10.4 Skill Mastery Criteria, Generalization Probes, & Maintenance

Skill Mastery Criteria, Generalization Probes, & Maintenance

Exam Tip: On the QASP-S exam, remember Stokes & Baer's (1977) foundational maxim: "Train and hope is not an acceptable clinical strategy." Generalization does not occur automatically; it must be systematically programmed from day one. When evaluating mastery criteria, ensure the standard includes at least two different interventionists across at least three consecutive sessions to prevent "technician-specific stimulus control." Differentiate clearly between Stimulus Generalization (emitting the same trained response across novel stimuli, people, or settings) and Response Generalization (emitting novel, untrained behavioral topographies that achieve the same functional outcome). In maintenance programming, fading from Continuous Reinforcement (CRF) to Intermittent Reinforcement must occur gradually to avoid ratio strain.

A common failure in autism intervention is the "mastered on paper, forgotten in life" syndrome: a learner demonstrates 90% accuracy on a target during structured 1:1 sessions with a primary technician, but cannot perform the skill when asked by a parent at home, when presented with novel materials, or two weeks after the program is archived. In Applied Behavior Analysis, a skill is not truly mastered until it exhibits durability (maintenance), portability (generalization), and fluency (speed and accuracy). A QASP-S is ethically and clinically responsible for engineering acquisition programs that withstand the tests of time, context, and environmental variation.

---

Setting Empirical Mastery Criteria

Mastery criteria represent the predefined, objective performance standard that a learner must achieve before an acquisition target is considered learned, moved to generalization, or transitioned to maintenance. Establishing arbitrary or overly lenient criteria leads to rapid behavioral extinction and skill regression, while setting overly stringent criteria wastes instructional time and induces satiation, boredom, and avoidance.

                      ANATOMY OF A DEFENSIBLE MASTERY CRITERION
  ┌─────────────────────────────────────────────────────────────────────────────┐
  │ [Accuracy Threshold] ──► Typically 80% to 90% for discrete skills           │
  │                          (Strictly 100% for critical health & safety skills)│
  │       ▲                                                                     │
  │       │ COMBINED WITH                                                       │
  │       ▼                                                                     │
  │ [Temporal Stability] ──► Sustained across ≥ 3 consecutive sessions / days   │
  │       ▲                                                                     │
  │       │ COMBINED WITH                                                       │
  │       ▼                                                                     │
  │ [Staff Generalization]─► Demonstrated across ≥ 2 distinct interventionists  │
  │       ▲                                                                     │
  │       │ COMBINED WITH                                                       │
  │       ▼                                                                     │
  │ [Stimulus Diversity] ──► Verified across ≥ 3 novel exemplar stimuli/settings│
  └─────────────────────────────────────────────────────────────────────────────┘

Clinical Components of a Complete Mastery Standard

1. The Quantitative Accuracy Benchmark:
   • Standard Developmental / Educational Skills (Receptive labeling, tacting, matching): Set at 80% to 90% unprompted (independent) correct responding. Requiring 100% on complex cognitive tasks can lead to over-training and rigidity.
   • Health, Safety, and Critical Adaptive Skills (Pedestrian street crossing, answering emergency contact, swallowing medication, unbuckling car seat, fire alarm evacuation): Set strictly at 100% independent accuracy across all trials. In safety repertoires, an 80% success rate represents a life-threatening 20% failure rate.
2. The Multi-Session Consistency Rule: Performance must be demonstrated across a minimum of 3 consecutive sessions (or across 3 consecutive days/opportunities). A single high-scoring session may simply reflect temporary motivational spikes, fortunate guessing, or transient caregiver presence.
3. The Multi-Interventionist Rule: The learner must meet the performance threshold across at least two different interventionists (e.g., Technician A and Technician B, or Technician and Caregiver). Requiring multiple observers prevents technician-specific stimulus overselectivity, where a learner responds only to the subtle idiosyncratic vocal inflections, body postures, or facial micro-expressions of one familiar adult.

---

Behavioral Fluency & Overlearning

Accuracy alone is an incomplete measure of competence. A student who correctly answers 10 math facts in 10 minutes (1 per minute) is 100% accurate, but hopelessly non-fluent compared to a peer who completes 40 math facts in 1 minute. In behavioral science, Fluency is defined as the combination of accuracy plus speed of responding (Frequency = Count / Time).

