Discrimination Learning

Discrimination Learning

This chapter on discrimination learning is the most important chapter in this manual. Although every effort is made to present discrimination learning in a detailed and stepwise manner, the procedures involved in discrimination learning can be difficult for the teacher to master. Therefore, be patient with yourself and proceed in small steps. It is helpful to read this chapter more than once and to review it before beginning any new program. The payoff for doing so is that discrimination learning procedures will facilitate the student's acquisition of all the programs presented in this manual. Discrimination learning is a process essential to helping the student acquire complex and flexible behaviours. In describing the various teaching programs in this manual that depend on discrimination learning, we review the basic steps involved in using discrimination learning as they apply to the particular tasks contained within those programs. We are especially conscientious and detailed in our presentation of discrimination learning in those programs that are difficult for the student to master, such as verbal imitation and prepositions. It is, however, to the teacher's advantage to be presented with these steps in a more general and abstract format, as done in this chapter.

Discrimination learning forms a basic teaching process that helps the student attend to (discriminate), and therefore learn about, all the subtle and complex stimuli to which human beings must respond in order to survive and develop. Over time, typical individuals learn to attend to an enormous range of stimuli without anyone having to explicitly and carefully teach them how to attend. Individuals with autism and other developmental delays, however, evidence extreme delays in learning to attend and must be taught to do so. This heightened attention gained through the use of discrimination learning procedures increases the student's rate of progress in the teaching situation.

Attention is considered a higher level mental process by most psychologists and educators, which speaks of its central role in the development of complex behaviours. It is by the use of discrimination learning that students with developmental delays and autism learn to attend to and imitate the behaviours of others, develop language, and acquire the concepts of such complex events as identifying and describing spatial relationships, differentiating between feelings within themselves and others, learning about the causes of different emotions, and much more.

In this chapter, we describe how to optimize discrimination learning by selecting appropriate teaching materials and achieving consistency across teachers. We then describe how to use differential reinforcement to help the student discriminate between different instructions, whether verbal (vocal, auditory) or visual. This is followed by a procedure called random rotation, which describes how to order the presentation of teaching material so as not to mislead the student but rather to help the student focus on the important parts of the teacher's instructions. Finally, we examine some of the problems teachers encounter in conducting discrimination learning and how these problems may be prevented or resolved. First, however, we illustrate how important it is for the teacher to be familiar with discrimination learning.

If you present the student with a particular instruction (vocal or visual) and teach the student to respond to that instruction, you may encounter two problems. First, there is no guarantee the student is attending to the salient or critical elements in your instructions. Although we use vocal instructions to illustrate the procedure, visual instructions (e.g., printed words) could just as well be used. If you reinforce the student for pushing a toy truck when you instruct him to 'Push truck,' and he eventually responds according to your instruction, it may seem as if he has learned to attend to your instruction. However, a simple test of whether the student responds to the critical elements of your instruction ('Push truck') can be made by instructing him to 'Clap hands.' If the student responds by pushing the truck, he may be responding to your instructions as undifferentiated auditory input (as vocal noise). Similarly, you may teach the student the appropriate response to 'Clap hands.' If you then ask the student to 'Push truck,' he may clap his hands. In technical terms, the student is not attending to (discriminating) the critical stimulus elements of the instructions.

The second possibility is that the student may not respond to your vocal instructions (the auditory stimuli), but rather to some visual cue. For example, when instructing, 'Push truck,' a teacher may inadvertently look in the direction of the truck and this visual cue may lead the student to seem to respond correctly. Similarly, if a teacher instructs the student to 'Clap hands' while looking at the student's hands, then this visual cue may guide his response. A test of whether the student responds to the auditory stimuli may be performed through the elimination of visual cues by fixating your gaze at the student's forehead while presenting verbal instructions. If the student is attending to the visual rather than the auditory cue, it is most likely that the student will either not respond or alternate between responses while looking at the teacher's face, searching for a significant visual cue. Some students learn to read lips, and this illustrates a similar problem. If the teacher's lips are covered while the instructions are provided, the student may fail to respond correctly, if at all.

Many persons with developmental delays seem to attend better to instructions that possess visual rather than auditory stimuli. There could be many reasons for this. One possibility is that auditory stimuli are of such short duration that they terminate the moment the teacher finishes the instructions. In contrast, visual stimuli, such as a teacher's gaze and pictures or printed material, last longer. Perhaps because they last longer, the student is given more of an opportunity to attend, and this may help him associate visual instructional input with behaviour. We introduced a similar possibility in Chapter 1, where we considered the student's echolalia to be a form of rehearsing instructions and prolonging the duration of auditory stimuli, thereby facilitating the student's acquisition of verbally transmitted information.

It is important to remember that it is easy for students to learn things that teachers do not intend for them to learn. Whenever a teacher delivers an effective reinforcer contingent on some behavior, the student learns. A skilful teacher is aware of potential teaching mistakes, how to avoid them, and how to teach the student in an effective and productive manner (see Chapter 35). Every teacher can learn to become skilful.

Instructions

We illustrate the basic steps of discrimination learning by using one of the first programs the student will learn: receptive language. Specifically, we use as an example the discrimination between the receptive instructions 'Clap hands' and 'Touch table.' Note that these are only examples, and that these examples help illustrate not only programs that teach receptive language but also nearly all programs in which a teacher presents stimuli, be these auditory (as in the Verbal Imitation Program) or visual (as in the Nonverbal Imitation, Expressive Language, and Reading and Writing Programs).

