On this page
You solve a full worksheet of 30 integration problems, then check the answers the next day and find you made the same mistake in 22 of them. B. F. Skinner saw this problem in a school classroom in 1953 and thought it was absurd. His explanation, and the machine he built that afternoon, still say something useful about how to practise.
Who was B. F. Skinner?
Skinner was born on 20 March 1904 in Susquehanna, Pennsylvania. According to the biography published by the B. F. Skinner Foundation, written by his elder daughter, he first wanted to be a writer. After a year in which he wrote very little, he came across books by Pavlov and Watson and enrolled in psychology at Harvard at 24. He later taught at Minnesota and Indiana and returned to Harvard in 1948. He died on 18 August 1990, the day he finished his last article.
What is operant conditioning?
At Harvard, Skinner built a device called the cumulative recorder, which traced every lever press by a rat as a rising line, so the slope showed the rate of responding. It revealed that the rate depended not on a stimulus before the press, as Pavlov's reflex work suggested, but on what followed it. He called this kind of behaviour operant, because it operates on the environment, and the process of shaping it through consequences operant conditioning. The studies appeared in his first book, The Behavior of Organisms (1938).
Three ideas from this work matter most for students:
Reinforcement: a consequence that makes a behaviour more likely next time.
Shaping: reinforcing small steps towards a target, adjusting what you ask for to the learner's current level.
Schedules: the timing of reinforcement changes behaviour, a topic his lab studied in depth in Schedules of Reinforcement (1957, with Charles Ferster).
For the earlier law his work built on, see our post on Edward Thorndike and the law of effect.
Why did Skinner build teaching machines?
On 11 November 1953, Skinner sat at the back of his younger daughter's fourth-grade maths class on Father's Day. The Foundation's account describes what struck him. Some children had no idea how to solve the problems while others raced through without learning anything new, so the work was not matched to each child's level. And the children did not find out whether one answer was right before doing the next. They finished a whole page, then waited, often until the next day, for feedback.
That afternoon he built his first teaching machine, which gave problems one at a time with feedback after each. Within three years he had developed programmed instruction: material broken into small steps, with the early responses prompted and less help given as performance improved. He set out the case in Teaching Machines (1958) in the journal Science.
Do rewards help or hurt learning?
Skinner's idea is sometimes reduced to "bribe students to study", which is not what he studied. A meta-analysis of 128 experiments by Deci, Koestner and Ryan (1999) found that expected, tangible rewards for doing or finishing a task tended to lower people's later free-choice interest in it, more so for children than for college students. Positive feedback did the opposite and raised interest. Our post on rewards and motivation covers the debate.
For a student, the useful reinforcer is usually information: knowing at once that a step is right.
What can a student take from Skinner's way of working?
Skinner was an engineer of learning. He watched what happened, then changed the set-up. Do the same with your practice.
Check after each problem, not each worksheet. Solve one, check it, then solve the next. If you got it wrong, fix the method before you repeat it.
Break hard topics into steps. For rotational dynamics, practise moment of inertia alone, then torque, then both together.
Fade the help. Start with a worked example open beside you, then with only the first line, then with nothing.
Make starting easy. Begin each session with one problem you can solve, so the first thing that follows sitting down is a success.
Avoid big tangible rewards for study itself. In the meta-analysis, expected prizes lowered interest while specific positive feedback raised it.
Skinner's ideas fed into mastery learning, where you move on only after you have shown you can do the current step. When you need to check a step at once rather than tomorrow, the Fermi Problem Library has a step-by-step solver that explains the reasoning, so you can compare your working one step at a time.
Tomorrow, do ten problems with a check after each one, and count how many errors you caught before they repeated.
