There is much discussion of what prompting strategies are most effective for children with autism. The two most-discussed prompting procedures are errorless learning and no-no-prompting. This is my
take on errorless learning (EL), which will make the most sense if it's
discussed in the context of no-no-prompting (NNP).
Prompting procedures differ across programs; while some DTT programs use NNP, others use EL; many use EL for acquisition skills and NNP for mastered skills; others use some combination of the two. I don't know of any ABA/VB programs that use NNP, though they may be out there; rarely is anything but EL used in such programs, if ever. However, DTT and NNP are not synonymous, which many folks falsely believe. I'll add that some folks adapt the NNP or errorless procedures
slightly, so this may not be what some versions look like, but this is the
gist of both. Determination as to the efficacy of any prompting procedure must be done on a case-by-case basis. And, again, this is my opinion, for folks to take or leave, although I've included some citations that support these statements (and from which much of this discussion stems). If anyone using NNP would like to email me with a contrasting opinion (preferably laid out similarly to this one) I'd be happy to discuss posting it here as another viewpoint.
NNP is intended, and used by those who do it correctly, to work on skills that are considered mastered, or (as some have said) for skills that have been in acquisition for a while, i.e., in random rotation. Unfortunately, this is not always done correctly; some begin teaching acquisition skills with NNP, which is not supported by anyone as an effective way to teach.
This also begs the question of what is considered mastery, which is discussed elsewhere on this site. I'm not convinced that NNP is effective at any stage in teaching, for the reasons described below, but some others feel differently.
That said, with NNP, the teacher delivers an antecedent (i.e., "Point to the
dog") and waits for the child to respond. Assuming the child in this example
responds incorrectly, the teacher's consequence is to say "No," often also
turning her head to the side. The antecedent is then delivered again and,
assuming the child again responds incorrectly, the teacher again consequates
with "No." The antecedent is then delivered again and the child is prompted
to give the correct response, which is then praised. It's important to
point out that most folks doing NNP use what's termed an "informational no," or "no equivalents" (such as "nope," "try again," etc.)
meaning that it's not harsh, just neutral.
With EL, in contrast, the teacher delivers an antecedent and either prompts immediately (zero-second time delay) or waits a
beat to see what the child will do. If the child begins to move to an
incorrect response, or doesn't respond, the teacher immediately prompts the
correct response and praises. The same antecedent is then presented again,
this time as a transfer trial (meaning it's an attempt to have the child
respond correctly without the prompt or with less of a prompt). If the child is correct
independently, the teacher reinforces more strongly (differentially
reinforcing the independent response) and moves on. If the child again
begins to respond incorrectly, the teacher prompts again, then usually moves
on for a few trials to other targets. However, she soon returns to the
missed target to try for an independent response, again prompting and trying
for a transfer trial as necessary.
NNP is a system that uses least to most prompting, which (as the name
implies) involves starting with less prompting and gradually increasing that
prompt in response to errors. EL is a system of most to least prompting,
which initially involves prompting with a 0 second time delay (meaning
immediately) and gradually fading the prompts to foster independence. There are many errorless learning strategies, i.e., progressive time delay, intrastimulus prompting, etc., but the goal (to minimize or eliminate errors) is the same regardless of technique. While
NNP may seem like it fosters independence, and may in some cases do so, in my opinion it can also teach a chain
of errors, thin the reinforcement schedule such that learning can become
aversive, slow the rate of teaching, make kids prompt dependent, and misuse
the child's learning time. EL done correctly can and likely will prevent these teaching
problems. Let me discuss why:
NNP May Teach a Chain of Errors: You get what you reinforce. That means that
if you allow two (or any number of) errors and then prompt and reinforce a
correct response, you may have chained those 3 behaviors together (wrong, wrong,
right). For example, I had a student whose teacher had been using NNP with
him. He'd been taught to answer the question, "What's your name?" this way.
The teacher explained to me that he typically would answer with "Buh," which
she'd consequate with "No"; this would happen a second time, after which
she'd prompt, "Billy." This explained to me why when I'd asked him his
name, he'd replied, "Buh-Buh-Billy." I thought he had a slight stutter, but it was
just the way he'd been taught.