The great majority of students need to be exposed to discrimination learning to master correct responses to even simple instructions, and, in our experience, all students with developmental delays have to undergo discrimination learning when exposed to the more advanced programs in this manual. It is important to abide by the following guidelines before beginning to teach discrimination learning:

Guideline 1

Start with two instructions that are maximally different. For example, 'Clap hands' and 'Touch table' appear to be more separate and different sounding than instructions such as 'Push truck' and 'Push car,' or 'Touch her nose' and 'Touch my nose.' The instructions 'Clap hands' and 'Touch table' differ in two important ways. First, the two instructions require actions that are different from one another (i.e., one involves an object whereas the other involves the body only). Second, the instructions sound different from each other. To further increase the dissimilarity of the instructions, you may abbreviate 'Touch table' to 'Table' and contrast this with 'Clap hands,' which renders the instructions different in length as well. If length is a prompt, make sure to fade it. During the early steps, use only the part of the instruction that is salient and critical for the student's mastery; avoid extra elements that are superfluous and may interfere with the student's solution. 'Please be so kind as to touch the table for me' is an example of a sentence with considerable redundancy that interferes with the student's learning to attend to the relevant stimulus element, namely 'table.' Although it is important to eventually generalize and teach subtle differences among instructions, it pays to start with the simple instructions and help the student learn to attend to gradually more subtle variations.

Discrimination Learning  125

• Guideline 2

Initially simplify the setting in which teaching occurs rather than trying to use naturally occurring situations such as a classroom environment. The more control you exercise over the stimulus presentation and the simpler this environment is, the less likely it is that distracting stimuli will interfere with learning and the more likely it is that the student will learn to respond to your instructions.

• Guideline 3

Use instructions that involve different and distinct behavioural topographies when appropriate. For example, touching a table requires a very different response from clapping hands, in contrast to similar behaviours such as pushing a toy truck and pushing a toy car. Pointing to one's eye and pointing to one's nose is another example of a difficult discrimination because both the instructions and the two responses are similar, and the student receives little differential feedback (in this case visual) from her responses. The response to the instruction 'Look at me' is more difficult to effectively reinforce than previously considered because the sensory feedback from the response is subtle and difficult for the student to discriminate (i.e., it is difficult for the student to connect the response to the reinforcing event). Vocalizations such as babbling may also be difficult to strengthen by reinforcement for the same reason. In short, start with maximally different stimuli and use responses with distinct proprioceptive feedback, which are easier to bring under reinforcement control.

• Guideline 4

Specify in detail what constitutes the correct response before beginning to teach, and decide exactly what your instruction is for that particular response. For example, if you decide to reinforce the student for clapping her hands, do not arbitrarily change the response requirement from one or two to several claps. If you are teaching the student to touch a table, it is enough for the student to touch the table with any part of her hand for 2 to 3 seconds. If that response is reinforced, do not arbitrarily alter the response requirement by, for example, teaching the student to touch the table with her finger, which will most likely interfere with mastery of either task. As another example, suppose you decide to teach the student to clap her hands, and the student stands up in response to your instruction. If you then instruct, 'Sit down and clap hands,' you may be unwittingly attempting to teach two tasks at the same time. It is better to separate the tasks, teaching one at a time, to simplify the learning situation. Later in the program, the student will need to learn more than one response at the same time and how to deal with inconsistencies, ambiguities, and differences in teaching styles. However, such skills should not be taught during the beginning stages.

• Guideline 5

Make certain that all members of your team understand the requirements for instructions and responses. If team members are not consistent or if a teacher shows hesitancy about what and when to reinforce, then correct these mistakes before going any further. Otherwise, the student is likely to become confused, and this will delay her mastery. Remember that it is important to have other team members present their teaching skills on a regular basis, as done in weekly meetings and visits to the house by senior team members, so that feedback and advice on how to proceed can be given. It is difficult or impossible for any one person to know how to proceed, and it is easy for any one team member to drift off criterion. We will provide numerous examples of potential teaching mistakes later in this chapter, in Chapter 35, and within each program, and suggest ways by which these mistakes can be remedied or avoided.

The first step in discrimination learning is teaching each instruction separately until the student masters the response to the instruction. This is termed mass trials and is done by the teacher helping the student achieve mastery through the presentation of several trials of a particular task in a consecutive and repetitive manner. After two separate tasks are mastered in this manner (e.g., 'Clap hands' and 'Touch table'), the teacher intermixes and alternates between these two tasks so that they are contrasted with each other, thereby helping the student tell them apart. This process of intermixing is likely to lead to a temporary loss of what the student has already mastered because the student cannot discriminate between the two instructions when they are first intermixed. This problem is solved by introducing the student to differential reinforcement and, later, random rotation. These procedures are described in detail in the next several sections.

Differential Reinforcement

One important teaching technique that was introduced in Chapter 10 concerns the use of differential reinforcement. Differential reinforcement helps direct the student's attention to the relevant aspects of the teacher's instructions. By definition, differential reinforcement consists of a procedure in which the teacher differentiates between the student's correct and incorrect responses by reinforcing and thereby strengthening correct responses while not reinforcing and thereby weakening incorrect responses. It is by the use of differential reinforcement that the student learns to tell the difference (discriminate) between the teacherss instructions.

It may be helpful to illustrate the process of differential reinforcement in the form of a diagram. Using the two instructions 'Clap hands' and 'Touch table,' let SD l (Discriminative Stimulus 1) represent the instruction 'Clap hands' and SD2 (Discriminative Stimulus 2) represent the instruction 'Touch table.' Let R l (Response 1) represent the student's clapping his hands and R2 (Response 2) represent the student's touching the table. If these two tasks have been taught separately, it is highly likely that the student will be unable to tell the instructions apart and will sometimes respond to SD l ('Clap hands') with R2 (touching table) and make similar errors (e.g., clapping hands) when SD2 ('Touch table') is presented. In Figure 16.1, let the arrow represent the bond or association between the SD (the teacher's instruction) and the R (the student's response).

