ACM Logo  An ACM Publication  |  CONTRIBUTE  |  FOLLOW    

A Science for e-Learning
Understanding B.F. Skinner's Work in Today's Education

By Julie S. Vargas / September 2010

Print Email
Comments (8) Instapaper

Advances in devices for e-learning lessons appear at a dizzying pace. Touchscreens, tablets, and phones now enable students to take lessons anywhere. Instructional procedures of e-learning, however, lag behind. The effectiveness and quality of online teaching would leap forward by taking more advantage of the science that originated in the work of B. F. Skinner.

Skinner discovered that behavior changes moment to moment through the selection of some actions but not others. A child does not learn to talk by listening to others talk (or we could all easily learn a foreign language). Listening is part of learning, but the critical part is behaving: making sounds and getting feedback. What is selected by adults, that is, what works, depends on the child's level of proficiency:

Toddler: "Mih"
Mother: "Do you want mi LK?"
Toddler: "Mi k"

For this toddler, the mother considers "mik" good enough to provide the milk.

Later, of course, "mik," no longer works. Adults require the pronunciation "milk." No child starts speaking like a college student. Gradually, over many years, sounds that increasingly approximate adult speech are selected from the many utterances a child makes.

Skinner coined the term "shaping" to describe how totally new behavior, like talking, comes about. First of all, learner behavior is paramount. From existing actions, only the "best" are selected by effects Skinner called "reinforcement." Think of learning tennis or Chinese. How proficient would you become by watching Wimbledon tennis matches or listening to recorded Chinese conversations, even with slow motion, a lot of explanation, and arrows to let you go on or to repeat sections? Teaching requires more than presentation or demonstration, followed by a test. Anything people need to learn, from accounting to zoology, is best learned through shaping, starting with the existing proficiencies of the learner.

Programmed Instruction and Teaching Machines

In the 1960s, Skinner designed what he called "programmed instruction." He and a colleague, Jim Holland, converted the text portion of his undergraduate course at Harvard into 1,901 steps he called "frames." Frames were organized into units. Students went to a learning center to take their units.

B. F. Skinner's Teaching Machine
Inside B.F. Skinner's Teaching Machine (left): "Frames" to answer and correct answers for comparison were printed on circular disks. Students wrote their answers on the paper strip. Students scored their own answers by moving the lever at the front of the machine one way if correct, another if incorrect.

Because there were no microcomputers in the 1960s, students worked through their units on mechanical "teaching machines." Frames and their answers were written on pie-shaped sections of a paper disk placed inside the machine. For each frame, students wrote missing words or answered questions designed to show understanding of critical points. The items built to increasingly complex skills. Analysis of student data showed where instruction was weak and sequences were revised until students were almost always successful at each step. In 1961 the program was published as The Analysis of Behavior.

Fast forward to 2010 and the internet. Online instruction has opportunities that Skinner could only dream about. Lessons can base a learner's next task on what he or she just did. Did the student misspell a word? A lesson can show the correct spelling, but also credit the answer as correct—at least the first time. Does a student answer several items incorrectly? Transfer him seamlessly to a sequence on prerequisites—without the stigma of "remediation." Did a student answer very rapidly and correctly? Skip him to a more advanced level. Did a student take a long time to answer but answer correctly? Timed practice would help. Does a student need to apply principles rather than just talk about them? Show videos and ask about what is going on. An ideal e-learning program would adapt on the spot to each individual's moment-to-moment performance.

These capabilities do exist on the internet, but not where needed for teaching academic material. Most computer games adjust dynamically to user input and do so instantly, with the behavioral results that science would predict. When responding is rapid and success frequent though not inevitable, even students diagnosed with attention deficit hyperactivity disorder, (ADHD) can play games for long periods of time. Programs to give practice on skills already learned also incorporate behavioral features. Those that require learners to type or select answers usually score each answer as it is entered. The better fluency-building programs also tell learners how fast they performed in addition to giving their percent correct. All fluency-building lessons, however, assume that students already can perform the skills on which additional practice is provided.

Most e-learning units that teach new skills follow the lecture model: Present quite a lot of information or video and then give a quiz or test. The internet improves on in-class lectures in that students take lessons on their own time, and quizzes are graded right away. But few online programs require the density of active participation needed for efficient learning. Back in the 1960s and 70s a high rate of student responding was achieved by including at least one student response for every 25 words presented. Just for fun, I checked The Analysis of Behavior to see how many words Harvard students read for each term they wrote. The book fell open to page 90. That page has an average of one student response for every 24.4 words of reading. With programmed instruction, students must constantly pay attention.