EL aims to keep the rate of errors to a zero- or near-zero level, prompting (hopefully) before they happen so that
the child doesn't spend his time practicing errors. Why is it useful to have a child be wrong twice before he's taught to be right? Were any of us
taught that way in school? If we were, did we learn anything from it? The purpose
of teaching is to show kids how to be correct so that they learn, not to
prove that they don't know something by letting them be wrong. That may not
be the aim of NNP, but it's sometimes the result of it. The logic of NNP really
escapes me and I'm still waiting for someone to provide a great explanation
of why it's done.
NNP May Thin the Reinforcement Schedule, Making Learning Aversive: We all want
kids to find learning itself reinforcing. I achieve this through an ABA/VB/NET
errorless learning approach by pairing myself with reinforcement, keeping
demands low and reinforcement high, and having a roughly 80% easy to 20%
difficult ratio of targets (and I make the difficult tasks easy through EL).
Kids (like all of us) are reinforcement-seeking missiles; they will go
wherever there is the highest level and best quality of reinforcement. At
any given moment, as Vince Carbone describes so well, there are three ways that a
child can get reinforcement: 1) Escape (socially-mediated negative reinforcement), which involves almost any behavior
that results in not working, such as aggressive behavior, disruptive
behavior, literal escape, etc.; 2) Stimming (automatic reinforcement),
which involves getting reinforcement from their own body; or 3) Learning (socially-mediated positive reinforcement),
which is, obviously, working with you. Given this, what the teacher has to
know is that she must constantly be at least as reinforcing as, and
preferably more reinforcing than, escape or stim for the child. (This is
called the matching law.)
Okay, so given all of that, let's compare quality and rate of
reinforcement in NNP vs. EL. Let's say the child is presented with 300
trials during a session. If the teacher is doing NNP, each antecedent might
be presented 3 times before the child is reinforced. If the child is making
a lot of errors, this might result in the child's being reinforced for only
30% of the session. This could mean that the child is only reinforced 100
times in those 300 trials. Look at it the other way: That child spends 70%
of the session being wrong. That low rate of reinforcement usually results
in the child's seeking reinforcement from escape and/or stimming, where
there's a higher rate of reinforcement.
With EL, the child is always right, whether as the result of prompted
or independent responses, so the child is reinforced potentially 100% of the
session, or 300 times in 300 trials. With that high rate of reinforcement,
there's no need for the child to go to escape or stimming for reinforcement;
it's right there in the learning.
Also, even difficult tasks are made easy through EL. No task in
itself is inherently difficult, if you think about it; tasks seem easy or
hard based on our skill set. For example, my husband is very good at math.
He had good teachers who helped him and reinforced him such that he learned
how to do math easily and well. He doesn't view math as a difficult thing.
I, on the other hand, am not good at math (so forgive any errors in this
post!) because I had terrible math teachers who told me to "figure things
out myself," so I didn't get any reinforcement for math and never learned how to
do math well. I, consequently, view math as a difficult thing. It's all in
your perspective. What seems difficult can be made easy through good
teaching, which involves effective prompting. Kids who are taught with
errorless learning don't know the difference between easy and difficult
tasks because the easy ones are those that they can do independently, for which
they receive reinforcement, and the difficult ones are those for which they
require prompting, for which they also receive reinforcement. In that
scenario, what's difficult?
With NNP, by contrast, the difference is usually very clear. When the child
is wrong, they hear, "No" (or something like it), get no reinforcement, and have to try again at
the risk of being wrong again. This is very clear and delineates easy and
difficult tasks sharply. Easy tasks are those that result in reinforcement
and difficult tasks are those that result in "No" and no reinforcement.
Given the percentages I mentioned above, that can mean 70% difficult and 30% easy,
which is far from the 80% easy and 20% difficult of EL.
NNP May Slow the Rate of Teaching: There are reams of data that show that a
high rate of teaching results in better learning for the student, so we aim
to keep that rate high. Rate of teaching is mostly determined by two things:
1) Latency, which is the amount of time between the end of the antecedent
and the start of the child's response. So if I say, "Point to the dog," and 2
seconds later the child raises his hand to point, that's a latency of 2
seconds; and 2) Intertrial intervals (ITI), which is the amount of time
between the end of one consequence and the beginning of the next antecedent.