If the association between SD l and R2 is as strong as the association between SD l and R l, the student will respond to SD l with R2 as often as he does with R l. This is a common occurrence prior to the use of differential reinforcement. However, if the teacher continues to reinforce the SD1-R1 relationship ('Clap hands' and the student claps his hands) and withholds reinforcement when SD1-R2 occurs, it is likely that the SD1-R1 association will be strengthened and the SD1-R2 association will be weakened. In technical terms, the SD1-R1 relationship is placed on acquisition while the SD1-R2 relationship is placed on extinction. With such a procedure, the SD1-R1 association eventually grows strong enough to block out errors.

In addition to withholding the reinforcer, the teacher may give an informational 'No' upon the student's incorrect response. For some students, however, the word no, may be neutral (i.e., have no effect) or be a positive rein-forcer (e.g., the student seems to enjoy the teacher's informational 'No'). Regardless, by combining (associating) the informational 'No' with the withholding (withdrawal) of a positive reinforcer (e.g., food), it is likely that the informational 'No' will eventually become an aversive stimulus, which will help to inhibit the incorrect response.

Once the SD1-R1 association becomes strong, the SD2-R2 association is practiced in the same manner as SD1-R1. The discrimination between SD l and SD2 is accomplished by intermixing and contrasting the two associations as described in the later section 'Intermixing Stimuli.' Once the student gives the correct responses to SD l and SD2 when they are randomly intermixed, it can be said that the student has learned to attend to the two instructions through the teacher's use of differential reinforcement.

Consider an illustration of a more difficult discrimination: At some stage into treatment, you teach the student pronouns such as my and your. Suppose the student is asked, 'What is your name?' in contrast to the question

SD l: 'Clap hands

R l: student claps hands (reinforced and strengthened)

R2: student touches table (not reinforced and weakened)

'What is my name?' For the student to answer correctly, he must attend to (discriminate) the relevant and significant elements in the question which centre on the use of the pronouns my and your. Consider more elaborate instructions, such as when a teacher asks, 'What animal gives us milk?' or 'Can you tell me what you did today?' There is no reason to believe that people are boom with the ability to answer such questions, but there is every reason to believe that the answers can and have to be taught, and that differential reinforcement plays a major role in this process. Although parents and other adults must provide corrective feedback, typical children learn such discriminations without anyone having to pay too much attention to the teaching processes; questions are asked and feedback is delivered in a more or less informal manner. However, individuals with developmental delays need specific teaching programs to acquire such discriminations. When provided with such programs, these students master many of the discriminations taught to them and some master all of them.

Mass Trials

As mentioned earlier, the first step in discrimination learning is the individual presentation of instructions in a mass trials format. In the mass trials procedure, repeated presentations of the same instruction are delivered to the student, and the student is prompted and reinforced for the correct response. Within a block of mass trials, the prompt is gradually faded, and mass trials are terminated when the student responds correctly without the prompt for a certain number of trials (the criterion for mastery). The receptive instructions 'Clap hands' and 'Touch table' are again used for illustrative purposes. Let SD1 be 'Clap hands' and SD2 be 'Touch table.' In other chapters throughout this manual, we provide detailed information on how to apply discrimination learning procedures to the various programs.

► Step 1

Present SD1 ('Clap hands') in mass trials; that is, repeatedly go through the process of presenting the instruction, waiting up to 3 seconds for the response, and then cons equating the response. If the student does not respond correctly (which includes a failure to respond), repeat the instruction, prompt simultaneously with the instruction, and then reinforce the correct response. The reason for prompting simultaneously with the presentation of the instruction is so that the temporal interval

Between the SD and the response can be as short as possible (ideally a second or less). A longer time interval is not optimal for the teaching of an association between the instruction and the response. For example, if you prompt the correct response after the student fails to respond within 3 seconds of the SD, the association between the prompt and the response is strengthened rather than the association between the instruction and the response.

Clapping may be prompted by placing your hands on the student's wrists and moving her hands together and apart. If the student has mastered clapping hands in the Nonverbal Imitation Program (Chapter 13), simply prompt the correct response by clapping your hands as a model for her to imitate. Continue mass trialling 'Clap hands.' Gradually fade the prompt over subsequent trials (by providing less and less assistance or by providing less and less of the model). Continue with mass trials until the student meets the criterion of 9 out of 10 or 19 out of 20 unprompted correct responses.

► Step 2

Present SD2 ('Touch table'). Mass trial SD2 and prompt the student to respond correctly by placing her hand on the table or by modelling the correct response. Gradually fade the prompt and continue reinforcement over subsequent trials. When the student responds correctly without a prompt, continue with mass trials until she meets the criterion of 9 out of 10 or 19 out of 20 unprompted correct responses.

Note that a round of mass trials resulting in mastery (e.g., 9 out of 10 correct) is unlikely to be achieved in one session of sitting. Although some students master this criterion in one session, most students require several separate sessions across several days (with different programs intermixed so as to reduce boredom and facilitate recall).

Intermixing Stimuli

Now that the student can correctly respond to both instructions when presented separately, intermix and differentially reinforce the instructions so the student can learn to discriminate between them. Intermixing is essential to ensuring mastery. It is a common mistake to believe that a student has mastered two labels when they are presented separately. To maximize the student's success when stimuli are intermixed, present each instruction in a strong, clear voice and set the inter-trial interval at 2 to 3 seconds initially, then gradually decrease the interval to 1 to 2 seconds. The closer the time interval between the two instructions, the closer the contrast between them and the more likely it is that the discrimination between the two instructions will be mastered. Moreover, by using a short inter-trial interval, you give the student little opportunity to self-stimulate. Do not present instructions while the student is engaged in self-stimulatory behaviour, such as gazing at lights, squinting his eyes, or flapping his hands. Rather, try to capture the student's attention before you present the instruction by showing the student the reinforcer he may earn. Or, present a warm-up trial of a simple task he knows well (e.g., imitating stomping feet). Remember that the student has mastered Steps 1 and 2 but may have temporarily lost either or both of the skills mastered in these steps because of an intervening activity. Once you have the student's attention, quickly move on to Step 3.

