Posts Tagged ‘Errorless learning’

The Errors we See- Discrimination challenges in children with autism and Learning Disabilities – Part II

The author is a Board Certified Behavior Analyst working with Behavior Momentum India. The views in this blog are his own. I  hope to keep in mind parents and therapists as I write and will try to keep the jargon to a minimum. Do feel free to contact me via

In Part 1 of the series we saw the types of discrimination challenges and issues we run into when teaching children with autism or other learning disabilities.  These are often a part of larger error patterns that children acquire inadvertently because of the way materials are used, the way responses may be prompted or the way they are reinforced. In this second part, we will look at inspecting and understanding error patterns. In part III we will look at strategies to avoid and overcome them.

Most of the receptive language tasks involve :

  • the teacher presenting an array of 3 or more items, asking for one of them
  • prompting and fading prompts if the child does not know the correct response
  • and rewarding ( reinforcing)  correct responses and ignoring incorrect responses.

Several potential errors creep in depending on how the training procedures are run.  Step one, in most cases, we must recognize that the student may not really interested in acquiring the discrimination between purple and Red or Tall and short. We believe, these are useful and functional and set up the teaching environment while the student could only aim to get a reward with minimum effort. The reward can be a preferred item or activity given by the therapist for correct responses or just completing the task as fast as possible and getting away from it all ( Technically. these are called reinforcers that increase future occurrence of the behavior that they follow) !! Behavior Analysts, both researchers, and practitioners have, in studies since the 1960s documented the kind of errors that could happen and newer and newer strategies are still being researched for improving outcomes. Laura & Grow ( 2013), have published a great article which details the type of errors that occur in teaching receptive language and strategies to overcome them.

Most parents and therapists miss the first step- recognizing that there is a problem. After doing a few tests it is very easy to be deluded into believing that the child ‘knows’ what is being taught though he sometimes makes incorrect responses. Some of the error patterns and how they creep in are discussed below:

I bet on a position ( Dominant position bias):

When you place pictures or dolls of cat, dog, and cow in an array and ask for one, say ‘cow’,  the student could adopt several strategies to complete the task which does not require him/her to listen carefully and make a correct selection – One is a position bias. As a student, I go for one particular position more often,  for example, the right most. Soon the therapist/mom inadvertently starts placing the target item there more often in that position, for, after all, the therapist or mom is also rewarded by students correct responses. The mistakes that happen when the target item is in other positions are overlooked and attributed to the student being ‘inconsistent’. In such a strategy, the student can completely avoid paying attention to what the instructor/ therapist/parent is asking to be selected( Technically called the Auditory stimulus presented by an adult which should serve as a discriminative stimulus for the student response).

I bet on a particular target item ( Dominant Response Bias):

Using the same example, possibly because I was taught ‘cow’ first and reinforced several times, even when the new targets of dog and cow are introduced I keep selecting cow more often regardless of what I am asked to select. On the occasions when I am correct, I get rewarded. Sometimes, this too could result in the teacher/ therapist asking for ‘cow’ more often creating an illusion of successful discrimination!

I bet on a familiar looking item :

There was a student who could identify spoon, cup, pen, socks, fork, lunch box, toy, and nearly a dozen such items from an array of 5 different objects. He never made a mistake until his teacher started presenting the array with 2 or more from his ‘known’ list. Then he started making mistakes. In the earlier procedure, the student merely had a rule that is something like ” If I see that familiar one, I will go for it”.

I respond based on body language or other inadvertent cues from the teacher:

While teaching receptive language the teacher may be inadvertently supplying additional cues which indicate where the correct responses lie. The student can be sometimes seen intently looking at the teachers face before selecting the item asked for. Some inadvertent prompts could be

  • the therapist’s eyes  dart at the target item before she asks for it (called eye-prompts or eye-gaze prompts)
  • the target item could be placed closer to the child than the other items, as the student’s hand hovers over an incorrect response the therapist’s features are taut and they relax as he switches to the correct response midway
  • the teacher uses exaggerated intonation correlated to a specific target ( Biiiiiig vs small, nineteen versus nine) but does not fade such intonation prompts.

Switching bias :

The student, after being asked to select big, switches the next response to small ( even if ‘big’ is asked for again) because historically his therapist has always switched the target from the previous teaching trial.

Random responding:

Here the student randomly selects any item without regard to what was asked. Afterall,  in an array of 3, he could be successful 33% of the times and that is fine!

Stimulus Overselectivity:

Stimulus over-selectivity or attending to one part of an item to the exclusion of all other parts of the whole could result in errors. For example, a child who looks only at wheels may find discrimination between a car, a truck and a van difficult.

Generalization Problems:

While discrimination is acquired with specific items taught it has not generalized to novel items 0r contexts.

Laura and Grow in their 2013 article talk about several evidence based strategies to avoid errors and teach receptive identification to children with autism.   These include counterbalanced assessments and presentations, requiring an observing response, avoidance of instructor errors, use of simultaneous targeting versus introducing targets one by one and use of appropriate array sizes.  In addition, Fisher, Pawich, Dickes, Paden and Toussaint (2014) conducted a very important study that explains how and why it is important to manage the consequences of correct responses in relation to errors ( they call it managing the behavior-consequence salience for children with autism who present persistent errors).

We will discuss these strategies to over come error patterns in part II of this blog series.


Selected References

Grow, L., & LeBlanc, L. (2013). Teaching Receptive Language Skills: Recommendations for Instructors. Behavior Analysis in Practice6(1), 56–75.