So if the child correctly IDs the dog, I say, "Great!" and then 2 seconds
later I say, "What's this?", that's an ITI of 2 seconds. Both latency and
ITIs need to be short to get the most teaching done in the least amount of
With NNP, the rate of teaching can be pretty low, due to the no, no, prompt structure. Since the
teacher may have to wait for the child to respond (as we've seen, often
incorrectly) twice before prompting, this can slow things down. Much of this depends on how long a latency is permitted before the teacher consequates.
With EL, usually there's no more than 2 seconds latency permitted
before a prompt is given, and teachers aim to have no longer than 2 second
ITIs, so in an hour-long session, that means that hundreds of targets would be presented. This is another consideration when choosing a prompting procedure.
NNP May Make Kids Prompt Dependent: The argument typically made against
errorless learning is that it makes kids prompt dependent. If EL is done
correctly, this is not at all true. After every prompted trial, the teacher
should go back and try for the transfer trial, in which there's no prompt,
in order to get an independent response. These independent responses are
reinforced much more strongly than the prompted ones, which typically
results in learning, since the child wants the better quality reinforcement that
comes with independent responding. EL will only make kids prompt dependent
if these transfer trials aren't part of the teaching.
NNP, however, can and sometimes does make kids prompt dependent. The
reason for this is that unless the child is correct the first time, he will
often be wrong until he's prompted. Kids quickly learn that they can be wrong
twice and then be prompted, so there's no need for them to respond
independently. This can be combatted with differential reinforcement, but it
often isn't. Think about it: If you were essentially told "No" almost every time you
responded to something, you'd stop responding pretty quickly. I know that
most NNP proponents say that the child learns from his errors, but all he may learn is that he was wrong, not how to be right. This also prevents the student from practicing being correct and from being reinforced for being correct, which is what leads to learning (behavior contacting reinforcement). Punishing incorrect behavior does not necessarily evoke correct responding that can then be reinforced.
NNP May Misuse the Child's Learning Time: I think the above illustrates this
point pretty well, but it's worth saying outright. Good teaching can eliminate
almost all errors so that the child spends the maximum amount of time
learning. NNP may do the opposite, which seems to me to demonstrate that it's
not necessarily effective teaching. When a child is wrong, it's rarely helpful, no matter what
people say. Being right, and more importantly, being reinforced for being
right, is how you learn. Instead of practicing errors, as kids may do with NNP,
they are practicing successes with EL. There's no downside to that.
My last point is that there are few, if any, data to support the use of
no-no-prompting; it's sort of an anomaly. If you want to read up on
errorless learning in the journals, so that you're not restricted to my
information, here are a few great studies explaining these issues:
Heckaman, K., Alber, S., Hooper, S., and Heward, W. (1998) A comparison of
least to most and progressive time delay on the disruptive behavior of
students with autism. Journal of Behavioral Education, 8, 171-202.
Touchette, P.E. & Howard, J. (1984) Errorless learning: Reinforcement
contingencies and stimulus control transfer in delayed prompting. JABA, 17,
Also see: Terrace, 1963; Sidman and Stoddard, 1966; Sailor, Guess,
Rutherford, and Baer (1968); Reese, Howard, and Rosenberger, 1977; Etzel and
LeBlanc, 1979; Altman, Hobbs, Roberts and Haavik, 1980; Carr, Newsom and
Binkoff, 1980; Weeks and Gaylord-Ross, 1981; Touchette and Howard, 1984;
and Durand, 1985; Lancioni and Smeets, 1986; Woolery, Bailey and Sugai,
Durand, 1990; Horner and Day, 1991; Woolery, Ault and Doyle, 1992; Cameron,
Luiselli, McGrath and Carlton, 1992; Cameron, Ainsleigh and Bird, 1992;
Sprague and Horner, 1992; Smith and Iwata, 1997; Woolery, Ault and Doyle,