► Step 3

Present SD1 ('Clap hands'). Because the student was most recently reinforced for touching the table (in Step 2), it is likely that he will make a mistake and touch the table when instructed 'Clap hands.' If the student makes a mistake, he loses out on a reinforcer and may tantrum. To avoid this, prompt the correct response as soon as the instruction 'Clap hands' is presented. Reinforce upon completion of the correct response. Present mass trials of SD1 while fading the prompt. Place mastery of Step 3 at 4 unprompted correct responses in a row.

► Step 4

Within 3 seconds of completing mastery of SD1 ('Clap hands'), present SD2 ('Touch table') in a loud and clear voice and simultaneously prompt the student's correct response. By quickly prompting the correct response, the student is prevented from making an error (e.g., clapping hands). Set mastery at 4 unprompted correct responses in a row.

► Step 5

Reintroduce SD1 ('Clap hands'). If the student is not prompted, the most likely result is that

the student will respond by touching the table because he has not yet learned to discriminate between the two instructions. Therefore, prompt the clapping of hands before the student makes a mistake. Use the minimal amount of prompting necessary to obtain the correct response. Fade the prompt and go on to Step 6 after 3 unprompted correct responses.

► Step 6

Reintroduce SD2 ('Touch table') and again prompt, reinforce, and fade the prompt. After 3 unprompted correct trials, go on to Step 7.

► Step 7

Reintroduce SD1 ('Clap hands'). This time, test for mastery by suddenly withholding the prompt altogether. The prompt is removed to avoid reinforcing the student for prompt dependency as discussed in Chapter 10. If the student responds correctly without the prompt, present one more trial and then go on to Step 8. If the student responds incorrectly, use the minimal amount of prompting necessary to reinstate the correct response. Fade the prompt over subsequent trials. After 2 unprompted correct responses in a row, go on to Step 8.

► Step 8

Reintroduce SD2 ('Touch table'). Test for mastery by suddenly withholding the prompt altogether. If the student responds correctly, present one more trial and then go on to Step 9. If the student responds incorrectly, present a minimal prompt on the next trial. Fade the prompt over subsequent trials. Minimize or withhold reinforcement for responding on prompted trials while maximizing reinforcement for correct responding on unprompted trials. After 2 correct responses without prompting, repeat Steps 7 and 8.

► Step 9

Continue to alternate between the two instructions until the student responds correctly to each of them the first time they are presented in contrast with one another. Provide differential reinforcement by maximally reinforcing correct responses and withholding reinforcers for incorrect responses by acting as if the response did not occur. If acting as if the response did not occur fails to reduce errors, cons equate incorrect responses with an informational 'No' (not a harsh 'No') and remove reinforcers from the student's view (e.g., by placing them under the table). 'No' paired with the temporary removal of food or other reinforcers may eventually establish 'No' as corrective feedback. However, an informational 'No' may act as a reward for some students. Likewise, the teacher's looking down contingent on incorrect responding may also be rewarding for some students because it may signal a momentary escape from the teaching situation. The type of consequence used for an incorrect response should not be rewarding and should not last for more than 3 seconds.

If the student responds incorrectly on any given trial, reintroduce the least amount of prompting necessary to reinstate correct responding on the next trial and then gradually fade the prompt. Over successive intermixed and differentially reinforced trials, the student will make fewer and fewer mistakes until he responds correctly without prompts the first time the instructions are contrasted with one another. This process occurs for two reasons. First, as the prompts are faded and minimized in strength, they become less and less reliable as cues while the instructions remain reliable and accrue strength. Second, as correct responses are rewarded and strengthened, incorrect responses are not rewarded and are weakened. If discrimination training is done correctly, the student has no choice except to succeed in making the discrimination (i.e., telling the two instructions apart).

In Step 7 we recommended testing for mastery by suddenly withholding the prompt instead of slowly fading it. Remember that this is a test and that some but not all students are likely to respond correctly without a prompt at this early stage of learning. We have also recommended that the teacher switch between instructions, moving from SD1 to SD2 after a gradually decreasing number of trials on any one instruction until the student can respond correctly without a prompt when the instructions are presented singly (as opposed to presented in blocks of 4, 3, and 2). There are three reasons for these recommendations. First, the teacher wants to avoid establishing prompt dependency. Second, the teacher wants to reduce perseveration by not reinforcing the student for merely repeating the same response. Finally, there may be some value in letting the student make a mistake, such as SD1-R2, in the sense that a non-reinforced trial is mildly aversive and thereby helps to speed up the discrimination.

The numerical values specified earlier (e.g., switching after 3, then 2, then 1 correct response) are based on our experience with the average learning rate of students with developmental delays. If the student has required several hours to master the first or second instruction (SD1 or SD2), or both, when presented in mass trials, the criterion should be increased (e.g., to 5 unprompted correct responses in a row or 9 out of 10). As the student's mastery increases, reduce the criterion to 4 in a row, then to 3, and so on, before switching instructions. A number of subtleties in the constructive use of instructions and prompts are unique to particular teaching programs and will be discussed in subsequent chapters.

Students with developmental delays typically experience difficulty learning to discriminate between the first two instructions in any program. However, additional instructions are usually acquired with increasing ease. It is as if the student learns how to learn.

Areas of Difficulty

One helpful step for students who experience problems differentiating two particular instructions (e.g., 'Clap hands' and 'Touch table') is to change one or both instructions. For example, introduce 'Block in bucket' instead of 'Touch table' or 'Clap hands.' Another possibility is changing instructions to those that contain a more substantial and direct payoff (reinforcer), such as the instruction 'Drink juice' if the student likes juice or 'Light switch' if the student likes to turn on and off lights and a light switch is handy. These are examples of instructions the student may have already learned, and this prior learning may facilitate mastery of the discrimination.