The effectiveness and efficiency of any lesson improves when the amount of presented material is cut, thus reducing the time between each learner response. Shaping requires more activity by the learner and less by the teacher. The teacher still must select appropriate actions and adjust the next step according to student progress. Live tutors in schools who follow these shaping principles enable students to make impressive gains. The science Skinner began continues to spread in one-to-one instruction.

Online, e-learning lessons designed by behavior analysts include responding by students, but rarely the high density of student actions of one-to-one tutoring or programmed instruction. That is a problem easily fixed. Increasing the density of student responding can easily be accomplished by e-learning designers.

More challenging than adding opportunities to respond is adjusting instructional sequences according to what students do moment to moment. Dynamic interaction turns pedantic instruction into stimulating and effective lessons. On the internet, techniques for interplay between user and "content" can be designed into application "engines" that take data and put it into interactive form. For example, from the spreadsheet below, an e-learn lesson can present Item 1 and evaluate what the student types by comparing it with all options in the Correct Answers column.

Item Number Item Correct Answers
1 A doctor taps your knee (patellar tendon) with a rubber hammer to test your ________. reflexes, reflex
2 If your reflexes are normal, your leg ______ to the tap on the knee with a slight kick (the so-called knee jerk) responds, reacts
The first two frames from Holland and Skinner, The Analysis of Behavior (1961). Used by permission of the B. F. Skinner Foundation.

Certain items can be flagged as possible branching points. If a student answers a flagged item incorrectly or takes too long, the "engine" can insert a remedial sequence. Students answering correctly without hesitation go on. Once procedures for an instructional program are completed they will deploy content from any spreadsheet that meets the required format. New courses can thus be designed solely by putting new content into spreadsheets.

Learning From Students' Learning

Instructional programs should record what students type in order to provide data for improvement. No sequence of items perfectly achieves its objectives. Every offering of a version of instruction should be considered a successive approximation towards a better unit. Too rapid responding with no errors might indicate a need for fewer steps or increased complexity. Errors on specific items reflect anything from a poorly designed item or a wrong comparison in the "Correct answers" column, to a weak sequence leading up to the missed item.

At this point more advanced aspects of behaviorogical science come into play, such as an analysis of what features on each screen determine how students respond. But that is another topic. If interested, see Behavior Analysis for Effective Teaching or go to When instructional designers combine the delivery technologies of the internet with existing behavioral technologies, they will harness the power shown by computer games. Students taking lessons on cell-phones, tablets, or computers will enthusiastically respond at each moment to exactly the instructional step they most need.


  • Mon, 04 Oct 2010
    Post by Ryan Tracey

    Behaviorism? Isn't that, like, so old? :oP

    Thank you Julie for re-deploying some classical knowledge and making it relevant in the modern world.

    I'm glad to see not everyone throws the baby out with the bath water!

  • Thu, 30 Sep 2010
    Post by Dr. Guy Bruce

    I was so pleased to read this article on how to design more efficient learning programs. The author clearly makes the point that internet technology, while offering the opportunity for more efficient learning,requires a "Technology of Teaching" (the product of a scientific and engineering approach to human behavior)for that opportunity to be realized!

  • Wed, 29 Sep 2010
    Post by Jason West

    Thanks for this article. I think you are spot on about instructional procedures online lagging behind the technology. Key to making the tools we have work for us more efficently, for example in my field which is improving English skills, is to create processes that mimic and support the natural learning processes of the brain.

    Our recent case studies, published in entirety on our free podcast, show that focused and comprehensible conversation between a learner and an expert (native or fluent speaker) can recreate the natural learning behavior of an infant and its mother (motherese). The difference is that the second language learner usually has a large amount of latent (inactive or non speech-forming) linguistic knowledge of the target language as a result of previous years of conventional study or exposure (as well as some acquired psychological pathologies about speaking in public or to a native or fluent speaker). This makes periods of structured and supportive one to one conversation with fluent and native speakers highly effective learning experiences.

    Our results (achieved with some PDF lesson plans, MP3s, Facebook (to meet friendly English speakers online) and Skype to talk to them and record your conversations for re-listening) are stunning and have drawn a comment on our podcast from Professor Stephen Krashen who called them 'Remarkable' and 'a major contribution to what we know works'.

    Our offer of help was randomly selected (she applied to a forum post)by an adult Chinese English learner but to get our assitance she had to agree to do what we asked her to do and agree to having it all recorded.

    The student, 27, had been stuying English formally for 16 years but still spoke like a beginner (clip 1.). After 18 hours of using our materials (13-15 hours) and 3-4 hours of online speaking practice based upon those materials she sounded completely different and could converse comfortably at what most English teachers would call 'intermediate' level (clip 2.).

