There are so many variables to control for when teaching children with autism. Rather than blaming the child when s/he isn't learning, we must always analyze the teaching. There are many data to indicate that certain teaching procedures are most effective in teaching language and other skills to children with autism. I'd like to discuss these procedures in some detail here and to explain why they are more effective than other popular procedures. You can find further discussion of some of these procedures on this site under How to Improve Your Existing Program. Please keep in mind that none of these teaching procedures are specific to teaching VB, although they marry very well with doing so.

Errorless Learning: There is an entire section of this website devoted to the use of errorless learning, especially as it relates to no-no-prompting. In addition to that explanation, I would like to say that errorless learning is one of the foundational teaching procedures of ABA/VB. Please see the discussion on this site for more detail.

Mixing/Varying Targets and Tasks: Mass trialing (presenting 10, 20, or some other number of consecutive trials on the same target) is not often used in ABA/VB. The preferred procedure is to mix and vary different targets and tasks. This means that we will switch among mands, tacts, intraverbals, receptive ID, RFFC (receptive by feature, function, class), TFFC (tact by FFC), motor imitation, echoics, etc. This makes "autipilot" (my affectionate term for the tendency of kids with autism to respond rotely once they've "figured out" what you want them to do) impossible, since the child cannot simply respond identically 10, 20, or however many times in a row without actually attending, discriminating, and learning.

Transfer Trials: One of the most common complaints about errorless learning (EL) is that it "makes children prompt dependent." This can be true if the teacher doesn't properly fade the prompts. Transfer trials, in which prompts are immediately faded to allow for independent responses, are critical to success with EL. Rather than following a set prompt level to criterion, it is preferable to use most-to-least prompting and adjust your prompting moment-to-moment according to the child's responses. A good rule of thumb to follow is that for every prompted trial you run, immediately run an unprompted, or transfer, trial. This procedure looks like this:

Teacher: "What is it? Cookie. (echoic prompt)"
Child: "Cookie."
Teacher: "Right. What is it? (no echoic prompt)"
Child: "Cookie." (Teacher immediately reinforces.)

You can see that the initial prompted trial was not reinforced, while the unprompted/transfer trial was. This differentially reinforces independent responding, which increases the probability that the child will respond before or without the prompt. This also prevents prompt dependency by making the prompt (which delays reinforcement) slightly aversive.

Fluency Teaching: Children need to be able to respond not only accurately, but quickly as well. Very few behaviors are truly functional when they occur with a long delay after the antecedent. Imagine asking someone you're meeting for the first time for their name and having them reply 10 seconds later. To be truly "mastered" and functional, behaviors need to have both accuracy and speed. This is teaching to fluency and is crucial for skills to be successfully acquired, retained, and generalized. To achieve fluency in teaching, the teacher needs to control for two variables: Latency and intertrial intervals. Latency is the time between the end of the antecedent and the beginning of the child's response. There should rarely, if ever, be a latency of longer than 2 or 3 seconds. The way to control for latency is with errorless learning. If the child does not begin to respond within those 2-3 seconds, the teacher should prompt, reinforce, and run the transfer trial. By prompting when necessary, you can keep latencies short and reinforce faster, independent responses. In addition to short latencies, you also want short intertrial intervals (ITIs). ITIs are the time between the end of the consequence for one trial and the beginning of the next antecedent. This can also be referred to as rate of teaching, since this variable is purely related to the teacher. Again, ITIs of no more than 2-3 seconds are preferable. Short ITIs allow for more teaching to be done in less time and keeps motivation to work high and motivation to escape or stim low.

Interspersal of Easy and Difficult Targets: Since the child always has three options for contacting reinforcement (escape, stim, or working with you) it is important to keep the value of working higher than the other two options. Another way to do this (besides EL and teaching to fluency) is to run about 80% easy targets and 20% difficult. This ratio may change slightly depending on the child, but the general idea is the same. By presenting many easy tasks and fewer hard ones (which are, in turn, made easier through EL) the child is reinforced more often, which results in both increased learning and elimination of the need to escape or stim.

There is much more to all of these teaching procedures, but this brief explanation will hopefully hint at some of the most effective ways to teach children with autism.