If difficulties continue, try contrasting one of the SD (e.g., SD1) with a contrasting stimulus (SDCS) from another program, such as the Nonverbal Imitation Program. That is, intermix SD1 with SDCS (e.g., from nonverbal imitation, give the SD 'Do this' and model stomping feet). The SD1-SDCS discrimination should be easier for the student to discriminate because it involves a contrast between a visual stimulus and an auditory stimulus instead of two auditory stimuli. Thus, this discrimination could be used as a warm-up trial for the SD1-SD2 discrimination.

A few students (1 out of 10 or 15) may respond to two or three elementary instructions when presented separately (e.g., 'Sit down,' 'Drop block,' or 'Drink juice') but fail to learn to discriminate between any two verbal instructions when intermixed despite teachers' most extensive efforts using what is now known about discrimination learning. These students may be referred to as visual learners, and they may excel at acquiring discriminations among visually presented instructions such as those introduced in the Nonverbal Imitation Program (Chapter 13), the Reading and Writing Program (Chapter 29), and the Picture Exchange Communication System Program (Chapter 30). The discrimination learning procedures described in this chapter are just as important to teaching responses to visual instructions as they are to teaching responses to vocal instructions. If the student has difficulty learning, examine the teaching style before attributing failure to the student.

Random Rotation

The intermixing and differential reinforcement of instructions (or any stimuli) have been subjected too much scrutiny by psychologists with special interests in learning processes in order to find a way to maximize the number of correct responses and minimize the number of errors made by an organism when learning discriminations. In working out an optimal procedure for intermixing presentations of instructional stimuli, the field has developed a procedure that we label random rotation (technically termed random stimulus presentation). Random rotation of instructional stimuli may be considered a safety check and a strengthening measure that ensures the student has acquired the correct discrimination through the employment of differential reinforcement procedures. It is absolutely essential that you become familiar with this procedure.

Random rotation is a procedure in which presentations of two or more SD are intermixed in a non-specific, random order. An example of a random sequence of two SD follows: SD1, SD2, SD2, SD1, SD1, SD2, SD1, SD2, SD2, SD2, SD1, SD1, SD2, SD1, SD1, SD1, SD2, SDL The sequence is called random because it does not allow the student to predict which SD will appear next. To illustrate random rotation in the Early Receptive Language Program, let SD1 represent 'Clap hands,' SD2 represent 'Touch table,' and SD3 represent 'Block in bucket.'

An illustration of a random sequence of trials follows. Let the correct response to SD1 be Rl and the correct response to SD2 be R2. Remember, if the student fails to provide a correct response when the SDs are switched (as in going from SD1 to SD2, or vice versa), the teacher should prompt to establish correct responding, and then fade it before moving on.

Note that in Trials 4 and 5, the teacher repeats SD2 twice before switching to SDL This is done to strengthen the SD2-R2 association because the intermixing has just begun and the SD2-R2 may need strengthening. In Trial 6, the student loses the SD l-R l association, perhaps because she had been reinforced for SD2-R2 in the preceding trials (4 and 5). In both Trials 8 and 10, the student again loses the SD l-R l association. The teacher responds to this error by prompting twice (Trials 11 and 12) and repeating reinforceable trials twice (Trials 13 and 14) so as to strengthen the association before going on to SD2. Intermixing SD1 and SD2 (Trials 15, 16, 17, and 18) indicates that the discrimination may have been mastered unless the student learned a certain pattern of alternating between responses, a problem we describe in more detail in the next section. The correct response in Trial 18 is an early test of whether this problem exists, and the student's correct response suggests that it may not. The student would have given Rl in Trial 18 had she been alternating and not attending to the teacher's instruction.

Before we proceed, it may be helpful to comment further on the loss of the correct response in Trial 8, which occurred after the student had been prompted and reinforced for the correct response in Trial 7. The same loss occurred in Trial 10. The teacher went from Trial 7 to 8 and from Trial 9 to 10 abruptly, withholding the prompt to probe whether the student had achieved mastery. It is impossible to probe without withholding prompts, and it is not possible to predict the exact point at which to suddenly withdraw the prompt to test whether the student has achieved mastery. However, the teacher will become more proficient at determining when to probe as he or she gains experience. Note that the teacher went on to establish the correct SD1-R1 association in Trials 11 through 14.

Common Problems When Intermixing Instructions

It is important to help ensure that the student is not misled when taught to discriminate among stimuli. As illustrated in this section, it is easy to mislead a student when intermixing instructions.

► Win-Shift Strategy

If the teacher systematically alternates instructions, as in SD1, SD2, SD1, SD2, SD1, SD2, and so on, it may seem that the student should learn to tell the two instructions (SD) apart. Such a systematic alternation was described previously when SD1 ('Clap hands') and SD2 ('Touch table') were intermixed. Some students may acquire the discrimination under this condition, but others will not. From the student's point of view, instead of attending to the auditory stimuli, it may be easier to solve the problem by learning that if he was reinforced for responding correctly on one trial, he needs to switch responses on the next trial to get reinforcement. In technical terminology, reinforcement of R l ('Clap hands') rather than the teacher's instruction becomes the SD for R2 ('Touch table'). This is called a winshift strategy. In short, such a procedure is not conducive to helping the student attend to the teacher's instructions.

WinStay Strategy

If the teacher repeats the same instruction across numerous trials, as in SD1, SD1, SD1, SD1, SD1, and so on, a procedure similar to the mass trials phase used during the acquisition of a response, the teacher's intent may be to strengthen the student's response to the instruction provided. However, the student may be reinforced for merely repeating what he did the last time and not for listening to the teacher's instructions. He may learn that if he responds the same as the last time he was reinforced, he will again be reinforced. The student is taught to perseverate. This is called a win-stay strategy. In technical terminology, the SD for the student's response is the reinforcement received for the performance of the same response done in the preceding trial and not the teacher's instruction. Random rotation of SD is critical to preventing this.

