Bryan Caplan  

Behaviorism Works Because It's False

Does Liberty Come from the Sta... Right Cut, Wrong Side: Obama B...
"Behaviorism" describes a range of positions, but they all claim that we should focus on "observable behavior" instead of mere mental states.  (For more, see here, here, here, and here).  While this position is counterintuitive, it seems to work: One of the easiest ways to change the behavior of an animal - or a human - is to reward the behavior you want, and punish the behavior you don't want.  You don't need a lengthy exchange of ideas; just deploy your sticks and carrots, and watch them do their magic.  If you want your child to sleep through the night, for example, let him cry for a while.  Once you stop giving attention to kids who don't sleep, the problem usually solves itself.

If you're paying attention, though, you'll notice a critical act of hand-waving.  How do you know that "attention" is a reward, not a punishment?  Indeed, how do you know that a carrot is a reward, and a stick is a punishment?  If you're trying to motivate a carnivorous masochist, you should use the stick when he pleases you, and the carrot when he doesn't.  Less obviously, but equally critically, the behaviorist usually tries to impose reinforcement soon after the desired/offending behavior.  But not always.  If he's dealing with a rat or a toddler, he'll want immediate consequences.  If he's dealing with a banker, he might confidently punish bad behavior with a monetary fine a year after the incident.

The lesson is that even the staunchest behaviorist makes implicit assumptions about subjects' mental states.  He makes assumptions about their preferences and emotions, such as: animals like food, kids like attention, adults like money, and most creatures aren't carnivorous masochists.  He makes assumptions about their beliefs, such as: rats and toddlers won't think X is a punishment for Y unless X happens soon after Y, but bankers can see more subtle connections.  Behaviorism works because (a) animals and people do have preferences, emotions, beliefs, etc., and (b) common sense gives us lots of insight into what these preferences, emotions, beliefs, etc. typically are.  Behaviorism works, in short, because it's false: We already know enough about thoughts and feelings to design rewards and punishments that animals and humans grasp and care about.

P.S. I wish this point were original to me, but it's not.  I first heard it in high school while reading The Ayn Rand LexiconSurprisingly successful Randian psychologist Edwin Locke explains it elegantly:
Behaviorism's substitute for the mind is certain entities in the environment called "reinforcers." A "reinforcer," say the Behaviorists, is an event which follows a response and makes subsequent responses of the same type more likely. "What type of events change the probability of responding?" we ask. "Reinforcing events," we are told. "What is a reinforcing event?" we inquire. "One which modifies response probability," they reply. "Why does a reinforcer reinforce?" we ask. "That's not a relevant question," they answer. . . . To understand why a "reinforcer" reinforces, Behaviorists would have to make reference to the individual's mental contents and processes--i.e., they would have to abandon Behaviorism.

Comments and Sharing

COMMENTS (17 to date)
Hyena writes:

A reward is anything which, once applied, tends to encourage the behavior. A punishment is anything which, once applied, tends to discourage the behavior.

"En/discouragement" is defined in reference to the prevalence of the behavior relative to the past.

We can perfectly well construct a behaviorist model without reference to mental states. We don't need to assume a secret reference to mental states, either, just a nascent observation of results however poorly described. Even introspection need not yield mental states: it is sufficient to ask your mental roster what you stopped or started doing when a stimulus was introduced.

Hyena writes:

Anyhow, the central problem with this attack on behaviorism is that it is at severe risk of vitalism. Bad mojo, that stuff.

FC writes:

[Comment removed pending confirmation of email address. Email the to request restoring this comment. A valid email address is required to post comments on EconLog.--Econlib Ed.]

Andrew writes:

Hyena is right. As described in the passage you excerpt from Edwin Locke, in behaviorism, reinforcement is defined entirely without reference to internal states. Reward is something that makes the behavior that is associated with it more likely, and punishment is something that makes the behavior that is associated with it less likely. Thus, reward and punishment are defined entirely operationally and don't require a priori knowledge of what an animal/human "likes" or "dislikes". Similarly, the time gap between behavior and reward/punishment is not assumed a priori ("animals/toddlers have short memories but bankers have long memories") but established experimentally. You do some training with immediate rewards/punishments and find that it works; you do some training with delayed rewards/punishments and find that it doesn't work. If we find that the results match our "common sense," that merely suggests that common sense provides a fairly good representation of reality, not that behaviorism is false.

Jody writes:

Seems like a Hidden Markov Process to me, which means you don't *have* to know what the states actually mean, just observe behavior and then construct an input / output relationship (thank you Baum-Welch)

But if the takeaway of the post is "side information helps process design", color me shocked. (That's a shade of orange.)

Miguel Madeira writes:

"Thus, reward and punishment are defined entirely operationally and don't require a priori knowledge of what an animal/human "likes" or "dislikes"."

In theory, yes. In practice, no - a behaviorist making experiments with a mouse will not make endless experiments trying to discover what is a reward for the mouse; in practice he will simply assume that the mouse likes mouse food (or dislikes electric shocks).

