Primary vs secondary reinforcers: the difference, and the one question that sorts most examples

A primary reinforcer strengthens behaviour without any learning. Food, water, sleep and touch are examples. A secondary reinforcer has no power of its own. It gets its power when it is paired with a primary reinforcer, or with another reinforcer that already works. Praise, money, tokens and the click of a dog trainer's clicker are examples. A secondary reinforcer is also called a conditioned reinforcer.

The definitions are short. The exam question is harder. Is praise primary because it feels good? Is money primary because everybody wants it? Textbooks also use two sets of names for the same two things. Here is the distinction, the question that sorts most examples, and the three places where the line is not sharp.

• Primary means unlearned. Secondary means learned by pairing. "Biological need" is a useful hint, but it is not the definition.
• Primary reinforcer and unconditioned reinforcer are the same thing. Secondary reinforcer and conditioned reinforcer are the same thing.
• A secondary reinforcer that depends on one reinforcer loses its power when that reinforcer stops. A primary reinforcer works only when you are deprived of it.
• Money is a generalized reinforcer: a secondary reinforcer that is paired with many primary reinforcers. B. F. Skinner gave a definition of the term in 1953.
• The pair is not the same as intrinsic and extrinsic motivation. It is also not the same as positive and negative reinforcement.

The difference in one table

Two sets of names for one distinction

The open textbook Psychology 2e from OpenStax says primary and secondary. The research reviews that this article cites say unconditioned and conditioned. The APA Dictionary of Psychology gives both sets. Its entry for primary reinforcement says "Also called unconditioned reinforcement". Its entry for secondary reinforcement says "Also called conditioned reinforcement".

Skinner mixed the two sets. In Science and Human Behavior, published in 1953, he wrote "primary reinforcer" for the first kind and "conditioned reinforcer" for the second kind. The word "conditioned" tells you how the second kind is made: by conditioning.

Thus, when an exam question says conditioned reinforcer, read secondary reinforcer. When it says unconditioned reinforcer, read primary reinforcer. The meaning does not change.

Primary reinforcers: no learning needed

The APA Dictionary defines primary reinforcement as a process in which a consequence makes a response more probable. The consequence does this "without the need for special experience with the stimulus or circumstance". OpenStax says the same thing in fewer words. Primary reinforcers "have innate reinforcing qualities" and "are not learned". Its list is "Water, food, sleep, shelter, sex, and touch, among others".

Two details are easy to miss.

First, a primary reinforcer does not always work. Skinner wrote that reinforcement with food shows its effect "only under an appropriate condition of deprivation". In his words, "when we reinforce with food, we gain control over the hungry man". After a large meal, food is a weak reinforcer.

Second, the definition says unlearned. It does not say necessary for survival. Many study pages define a primary reinforcer as something that satisfies a biological need. That is a good hint, because most textbook examples are needs.

But OpenStax adds that "Pleasure is also a primary reinforcer", and it puts touch on its list. Skinner also allowed for unlearned reinforcers that are not needs. He thought it possible that some of the reinforcing effect of "sensory feed-back" is unconditioned, and his example was a baby's rattle.

How do you know that an event is a reinforcer at all? Skinner's answer was "a direct test". You make the event follow a response, and you measure if the response becomes more frequent. A reinforcer is defined by this effect, not by how pleasant it looks.

Secondary reinforcers: learned by pairing

OpenStax defines the second kind in one sentence: "A secondary reinforcer has no inherent value and only has reinforcing qualities when linked with a primary reinforcer." The APA Dictionary says that a neutral stimulus gets its effect "by virtue of being paired with another stimulus". The stimulus then works "only after special experience or training".

Skinner's example was a pigeon and a light. "If each time we turn on a light we give food to a hungry pigeon, the light eventually becomes a conditioned reinforcer." You can then use the light to teach a new response "just as food is used". He named three conditions:

• More pairings make the light more reinforcing.
• The food must follow the light after a short interval.
• The light loses its power quickly when all food is withheld.

This pairing is the process of classical conditioning. Skinner wrote that the process "appears to be the same". In classical conditioning, a stimulus that predicts food starts to cause a response. Here, a stimulus that predicts food starts to reinforce a response.

The APA Dictionary's example is a dog trainer. Each time the dog sits on command, the trainer gives a treat and makes a noise with a clicker. "Eventually, the clicker noise itself (the conditioned reinforcer or secondary) can be used alone to maintain the desired behavior."

A secondary reinforcer does not need a primary reinforcer directly behind it. In a 2009 review of token reinforcement, Timothy Hackenberg described how a token gets its function. It does so "through relations to other reinforcers, both unconditioned (e.g., food or water) or conditioned (e.g., money or credit)". A voucher that you exchange for money is still a secondary reinforcer.

Why are secondary reinforcers so important? Skinner's answer was that "only a small part of behavior is immediately reinforced with food, water, sexual contact, or other events of obvious biological importance". Conditioned reinforcers fill the time between an act and its final result. His example was a man who puts storm windows on his house in October. The warm house comes in January.

Wolfe's chimpanzees and the poker chips

The first laboratory studies of tokens used chimpanzees. Hackenberg's review describes them. Wolfe (1936) and Cowles (1937) worked at the Yerkes laboratories. Their chimpanzees learned to put poker chips into a vending machine that gave grapes or water.

Wolfe first showed that the pairing was important. White chips bought grapes, and brass chips bought nothing. The chimpanzees started to choose only the white chips, and they stopped putting brass chips into the machine. When both kinds lay on the floor, the animals picked up only the white chips.