► Lose-Shift Strategy

If the teacher presents SD1 ('Clap hands') and the student responds with R2 (touching the table) and is not reinforced, and the teacher then repeats SD1 ('Clap hands') and the student responds with Rl (clapping hands) and is reinforced, the student may learn that if reinforcement was withheld for touching the table in the last trial, he needs to switch to clapping in next trial. This is called a lose-shift strategy in which the loss of a reinforcer becomes the SD for shifting to an alternate response. One way to break up such a strategy is to insert an already mastered SD-R trial after the reinforced SD1-R1 trial. The contrasting stimulus should consist of an instruction to which the student can already respond, such as 'Stand up,' or a nonverbal imitation trial as in waving in response to a model. The sequence of SD may then proceed in the following manner:

Trial  SD and Result

SD l-incorrect R, do not reinforce

SD1correct R, reinforce

Present a contrasting stimulus and reinforce correct R.

Present SDL If the student responds with R l, the teacher may be more confident that the student's correct response is associated with SD1 and not the loss of a reinforcer on a preceding trial.

Another example of inadvertently teaching a lose-shift strategy occurs when the teacher lets the student self-correct. Self-correction may be helpful for typical individuals and for those who master advanced programs, but it may lead to serious problems for students with developmental delays in the early stages of learning. Self-correction is discussed later in Mistake 12 of the section 'Areas of Difficulty.'

The illustrations provided in this section show that it is easy for the student to learn the wrong associations. He seems efficient at gaining reinforcers, perhaps because he is expending less effort than would be needed for paying attention to what the teacher is saying. Maximizing gains while minimizing effort is not only characteristic of students with developmental delays, but of all living organisms.

Introducing the Third Discrimination

Mastery criterion when SD1 and SD2 ('Clap hands' and 'Touch table') are randomly rotated may be set at 9 out of 10 or 19 out of 20 unprompted correct responses. At that point, we strongly advise teachers to further strengthen mastery of the discrimination between these two instructions by actively rehearsing the discrimination over the next 2 or 3 days, generalizing it across teachers and across different physical locations (e.g., across rooms of the house), and intermixing other programs such as Matching and Sorting (Chapter 12) and Nonverbal Imitation (Chapter 13). Intermixing the discrimination with other tasks helps strengthen recall, a phenomenon more fully discussed in Chapter 31. The discrimination between the first two stimuli (SD1 and SD2) is the most difficult, and success should be considered a great achievement for both the student and the teacher. Once the first discrimination is firmly established, new stimuli may be added at a faster rate and with less likelihood that the first discrimination will be lost.

Once you feel assured that the SD1-SD2 discrimination is mastered, a third instruction (SD3, such as 'Block in bucket') can be introduced. Mass trial SD3, then intermix and differentially reinforce SD3 with SD1. Once SD3 is randomly rotated with SD1 and mastered, intermix SD3 with SD2. Follow up with random rotation. Use the same procedures to introduce new SD.

A typical outcome from discrimination learning is that, with the introduction of SD3 and for each new SD thereafter, there is a concomitant decrease in the number of errors the student makes. This happens because the student learns to attend to (discriminate) the teacher's instructions. This is an inevitable outcome when the correct response is reinforced and thus strengthened and incorrect responses are not reinforced and thus weakened. Eventually the student may become so attentive to the teacher's instructions that one prompt of the correct answer is sufficient for learning. This is known as one-trial learning.

Areas of Difficulty

The student may demonstrate little or no progress in learning discrimination after several trials. For example, the student may not master the discrimination between the first two instructions after 3 hours of instruction even though you adhere to correct discrimination learning procedures. It is very easy to make mistakes when teaching children with developmental delays. Mistakes are likely to reinforce behaviours that will hinder or block new and appropriate learning. Consider the following teaching mistakes:

► the stimuli being contrasted are too similar. If this could be the problem, re-examine whether the teacher is in fact beginning with stimuli that are maximally different and that require maximally different responses (when appropriate).

► Members of the team are inconsistent in what and how they teach. Check consistency at the weekly team meeting by having each teacher demonstrate his or her skills with the student.

► the particular task the teacher employs is especially difficult for the student. If this is a possibility, try another task for contrast. For example, if the student mastered SD1 ('Clap hands') but continuously makes errors when this instruction is mixed with SD2 ('Touch table'), then shift to another instruction such as 'Block in bucket.' Reintroduce 'Touch table' later.