    The process, i.e. the behaviour of the learner is crucial and that can be guided by materials that work with widely and freely available technologies.

    If anyone is interested to learn more you should look for us online as English Out There. I will be happy to answer questions.

    So behavioural processes and instant adaptation to the comprehension level of the student seem to be crucial.

  • Wed, 02 Jun 2010
    Post by Clark Quinn

    Anna, you point out a couple of the barriers people encounter: time pressures, mistaken beliefs about what adds value. Yet people who take time for reflection are more productive than those who don't, and content wiithout meaningful practice is worthless. We just have to keep fighting the good fight! Thanks for the feedback.

  • Tue, 01 Jun 2010
    Post by Anna Sabramowicz

    Hello Dr. Quinn! I agree with the fact that we can go minimalist and meaningful and it would get us a lot farther than we are with a lot of courses today. It is difficult to change your habits and when under pressure most of us fall into the trap of doing what we are comfortable with, even though in the back of our minds we know it is poor design. The worst part is that you almost never have time to innovate and create those new checklists, you are just keeping your head above the water with the workload. I think a lot of clients these days still feel their courses need a lot of "pulp" to be meaningful and are totally uneducated about the immense benefits of practice and feedback. The funny thing is that you don't have to be so creative anymore, there are a lot of people out there that have figured it out, have designed some fantastic activities and are sharing them. You, as a designer, instructional or otherwise, just have to provide the opportunity for those pieces to have a chance to "live" in your course. Thank you for the great read and the breakdown... that plain table you created is proof that simple visualization is key to helping learners digest information.

    Take care, Anna

  • Thu, 13 May 2010
    Post by Clark Quinn

    Thanks for the feedback. I'm not talking about designing more at the expense of development. And I *am* saying that the creativity doesn't take more time after it becomes habit. I believe (and my personal experience has been) that once you get your mind around the content (which you have to do in either case), the practice of thinking towards meaningful application and model-based exposition, etc ends up being no more time or work than thinking about rote recitation and knowledge presentation, etc.

    I would agree that the nuances between well-designed and typical elearning are subtle: you have the same elements, it's just how they're put together, and the client may not know or care (I recall a situation where the client's boss complained until I walked him through the underlying rationale, and then he was singing the hallelujah chorus).

    Visual design is a part of it, but to me it's more core about designing the cognitive and emotional experience. I still think that, with practice, that becomes as easy as regular design, but there is that initial investment. I posit that if you incorporate minimalism, engagement, meaningful practice, and model-based thinking, you have a chance to take no more time to design and develop really effective learning.

  • Wed, 12 May 2010
    Post by Kevin

    I agree with both of you on different levels. Referring to the same Learning Solutions conference, I attended a session by Ellen Wagner and Cammy Bean about the ID process. They presented a concept of the big "D" and the little "d" (not sure if this is a new concept though as I had not heard of it before then). The big "D" refers to the Instructional Design in your improved process of less writing. However, the little "d" requires creativity in the presentation of the big "D" - of for lack of a better phrase: Visual Communication.

    As Simon points out, there is no argument that we 'want' to design better. Better quality; less risk. In your table where you show Practice, the Traditional method is to provide a collection of simple knowledge questions. In the Improved method you suggest providing a few scenarios applying the knowledge to a problem. Less writing, and perhaps less development time, but it does take more creativity. A rare skill in my opinion. Again to Simon's point, clients don't much care about the overall experience as they do the 'deadline.'

    We need to change that! I'm a design guy and whether its instructional or visual, I think a big gap in this area starts at the academic level where ID grad students are being taught everything there is to know about "Instructional" design but zero in "Visual" design. If we can get ID's to think more creatively about instruction, perhaps that creativity will carry over to the visual presentation and how to communicate with less words altogether.

    Good read, Clark!

  • Tue, 11 May 2010
    Post by Simon Egenfeldt-Nielsen

    As far as I can read this basically boil down to what I would label pre-production (that is well-known in most games development although also often suffering cuts), namely to prepare your project. I think it is fairly obvious that this will lead to better quality. However, I think the article misses the problem in reality - we all want to 'design' more. However, usually we are faced with tight deadlines, clients in a hurry and a desire to 'show' something.

    Also, I think that the problem with 'designing' more is that it requires a lot of skill from the developer if you are not to be sucked into the abyss. You can spend endless amounts of time. Especially with clients with little experience and high expectations.

    I think basically 'designing' more will improve quality and reduce risk, but extend deadlines. This will cost at a limited extra development cost. I also think that sometimes clients may not appreciate that they get better quality and lower risk but merely 'focus' on the deadlines being pushed.