Then the chimpanzees had to work for the chips. They lifted a bar with a weight on it. In one experiment, sessions that paid in chips alternated with sessions that paid in food. For 3 of the 4 animals, there was "little difference" in the time they needed to complete the trials. The fourth animal worked faster for food.

Generalized reinforcers: why money works when you are not hungry

A secondary reinforcer that depends on one primary reinforcer has a weak point. Skinner wrote that behaviour strengthened by a conditioned reinforcer "varies with the deprivation appropriate to the primary reinforcer". If the chip buys only food, the chip is worth little to an animal that is full.

Wolfe saw this effect. He gave three chimpanzees a choice between black chips for peanuts and yellow chips for water. Before a session, the animals had no food or no water for 16 hours. According to Hackenberg, all three "generally preferred the tokens appropriate to their current deprivation conditions".

Skinner's answer to the weak point was the generalized reinforcer. "A conditioned reinforcer is generalized when it is paired with more than one primary reinforcer." At least one of the matching deprivations is then probably present at any moment. In his words, "When we reinforce with money, for example, our subsequent control is relatively independent of momentary deprivations."

He listed these generalized reinforcers:

• Attention, because only people who attend to you reinforce your behaviour.
• Approval, such as a smile or the words "That's right".
• Affection.
• The submissiveness of other people.
• Tokens, such as money, marks, grades and diplomas.

Money was his clearest case. "It is the generalized reinforcer par excellence because, although 'money won't buy everything,' it can be exchanged for primary reinforcers of great variety." He added that we can usually watch how a child comes to be reinforced by money. A generalized reinforcer can also keep its power after the primary reinforcers stop. Skinner's example was severe: "A miser may be so reinforced by money that he will starve rather than give it up."

Skinner also explained why people work so hard for attention and approval. These reinforcers are difficult to see, so they follow behaviour only some of the time. He connected the result to intermittent reinforcement, which makes behaviour persistent.

One caution applies. Skinner's account is logical, but direct tests are rare. Hackenberg wrote that a generalized token should be more durable, "although there is surprisingly little empirical support for this". He added that the generalized functions of tokens "have yet to be explored".

The one question that sorts most examples

Ask one question: would this consequence work on a person or an animal that never learned anything about it?

• If the answer is yes, it is a primary reinforcer.
• If it works only because of a pairing, it is a secondary reinforcer.
• If it was paired with many different reinforcers, it is a generalized secondary reinforcer.

Do not sort by strength. Skinner's miser starves before he gives up his money, and money is still secondary. Do not sort by how pleasant the consequence is. Sort only by learning.

Where the line blurs

The line is not sharp at three points.

The first point is affection and touch. Skinner treated attention, approval and affection as learned reinforcers. He argued that a capacity to be reinforced by them "could scarcely have evolved" in the short time that the necessary conditions have existed. Harry Harlow disagreed in 1958. He said that psychologists of his time treated love and affection as "derived or secondary drives", learned because the mother gives food.

Harlow raised infant rhesus monkeys with two artificial mothers, one made of cloth and one made of wire. For four monkeys the cloth mother gave milk, and for the other four the wire mother gave milk. With age, the monkeys fed by the wire mother spent less time on her and more time on the cloth mother. Harlow concluded that "contact comfort is a variable of overwhelming importance" and that feeding was "a variable of negligible importance".

That result agrees with the OpenStax list, where touch is primary and praise is secondary. But Harlow's subjects were monkeys. For human attention and approval, the question is not closed.

The second point is sensory feedback. Skinner thought it possible that the reinforcing effect of a baby's rattle is unlearned. If he was right, a primary reinforcer does not need a connection with hunger or thirst. The border of the primary group is thus not fully clear.

The third point is what a secondary reinforcer does. In the usual account, it strengthens a response in the same way as food does. In a 2010 review, Timothy Shahan described a different account. In that account, the stimulus is a signpost: it serves "to guide rather than strengthen behavior". It tells the animal how or where to get the primary reinforcer. Shahan called the signpost account "a plausible alternative interpretation".

For an introductory exam, use the usual account. For a seminar, know that the question is open.

What the distinction is not

The pair is not intrinsic and extrinsic motivation. Some pages say that primary reinforcers cause intrinsic motivation and secondary reinforcers cause extrinsic motivation. That mixes two different questions.

Ryan and Deci (2000) defined intrinsic motivation as "the doing of an activity for its inherent satisfactions rather than for some separable consequence". Candy for finished homework is a primary reinforcer. But the candy is a separable consequence, so the motivation is extrinsic. Our article on intrinsic and extrinsic motivation explains that pair.

The pair is also not positive and negative reinforcement. Positive and negative tell you if a consequence is added or removed. Primary and secondary tell you where the consequence got its power. Our article on negative reinforcement and punishment explains the first pair.

Last, learned consequences do not only reinforce. They can punish too. Skinner wrote that "we punish by disapproving, by taking money away, as in a legal fine". Hackenberg's review makes the same point for tokens: "If tokens function as conditioned reinforcers, then the loss of tokens should function as conditioned punishers."

Keeping the terms apart

Three short rules help in an exam:

• Primary comes first, because it needs nothing before it. Secondary comes second, because it needs a pairing first.
• Conditioned means made by conditioning, so a conditioned reinforcer is a secondary reinforcer.
• Generalized means paired with many reinforcers, and money is the model.

These terms are easy to mix up because you meet them as pairs. Our guide to memorising psychology terminology shows how to study pairs of terms that look alike.

Psychology World is a psychology dictionary made for this kind of practice. You look up a term, save it to one of your own lists, and then drill the list. You can drill it as flashcards, as multiple-choice questions when the list has four or more terms, or by typing the answer. A daily word also brings a term back without a search. The four steps of the learning loop and what each study mode does are on the home page.