► there are problems with reinforcers and motivation. If the student reaches a plateau or begins to lose mastered tasks, try the following possible remedies: (a) if this problem is caused by ineffective reinforcers, search for new and more powerful reinforcers or avoid satiation on older ones. It is a common mistake for teachers to repeatedly use consequences, such as saying, 'Good,' 'Nice job,' 'Fine,' 'Splendid,' 'Fantastic,' 'Super,' 'Right on,' 'Excellent,' and 'Yes,' all to no avail, (b) If self-stimulatory behaviours have shown a marked increase, engage the student in attention exercises (also known as wake-up exercises) for a 10- to 15-second duration by rapidly alternating between SD of already mastered tasks (such as 'Sit down,' 'Stand up,' 'Turn around,' or nonverbal imitation tasks) and then suddenly presenting the SD you are training, (c) Make certain the reinforcement is immediate; there should be no delay between the student's response and the reinforcement. The effectiveness of the reinforcer decreases exponentially over the time elapsed between a response and the delivery of reinforcement. Data suggest that a 1-second delay between the response and the reinforcer is only 25% as effective in strengthening a response as a .5 second delay (Reynold, 1968). If the delay is longer than .5 second, the student is likely to have started doing something else and that behaviour will inadvertently be the one that receives reinforcement (i.e., the behaviour performed immediately before the delivery of reinforcement will be strengthened). If a teacher uses a reinforcer that cannot be delivered immediately (e.g., food), it is recommended that the teacher bridge the delay by immediately giving verbal praise while reaching for and providing a food reinforcer. Through such a procedure, verbal praise (e.g., 'Good') is likely to acquire reinforcing properties because it signals (becomes associated with) an already powerful reinforcer (i.e., food consumption), (d) Make sure the consequences for incorrect trials do not inadvertently function as positive reinforcers and thereby increase incorrect responding. This could happen if the student receives a pause in teaching by, for example, the teacher talking to the student ('I know you are trying hard and you are getting upset, but we have to continue'), (e) Do not reinforce the student before she completes her response. If you do so in your eagerness to help the student, the student is reinforced before she commits to a response. If this occurs, the student does not have the opportunity to complete, be reinforced for, and learn the correct response, (f) Be sure not to overwork the student. If the student makes rapid progress, she has reinforced the teacher for introducing an ever-increasing number of new and perhaps more difficult tasks. The student, however, often cannot vocally communicate to the teacher when it becomes too much and may fall back on an earlier, well-practiced strategy of tuning out or tantruming. If this occurs, simplify the program, reinstate motivation, and then move forward at a slower pace, (g) If the student's entire curriculum, not only a specific program, is too academically oriented, behaviours such as tantruming toward the end of the day and non-responsiveness across programs may become evident. To help avoid these problems, introduce play breaks between sittings and intersperse other reinforcing activities between academically oriented programs.

► the teacher may give inadvertent prompts, such as looking at the student's hands when asking the student to 'Clap hands' or looking at the table when asking the student to 'Touch table.' If the student is reinforced for responding to inadvertent prompts, she will not learn the desired association between the teacher's request and her response. Also, if the student is not inadvertently prompted in the same way at another time or by other teachers, then she will fail to respond correctly. Other kinds of inadvertent prompts also may delay mastery. For example, the instructions may differ in length, as when a one-word SD such as 'Table' is contrasted with a two-word SD such as 'Clap hands.' Length of the instructions may help prompt the correct response, but the teacher must be aware of such prompts and must fade them by changing the instructions to make them of equal length.

► the student responds to some element of the instruction not intended to be solely responded to, such as the first part of the instruction (e.g., 'clap' in 'Clap hands' or 'touch' in 'Touch table'). To check for this possibility, merely present the instructions as 'Hands' and 'Table.' Place the instructions 'Hands' and 'Table' into discrimination training if the student fails to respond correctly. Once mastered, fade in 'Clap' and 'Touch' by first presenting these words at a very low decibel level (whispering them) and then gradually increasing their volume until they are stated as loudly as 'hands' and 'table.' Note, however, that is the student is more likely to respond to the last part of the instructions than the first part because the last part is the most recent stimulus element. Therefore, in the case of 'Clap hands' and 'Touch table,' it is more likely that 'hands' and 'table' will become associated with the correct response than 'clap' and 'touch.' The student is taught to 'store' instructions of increasing length as she progresses through language programs.

► the teacher shifts to a new SD after a prompted response on the preceding trial. For example, in Trial 1 the teacher presents SD l and the student responds with R2 (an incorrect response) or not at all. In Trial 2, the teacher presents SD l once more, prompts, and reinforces the correct response, R l. The teacher then goes on to Trial 3 and presents SD2 with the intent of contrasting SD l with SD2. The mistake occurs in Trial 3, where the prompt was not faded for SD l. The student was not reinforced for responding to SD l with R l, but rather for responding with R l to the prompt. In such a situation, SD2 is contrasted with a prompt instead of SD l. This situation does not help the student learn to discriminate between the relevant stimuli (SD l and SD2 in this example). In other words, very little learning occurs if the student simply gives responses to prompts. It is easy to make this mistake in one's eagerness to help the student move along.

► the teacher inadvertently reinforces incorrect responses by, for example, providing encouragement to the student when she gives a mistaken response. Examples of such encouragement include 'Good try,' 'Almost right,' 'Try again,' and 'We don't want to do that, do we?' If the student responds to such consequences as rewarding, the teacher may inadvertently strengthen incorrect responses. A typical individual may know the meaning of such feedback and not respond as if being reinforced for an incorrect response. However, it is best not to assume that a person with language delays comprehends such comments. As mentioned earlier, even a consequence such as 'No,' which is aversive for many students, may serve as a positive reinforcer for some students. If in doubt, withhold any and all potential reinforcers for incorrect responses so as to avoid reinforcing errors.

► an extended time interval occurs between trials. For example, the teacher talks too much between trials as in commenting, 'That is good clapping hands. Let's try it one more time, you are doing so well.' Or, the teacher takes time away from teaching to record whether the student made the correct response. Too much time may also pass if the student is given a chewy food or a large amount of food that takes a long time to consume. In each instance, time between trials may result in the student drifting off into self-stimulatory behaviour, making it difficult for the teacher to regain her attention to the appropriate stimuli. Short intertrial intervals allow for the presentation of more trials than is possible when intervals between trials are longer. In general, the more trials, the more often a response is reinforced and the stronger that response becomes. The practice of lengthening intertrial intervals should be employed only when introducing the procedures intended to maintain already mastered material (Chapter 31).

► the time interval between the teacher's instruction and the student's response is too long. This interval should be as short as possible (not exceeding 3 seconds). Avoid prolonged intervals by providing feedback (informational 'No') if the student does not respond 3 seconds after the instruction, and then immediately start another trial, prompting the desired response if necessary. Technically speaking, for a stimulus-response association to be made, the two events should occur concurrently (i.e., overlap) in the student's nervous system. If a teacher presents an instruction and the student delays responding, other intervening stimuli may become associated with the response instead of the teacher's instruction.

► The SD and the prompt do not overlap. Rather, the SD is presented, 2 or more seconds pass, and then the teacher prompts the correct response. For the SD to take control over the student's response, the SD and the prompt must occur simultaneously (overlap) in the student's nervous system. If the student does not respond to the SD, repeat the SD and simultaneously present the prompt.

► the teacher allows the student to self-correct. That is, the student makes the incorrect response (either in whole or in part), hesitates for a brief moment as if waiting for the teacher's reaction, and then, if no reinforcement is forthcoming, switches to the correct response. Reinforcement of the correct response may seem like the appropriate strategy except that the correct response may now be associated with the teacher's withholding of reinforcement for the aborted incorrect response. This is similar to the lose-shift strategy, which was discussed earlier, and to the problem discussed in the paragraph numbered 7. If the student self-corrects, quickly end that trial and then provide the SD again, prompting the correct response to avoid self-correction.

►A variation on intermixing SD1 with SD2 for the student who experiences serious problems during acquisition is to intermix SD1 with an already mastered SD from another program, such as Nonverbal Imitation (this was earlier referred to as the use of a contrasting stimulus). This step should be considered a pre-training step or a warm-up. Mastery of such discrimination should not be accepted in place of the target discrimination; it is intended to facilitate the student's mastery of subsequent discriminations. There are three advantages to a procedure that intermixes different kinds of SD: (1) The student is more likely to acquire a discrimination between two instructions when one is auditory and the other is visual; (2) SDCS may have reinforcing properties, thereby helping to maintain the student's cooperation with the teaching situation; and (3) inserting SDCS helps improve recall of SD1 (see Chapter 31).

► The discrimination learning procedures described in this chapter use prompts that are ideally suited to help teach discriminations in programs containing auditory instructions or simple visual stimuli, such as in the Early Receptive Language, Matching and Sorting, Nonverbal Imitation, and Verbal Imitation Programs. The prompts are reasonably simple, such as moving the student's hand to help execute the appropriate action or emphasizing a critical element in the teacher's instruction. However, several of the programs outlined in this manual require the student to make a selection from several visual stimuli, such as in the Receptive Identification of Objects and Receptive Identification of Behaviours Programs. For example, in the Receptive Identification of Objects Program, the teacher names an object and teaches the student to select that object from among several objects placed on the table in front of the student. To facilitate correct responding in such a program, certain additional prompts must be introduced in discrimination learning procedures to help the student make the association between a particular verbal or printed label and the object that the label denotes. Prompts specific to particular programs are described as each program is introduced.

Other Problems

If one has exhausted all known possibilities and the student still makes little or no progress, consider the possibility that the student may have a neurological deviation that makes it difficult for her to associate certain kinds of stimuli, such as auditory stimuli, with behaviours. In our experience, 1 out of 10 to 15 students appears to experience this problem to an extreme degree, acquiring a reliable response to only one or two verbal instructions after a month or more of instruction. Such students appear to be primarily visual learners and are likely to make progress in programs involving visual stimuli, including the Matching and Sorting, Nonverbal Imitation, Reading and Writing, and Picture Exchange Communication System Programs (Chapters 12, 13, 29, and 30, respectively). The student's exposure to visual instructions may later facilitate her recognition of verbally presented material.

Another possibility is that the student has a hearing or visual impairment. Most parents have already suspected such impairments and have taken their child for a comprehensive neurological evaluation. It is recommended that a neurological evaluation be scheduled as part of the intake (pre-treatment) assessment. If this has not been addressed, it should be at this point.

In the context of obtaining a comprehensive neurological examination, we raise a warning against 'over-pathologizing' because one can risk handicapping both the student and the teacher. Most teachers have been exposed to alternate treatment approaches (see Chapter 2). Various instructional sets or philosophies may have been acquired through this exposure and could interfere with the effective use of the teaching principles advocated in this manual. Although rewarding correct responses and withholding rewards for incorrect responses may sound reasonable, basic, and commonsensical, many adults find it difficult to follow such advice because individuals with developmental delays have historically been conceptualized as qualitatively different from typical individualsas extremely frail, anxious, and the like. Therefore, qualitatively different treatments, such as non-contingent love and acceptance, have been applied, often independent of how the person behaves. Gentle Teaching, Options, Holding Therapy, and psychodynamic intervention illustrate such treatments. Research has indicated that many persons, even when not adhering to any of these philosophies, have a tendency to reinforce any responses of individuals with developmental delays, including incorrect ones (Eikeseth &. Lovaas, 1992).

Teachers may also have been taught to rely on theories that place an emphasis on the idea that development unfolds according to maturational stages. Such theories have been influential in determining education for both typical persons and persons with developmental delays. Some of these theories are based on developmental psychology and include the views of Chomsky, Piaget, and Kohlberg. These theories imply that a teacher may have to wait for certain maturational stages to manifest themselves and then, with minimal exposure to language and educational material, competence will occur. Whether or not these theories are true of typical children is a matter of heated debate, but they clearly have been found to be inapplicable to children with developmental delays.

Finally, there are those who propose that the goal of a teacher is to free a frightened little person out from behind an autistic shell, a person capable of understanding what the teacher may be communicating. Once out of the shell, the person would be expected to speak or write in full sentences and otherwise behave like a typical individual. Such were the theories of Bettelheim (1967).

What this and the theories presented earlier have in common is that they are well known; sound attractive, comforting, and humanistic; can be mastered by a teacher attending a workshop of a 1-week or less duration; and allow the teacher to assume a more passive role toward the individuals with developmental delays. Although all of these interventions may be well meaning, none of them have demonstrated effectiveness, and several seem to cause more harm than good (see Chapter 3). A less speculative approach is to assume that the individual has a potential for learning and that there is not a qualitative difference between the processes by which individuals with developmental delays and typical individuals learn. However, with a learning-based behavioural approach, the teacher will need a great deal more training to become competent and the child, the teacher, and the parent will have to work much harder.