Philo writes:

Well said. A banker can appreciate the connection between his behavior and the year-later punishment, a very small child will appreciate the connection only if the punishment comes within minutes, some animals require that the interval be no more than a few seconds, and in the limit we have an organism that is so stupid (perhaps because it is dead?) that behaviorism doesn't work at all.


Does behaviorism say anything more than that we can get more of a certain behavior if we provide the agent with causal factors that tend to produce more of the behavior, and we can get less if we supply factors that tend to produce less?

Doc Merlin writes:

Pfft, its not behaviorism you don't believe in, Bryan, but behaviorists. From your formulation its pretty clear that behaviorists don't exist, not that behaviorism is false.

mick writes:

To invoke another waving hand, what causes the behavior of studying behavior?

Hyena writes:

Mr. Madeira,

He does that because he already knows how organisms like it, or mice specifically, tend to react to different things. He's aware that they tend to go towards food and away from shocky things, so he'll probably start there.

Mr./Ms. Philo,

No. In fact, no scientific theory can ever say more than that because it is the limit of empirical observation. To the extent it says more, it ceases to be scientific and becomes a metaphysical theory, that is, becomes philosophy.

Behaviorism should, however, shade cleanly into neuroscience and its concern about functional relationships in the brain. Well, should once we know enough about those to build stronger ties between psychology and neurology.

David P writes:

Interesting take, I had never thought of the circular nature the reward system. It's way too easy to think of the outside, third party examining the situation as almost god-like and without fault.

My problem with behavioral economics has always been that there's an implied normative assumption behind the outcome. When someone exhibits a present bias there's an implication that they should value two outcomes equally and because they don't they're irrational. Who knows, maybe there's a very good evolutionary reason we do value these two outcomes differently that behavioral economics just hasn't figured out yet.

Bob Murphy writes:

It's the occasionally rewarding blog post like this one that keeps me coming back to Caplan, even though more often than not I am shocked.

Wes Bertrand writes:

Behaviorism, whether formal or "pop," is antithetical to reasoning beings, as it's based on extrinsic motivation instead of intrinsic motivation.

To contend that behaviorism "works," is beyond false because it overlooks all the facets of mind that are affected by one mind trying to manipulate or dominate another mind. This has been catalogued in extensive detail by psychological researcher Alfie Kohn ( Even if you're interested in solely the empirical side of this issue, scientific understanding is not on the side of behaviorism: Kohn references a plethora of ingenious studies in his book "Punished By Rewards: The Trouble with Gold Stars, Incentive Plans, A's, Praise, and Other Bribes."

And, as founder of nonviolent communication Marshall Rosenberg has noted, those who use behaviorist methods to "get persons to do things" overlook a very important aspect: What reasons do you want the person to have for doing what you want them to do?

Curt writes:

Hayena, you state: that "it is sufficient to ask your mental roster what you stopped or started doing when a stimulus was introduced." But surely this cannot be done without reference to some mental state.

After all, there is an innumerable set of things I was doing and not-doing when a "stimulus" was introduced. I was sitting, blinking, wearing a tie, etc. I was also not talking, not reading Shakespeare, not combing my hair, and so on. You can't know which of these things is being reinforced without some tacit understanding of the mental process that the person is going to associate with the reinforcement.

Reminds me of the old story about the mother who wanted to cure her two sons of swearing. She consulted a psychologist -- a strict behaviorist -- who advised her to immediately punish the boys whenever the undesired behavior occurred.

The very next morning at breakfast, son #1 sits down and says, "pass the goddam Cheerios". Without blinking, the mother swats the kid across the head and sends him tumbling out of his chair. She then turns and glares at son #2. Now what do YOU want for breakfast?" she demands.

Stunned, the boy looks at his brother, thinks for a moment and then responds: "I dunno, but you can bet your sweet ass it ain't Cheerios."

RS writes:


Behaviorism is nothing less than the blind leading the blind. ALL of you overlook the simple fact that on a behaviorists own premise, ANY AND ALL inferences that a behaviorists can make from the mere observations of someones "observable behavior" or "mere mental states" is itself just another "behavior" or in other words, according to a behaviourist, we are all just peices of meat who happen to stimulate each other randomly so therefore nothing could be conciously observed or learned from the stimulation, including the observers observations. that about sums up the entire theory. what a waste of time.

Hyena writes:


So... you're saying you have a priori knowledge sufficient to craft a complete psychological theory?

If you're not, then you don't have a problem with behaviorism.

If you are, then you have a problem with the broader scientific project of the modern world.

Gene Callahan writes:

Incredible to find that there are still people who defend behaviourism, which has been dead as a psychological theory for half a century or so.

And what Hyena thinks is the "broader scientific project of the modern world" is absurd and impossible. Limiting ourselves to strictly empirical observations, we could never even arrive at the notion of the entities that inhabit scientific theories. For instance, "trees" are a mental construct: all that we empirically observe is a patch of green colour there, a patch of brown there, etc. As Hegel put it, "objective reality" is an idealistic construct.

Comments for this entry have been closed
Return to top