Fisher, W. W., Pawich, T. L., Dickes, N., Paden, A. R. and Toussaint, K. (2014), Increasing the saliency of behavior-consequence relations for children with autism who exhibit persistent errors. Journal of Applied Behavior Analysis, 47: 738–748. doi:10.1002/jaba.172

Errorless learning is the way- time to think outside trial and error

Trial and error learning is often counterproductive and too often we cannot afford the cost of errors especially when teaching children with learning disabilities. The proponents of trial and error learning will show you the example of throwing a baby into water without any training and behold! it swims!!!!. This is not an entirely  accurate proposition as babies may have certain reflex actions in their repertoire that may look like they are born with an ability to swim but it  is not true. such an attempt would be downright dangerous and irresponsible. Check out

We see examples of the danger and disasters that trial and error learning cause all around us. Recall the girl who, while learning to reverse crashed her car into a pillar in the basement and could never learn to drive after that. Do you not come across hundreds of children who ” hate math” ? Well they tried to learn by trial and error  and came across failures that stunt learning. If this is the case with typically developing chidren ( and adults) the problem with errors is much higher in children with learning disabilities trying to learn a new skill. If you have tried to teach them or have observed them being taught and you find that they engage in escape behaviors, fright reactions , aggression or behaviors suggestive of anxiety it is very probable that error rates are high. The necessary pre-requisites for successful performance may be missing or some irrelevant aspects of teaching procedures could be dictating the child’s response.These are also true about typical population where students fail in exams or people fail in their jobs without adequate training.

Dr. Murray Sidman, a pioneering behavioral scientist in  a 2010 publication argues that the typical learning curves which show an acceleration towards a peak performance (during which period learning is progressively happening by trial and error) followed by a plateau appears like pat explanation only because  such curves are developed by aggregating data of the learning process of hundreds of individuals. If an individual’s learning is tracked there will be seen a lot discontinuities and irregularities. He then presents an alternative learning path which is that when all pre-requisite skills are taught there is an instantaneous vertical climb (no acceleration or curve) to performance with all learning and no error. He quotes two simple and elegant examples from the non-human kingdom.

B. F. Skinner, taught rats to press a lever in a chamber in a errorless fashion by teaching 5 pre-requisites. He first let them explore a chamber and experience that the environment was safe. By mixing food pellets with the rats regular food, he taught them that pellets were food too. He then dropped the pellets in a food tray occasionally and the clanking noise from the dispenser was a indication of when pellets would be available. soon the rats started going to the dispenser whenever the dispenser made a sound and they picked up their food pellets.  Next a  lever was introduced in the chamber. When the rat pressed the lever the first time, the dispenser operated making a sound, the rat went to the food tray and ate the pellet. There after it continued pressing the lever at a high rate pausing only to eat the food pellets dispensed. The rat acquired the ‘lever pressing skill’ in a immediate fashion ( see figure 2 below). Imagine another rat which was not trained in the pre-requisites being let inside a similar chamber. How long could it be before it starts pressing the lever and accessing food is anybody’s guess. The latter rat may even give up after the fist lever press and stay hungry and anxious because it does not have the pre-requisite training to look into the food tray or even identify the food pellet as a reward.

Terrace in his Doctoral dissertation in 1963 built on the foundation and demonstrated error-less learning with pigeons. The first goal was to teach them to peck a key only when red light was on and not when green light was on in a chamber. He achieved this with a set of pigeons errorlessly by having the red light on in normal intensity and rewarding lever presses. He introduced the green light ( to which the pigeons should not respond) in a gradual fashion from low intensity to eventually an intensity equaling the red light. These pigeons made no error responses – never pecked the key when green light was on and always pecked when red light was on. Compare this with another group of pigeons which were introduced to both the lights at equal intensity from the start and they made a large number of error responses before acquiring the discrimination that it is worth pecking only when red light is on. There were many other brilliant facets to the great experiments done by Terrace. Suffice to note that the pigeons that learnt errorlessly also performed better later much after the training was withdrawn. Their learning endured. Another important note is that errors do not produce rewards and hence can produce fright reactions or aggressive behaviors.

Figure from sidman (2010)

The dad who successfully manages to teach his young son to cycle was not only patient but also manipulating the learning in such a way as to avoid nasty falls or unsuccessful turns. Every successful coach works hard on pre-requisites training and minimizing errors though he may seen to be commandeering. You would have doubtless come across the story of someone who succeeded in the eighteenth attempt. If you are able to examine closely you will find that the previous 17 attempts involved gaining some new ground everytime , identifying missing pre-requisites and working on them diligently and the 18th is a success. Otherwise if the previous attempts were a utter failure with no rewards contacted or no new pre-requisites identified it would have stopped at 2 or 3?

Behavior Analysts work towards promoting error less learning by a variety  of research based procedures such as shaping, prompts combined with effective and timely prompt fading procedures, evaluating the use of time delay before providing prompts, adding cues that can be faded to stimulus materials before fading and arranging rewards in proportion to task difficulty. There is now a large body of research that confirms the efficacy of these procedures not only in effectively facilitating skill acquisition but also in minimizing  escape responses and aggression during the teaching process. Pairing of the instructor with positive experiences and rewards also plays a key role. If  errors still occur then a Behavior Analyst would go deeper into component skills and focus training on those.

Play, A., Activities, G., & Team, T. (2017). Is it true that babies are born with the ability to swim? | BabyCenter.            BabyCenter. Retrieved 3 February 2017, from

Sidman, M. (2010). Errorless learning and programmed instruction: The myth of the learning curve. European Journal of Behavior Analysis, 11(2), 167-180.

Terrace, H. S. (1963). Errorless transfer of a discrimination across two continua. Journal of the experimental analysis of behavior, 6(2), 223-232.



%d bloggers like this: