12.2 Teaching Simple and Conditional Discriminations
Key Takeaways
A simple discrimination is taught by reinforcing a response in the presence of the SD and withholding reinforcement in the presence of S-deltas, which is differential reinforcement across stimuli.
A conditional discrimination, such as matching-to-sample or receptive identification, requires choosing a comparison according to the sample presented, so the learner must attend to the sample on every trial.
Prevent faulty stimulus control by counterbalancing comparison positions, rotating which stimulus is correct, removing unintended cues, and adding observing responses (Green, 2001; Grow & LeBlanc, 2013).
Teaching targets together in conditional-discrimination format from the start was more efficient than first teaching each target alone (Grow et al., 2011), and errorless procedures can reduce errors during acquisition.
From Stimulus Control to Teaching
Section 3.3 explained stimulus control as a principle. Task G.4 asks you to build it on purpose. Almost every academic and language skill is a discrimination: saying "red" only for red objects, reading "cat" only for the printed word cat, or handing over a spoon only when someone asks for a spoon.
Simple Discriminations
In a simple discrimination, one stimulus (the SD) sets the occasion for reinforcement, and other stimuli (S-deltas) do not.
Teaching steps:
- Present the SD and reinforce the target response.
- Present S-deltas and withhold reinforcement for the same response (extinction), or reinforce a different response.
- Alternate SD and S-delta trials unpredictably until responding occurs mainly in the presence of the SD.
Simple discriminations can be taught successively (one stimulus presented at a time, such as a light that is either on or off) or simultaneously (two or more stimuli present at once, and the learner selects one). Everyday examples include crossing the street only when the walk signal is lit and answering the phone only when it rings.
Conditional Discriminations
In a conditional discrimination, the correct response to a stimulus depends on another stimulus. The classic procedure is matching-to-sample (MTS):
- A sample stimulus is presented (a picture of a dog, or the spoken word "dog").
- Several comparison stimuli are presented (pictures of a dog, a car, and a cup).
- Selecting the comparison that goes with the sample is reinforced; which comparison is correct changes from trial to trial with the sample.
Identity matching uses physically identical stimuli (picture to picture). Arbitrary matching relates stimuli with no physical similarity (spoken "dog" to a dog picture; printed dog to the picture), which is the basis for language comprehension and for equivalence-based instruction (section 12.4).
Receptive identification ("Touch the cup") is a conditional discrimination with an auditory sample: the learner must listen to the instruction, then scan the comparisons. If the learner ignores the sample, they may still be correct by chance or by using an unintended cue.
Recommended Practices for Arrays
Research reviews of receptive and matching instruction (Green, 2001; Grow & LeBlanc, 2013) recommend:
- Use arrays of at least three comparisons so chance accuracy stays low.
- Counterbalance positions so each target appears equally often in each location.
- Rotate the correct stimulus so each comparison is sometimes correct and sometimes an S-delta.
- Present the sample in a consistent order relative to the comparisons and secure attention before presenting the SD.
- Remove unintended cues: the instructor's gaze toward the correct card, a smudge on one card, a predictable left-right pattern, or always placing the newest item closest.
- Add an observing response when samples are similar: having the learner touch, name, or echo the sample before choosing ensures they attend to it (a differential observing response).
- Teach targets together in conditional format. Grow and colleagues (2011) compared the simple-conditional method (teaching each target alone first, then adding distractors) with the conditional-only method (teaching targets together in arrays from the start) and found the conditional-only method more efficient with fewer errors.
Errorless Discrimination Training
Errorless procedures arrange instruction so the learner rarely makes errors during acquisition:
- Stimulus fading: make the correct comparison more salient at first (larger, brighter, closer), then fade the extra dimension.
- Stimulus shaping: gradually change the form of the stimulus from an easy discrimination to the target one.
- Prompt delay: begin with a 0-second delay (prompt delivered immediately), then insert a fixed or progressive delay so the learner can respond before the prompt.
Errorless procedures reduce frustration and error patterns, but probes without prompts are needed to confirm that the intended stimulus, not the prompt, controls responding.
Diagnosing Faulty Stimulus Control
| Data pattern | Likely problem | Fix |
|---|---|---|
| Correct only when the target is in one position | Position bias | Counterbalance positions; reinforce only correct selections in other positions |
| Correct when the instructor looks at the target, not otherwise | Instructor cue | Neutral gaze, screen, or a second instructor |
| Selects the last reinforced item regardless of the sample | Not attending to the sample (win-stay pattern) | Rotate targets; add an observing response to the sample |
| Responds to one feature only (e.g., the color, not the shape) | Stimulus overselectivity | Teach with multiple exemplars that vary the irrelevant feature |
| High accuracy in teaching, low accuracy in natural settings | Control by teaching materials | Vary materials, settings, and instructors (train loosely) |
Data and Mastery
Record accuracy per trial and review position and target patterns, not just overall percentage. A learner who is 66% correct in a three-item array might be correct on every trial in two positions and wrong in the third, which points to a position bias rather than slow learning. Use unprompted first-trial (cold) probes to judge mastery, and probe with new exemplars before calling a discrimination generalized.
Common Exam Traps
- Calling receptive identification a simple discrimination. The correct choice depends on the spoken instruction, so it is conditional.
- Using one-item "arrays." With only one comparison, the learner can be correct without discriminating anything.
- Treating high percentages as proof. Always check for positional or instructor cues.
- Assuming errorless means no testing. Prompt-free probes are required.
On each trial, a learner hears "Show me the horse" or "Show me the pig" and must touch the matching picture in an array of three animal pictures. What type of discrimination is being taught?
Response differentiation, because only one pointing topography is reinforced
Respondent discrimination, because the spoken word elicits a pointing reflex
A simple discrimination, because the horse picture is always the SD
A conditional discrimination, because the correct picture depends on the spoken sample
A learner is 100% correct when the target card is on the left, 90% in the middle, and 10% on the right. What is the most appropriate change?
Teach each card alone with no distractors until accuracy reaches 100% again.
Move the target to the middle position for every trial to reduce errors.
Counterbalance target positions and reinforce only correct choices in all positions.
Raise the reinforcer magnitude for correct responses so motivation improves.
Two samples, the printed words "bat" and "bag," are often confused in a matching task. Which procedure is most likely to improve attention to the sample?
Increase the array from three comparisons to five comparisons right away
Require the learner to say each sample word aloud before choosing a picture
Reinforce every selection regardless of accuracy to keep the learner engaged
Present only the picture comparisons and remove the printed sample word
Sections you finish are checked off in the contents.