Teaching simple and conditional discriminations

Much of what we teach comes down to one question: is the learner responding to the right stimulus? This lesson covers how to set up discrimination training so the answer is yes.

Why it matters

Listener skills, matching, sorting, reading, and many daily living skills are discriminations. When a program is set up poorly, a learner can score well on data sheets while responding to the wrong thing, such as where the card sits, the order of trials, or where the teacher is looking. That skill falls apart as soon as the setting changes.

As an analyst, you need to name the type of discrimination, spot faulty stimulus control in the data, and pick the program change most likely to put responding under control of the intended stimulus.

Key ideas

• Discrimination training is differential reinforcement across stimuli. You reinforce a response when the discriminative stimulus (SD) is present and withhold reinforcement when an S-delta is present. Over time the response occurs more in the presence of the SD.
• In a simple discrimination, one stimulus is always correct. The SD and S-delta keep their roles from trial to trial. The relation is the familiar three-term contingency: SD, response, consequence.
• In a conditional discrimination, which stimulus is correct depends on another stimulus. Picking the cup is reinforced when the sample is "cup" and not when it is "spoon." The sample adds a fourth term that sets the occasion for the whole three-term relation.
• Matching-to-sample (MTS) is the standard format for conditional discriminations. You present a sample, then an array of comparisons. One comparison is correct (S+) and the others are incorrect (S-) for that sample.
  • Identity matching: the correct comparison is physically the same as the sample (photo to identical photo).
  • Arbitrary matching: the sample and the correct comparison share no physical features (spoken word "dog" to a picture of a dog, or a printed numeral to a set of objects).
• Arrange trials so only the intended stimulus predicts reinforcement. Rotate which sample is presented, counterbalance the position of the correct comparison, avoid predictable patterns (such as strict alternation), and use arrays of at least three comparisons when you can.
• Teach the conditional discrimination from the start when that is the goal. Teaching one target alone in massed trials, then adding others later, can build control by the comparison alone ("always pick the car"). Many current recommendations favor rotating samples within a mixed array from the beginning.
• Make sure the learner observes the sample. A differential observing response is a response that differs for each sample, such as naming it or picking its identical copy, before the comparisons appear. Simply touching the sample shows orientation, not discrimination.

Definitions

• Simple discrimination: a response occurs reliably in the presence of one stimulus (SD) and not in the presence of another (S-delta), and the SD stays the same across trials.
• Conditional discrimination: the stimulus that functions as the SD for a response depends on a second stimulus, called the sample or conditional stimulus.
• Sample stimulus: the stimulus presented first in a matching-to-sample trial, which determines which comparison is correct.
• Comparison stimuli: the array of stimuli the learner chooses from on a matching-to-sample trial.
• Faulty stimulus control: responding controlled by an irrelevant feature (position, trial order, teacher cues) or by only part of the target stimulus (the first letters of a printed word) rather than the whole stimulus the program targets.
• Differential observing response (DOR): a response to the sample that differs depending on which sample is shown (for example, naming it), so it shows the learner has discriminated the sample before choosing a comparison.

Examples from a session

When you review a discrimination program, look past percent correct. Check whether positions were counterbalanced, whether samples were rotated, and whether errors show a pattern, such as always choosing the left card or always choosing the last target trained.

Common mistakes / exam traps

• Calling receptive identification a simple discrimination. If the correct picture changes with the spoken instruction, it is a conditional discrimination with the instruction as the sample.
• Reading high accuracy as mastery. If the correct card was usually on the left, 90% correct may only show a position bias. Probe with new arrangements.
• Teaching targets in isolation and calling it a head start. Massed trials of one target, with no rotation, can build a habit of selecting that item regardless of the sample.
• Confusing identity and arbitrary matching. Identity matching requires physical similarity. Word-to-picture and numeral-to-quantity are arbitrary.

Quick check

Related topics

• Recognizing stimulus control (B.12)
• Recognizing stimulus discrimination (B.13)
• Building motivating operations and discriminative stimuli into interventions (G.5)
• Choosing prompting procedures: errorless, most-to-least, and least-to-most (G.7)
• Fading prompts: prompt delay and stimulus fading (G.8)
• Teaching for emergent relations and generative performance (G.19)