly broad psychological theories about how you should always reward and never punish all humans in all cases.
In the case of positive behavior/reinforcement, there’s nothing particularly problematic about this, but for the negative case, you’re also punishing being caught/noticed/seen, which can be problematic.
This is a kind-of-sounds-good argument, but one could make equally kind-of-sounds-good arguments on the other side. For example, rewards decrease intrisic motivation by replacing them with extrinsic motivation. Or anything Alfie Kohn ever said (I’m not necessarily endorsing him, just holding him up as an example of arguments that rewards too have their problems).
Finally, I am very suspicious of the studies in this area. Not only is developmental psych generally a minefield, but “rewards work better than punishments” is so in tune with modern sensibilities that I would be really surprised if the average researcher doesn’t go out gunning for that result and quietly file anything that shows punishments are better than rewards into a file drawer never to see the light of day.
(one counterexample is a recent study that shows spanking children generally gives them better life outcomes. I’ve seen a zillion studies that say the opposite, and I consider the entire project of observational studies of spanking totally useless because of the confounding effects of parental choice—ie parents who decide to spank or not spank their children are probably different in many other correlated ways which are much more important—but the study exists and I don’t consider the spanking field nearly as settled as you do)
To speak to goal 1 first, Bryan Caplan claims flat outcomes for goal #1 under commonly tried parenting interventions, which seems counter-intuitive. More explanation of what exactly the studies in question proved would be welcome.
I’ve never read Caplan, but I’m working my way through The Nurture Assumption which is probably what Caplan’s working off of. Nurture Assumption claims pretty much everything is flat for pretty much all kinds of parenting and we need to drop the assumption (hence the name) that parenting matters much beyond just not starving or abusing your kid. The books presents lots of complicated evidence but the strongest is adoption studies, where children raised together aren’t much more similar than children raised apart (with degree of genetic relatedness held constant). Right now I believe her about 60%.
When trying to maintain order in a class, ignore unruly behavior and praise good behavior
Problem with this advice is that even if you ignore the unruly behavior, if other classmates don’t ignore it, it is rewarded anyway. When you have only one unruly child in the classroom, you can use this rule. When you have two or three of them, and they encourage each other, you are out of luck.
I also don’t recall claiming that positive reinforcement works better than negative reinforcement.
Also, it looks to me like what MBlume means to say is that positive reinforcement works better than punishment (a different concept than negative reinforcement), which is also something I don’t recall claiming.
In any case, thanks again Yvain for your analysis of some of the relevant studies.
I’ve seen this claim all over the comments to Luke’s post, and I don’t see Luke as asserting it (or being able to). In fact, there are only two things in Luke’s post that even slightly resemble this claim:
First is a warning against punishing people in a way such that the punishment could negatively incentivize the effort instead of the failure. This is a good point, but doesn’t apply in a lot of cases—for example, punishing a child for hitting a sibling.
Second is the following quote:
So I looked up McNamara 1987. It’s a book chapter that reports a few studies on students getting better when praised, then says they’re very weak, no one’s been able to replicate them, and then goes into how a better study McNamara clearly finds more credible shows that students do worse with the sort of hokey planned praise used in the previous studies.
Then I looked up Madsen et all 1968. It’s a study on three problem children, Cliff, Frank and Stan. The relevant conditions were Rules, where the teachers wrote some rules on the blackboard and made the children repeat them a bunch of times, and Praise, in which the teachers praised the students for following the rules. There was no punishment condition. Writing the rules on the blackboard didn’t help much, but Cliff and Frank seemed to get better after being praised a lot (Stan was equivocal).
Poor Cliff and Frank are not a sufficient basis on which to assert incredibly broad psychological theories about how you should always reward and never punish all humans in all cases.
This is a kind-of-sounds-good argument, but one could make equally kind-of-sounds-good arguments on the other side. For example, rewards decrease intrisic motivation by replacing them with extrinsic motivation. Or anything Alfie Kohn ever said (I’m not necessarily endorsing him, just holding him up as an example of arguments that rewards too have their problems).
Finally, I am very suspicious of the studies in this area. Not only is developmental psych generally a minefield, but “rewards work better than punishments” is so in tune with modern sensibilities that I would be really surprised if the average researcher doesn’t go out gunning for that result and quietly file anything that shows punishments are better than rewards into a file drawer never to see the light of day.
(one counterexample is a recent study that shows spanking children generally gives them better life outcomes. I’ve seen a zillion studies that say the opposite, and I consider the entire project of observational studies of spanking totally useless because of the confounding effects of parental choice—ie parents who decide to spank or not spank their children are probably different in many other correlated ways which are much more important—but the study exists and I don’t consider the spanking field nearly as settled as you do)
I’ve never read Caplan, but I’m working my way through The Nurture Assumption which is probably what Caplan’s working off of. Nurture Assumption claims pretty much everything is flat for pretty much all kinds of parenting and we need to drop the assumption (hence the name) that parenting matters much beyond just not starving or abusing your kid. The books presents lots of complicated evidence but the strongest is adoption studies, where children raised together aren’t much more similar than children raised apart (with degree of genetic relatedness held constant). Right now I believe her about 60%.
Problem with this advice is that even if you ignore the unruly behavior, if other classmates don’t ignore it, it is rewarded anyway. When you have only one unruly child in the classroom, you can use this rule. When you have two or three of them, and they encourage each other, you are out of luck.
I also don’t recall claiming that positive reinforcement works better than negative reinforcement.
Also, it looks to me like what MBlume means to say is that positive reinforcement works better than punishment (a different concept than negative reinforcement), which is also something I don’t recall claiming.
In any case, thanks again Yvain for your analysis of some of the relevant studies.
For the record, I, too, thought that was a claim of your post. Maybe it’s worth adding a note to it to clarify since this mistakes seems wide spread.