                              THE BINDER RESA MODEL
  ┌─────────────────────────────────────────────────────────────────────────────┐
  │ R ──► RETENTION: Skill is retained across long intervals without practice.  │
  │ E ──► ENDURANCE: Performance is maintained over prolonged work periods      │
  │                  without cognitive fatigue or behavioral deterioration.    │
  │ S ──► STABILITY: Responding resists environmental distractions and noise.   │
  │ A ──► APPLICATION: Component fluent skills combine easily into complex      │
  │                    composite repertoires (e.g., letter sounds into reading).│
  └─────────────────────────────────────────────────────────────────────────────┘

The Concept of Overlearning

Overlearning refers to deliberate, continued instructional practice on a behavioral repertoire beyond the initial achievement of the mastery criterion (e.g., providing an additional 50–100 trials of practice after reaching 90% accuracy across 3 sessions).

• Neurological and Operant Benefits: Overlearning automatizes motor and cognitive chains, driving the response toward behavioral fluency. It significantly enhances resistance to forgetting, decreases response latency, and ensures that the behavior will withstand stress, emotional dysregulation, and environmental distractions.

---

Generalization Probes Across Stimulus & Response Dimensions

In their classic 1977 paper, Donald Baer and Trevor Stokes reviewed the literature and identified that generalization was frequently treated as an unexpected, passive byproduct rather than an active operant technology. They established that generalization must be programmed and verified through systematic probes.

                               STIMULUS VS. RESPONSE GENERALIZATION
 ┌──────────────────────────────────────────────┬──────────────────────────────────────────────┐
 │            STIMULUS GENERALIZATION           │           RESPONSE GENERALIZATION            │
 ├──────────────────────────────────────────────┼──────────────────────────────────────────────┤
 │ DEFINITION: The emission of the SAME trained │ DEFINITION: The emission of NOVEL, UNTRAINED │
 │ response in the presence of novel, untrained │ behavioral responses that achieve the SAME   │
 │ antecedent stimulus conditions.              │ functional outcome as the trained response.  │
 ├──────────────────────────────────────────────┼──────────────────────────────────────────────┤
 │ EXAMPLES ACROSS 3 AXES:                      │ CLINICAL EXAMPLES:                           │
 │ 1. People: Child says "Hello" to novel peer. │ • Trained Response: Saying "Hello".          │
 │ 2. Setting: Child uses spoon at restaurant.  │   Novel Untrained Responses: Saying "Hi",     │
 │ 3. Materials: Child identifies "Cup" using a │   "Hey", "What's up?", or waving hand.      │
 │    ceramic mug, paper cone cup, and tumbler  │ • Trained Response: Wiping table with cloth. │
 │    after being trained only on a plastic cup.│   Novel Untrained Responses: Wiping with     │
 │                                              │   paper towel, sponge, or sanitizing wipe.   │
 └──────────────────────────────────────────────┴──────────────────────────────────────────────┘

Stokes & Baer (1977) Programming Strategies for the QASP-S

1. Programming Common Stimuli: Bringing physical stimuli from the natural generalization setting into the training environment (e.g., bringing school cafeteria trays, lunchboxes, and ambient cafeteria audio into the clinic cubicle to teach lunchtime opening skills).
2. Teaching Sufficient Multiple Exemplars: Training across a wide variety of stimulus topographies before probing (e.g., teaching receptive identification of "car" using sedans, SUVs, sports cars, convertibles, and pickup trucks, rather than a single plastic toy car).
3. Training Loosely: Systematically varying non-critical incidental features of instruction (varying clinician tone of voice, lighting, seating positions, background noise, and wording of $S^D$s) so stimulus control does not become bound to irrelevant environmental cues.
4. Indiscriminable Contingencies: Moving from predictable reinforcement schedules to variable, intermittent schedules where the learner cannot predict which specific response will yield reinforcement, promoting sustained responding.

---

Maintenance Schedules & Reinforcement Thinning

During initial skill acquisition, the learner is typically supported by a Continuous Reinforcement Schedule (CRF / FR1), where every single correct response contacts immediate, high-magnitude reinforcement. However, the natural world does not operate on an FR1 schedule. If an acquisition program is terminated abruptly on an FR1 schedule, the sudden cessation of dense reinforcement triggers extinction-induced behavioral breakdown.

                          SCHEDULE THINNING TRAJECTORY
  ┌─────────────────────────────────────────────────────────────────────────────┐
  │ Acquisition Phase: Continuous Reinforcement (CRF / FR1)                     │
  │   Every unprompted correct response receives immediate premium reinforcer.  │
  │         │                                                                   │
  │         ▼ Step 1 Thinning: Fixed Ratio 2 / Variable Ratio 2 (VR2)           │
  │   Reinforcement delivered on average every 2 correct responses.             │
  │         │                                                                   │
  │         ▼ Step 2 Thinning: Variable Ratio 4 (VR4) / VR5                     │
  │   Reinforcement thinned; praise delivered intermittently; tokens used.      │
  │         │                                                                   │
  │         ▼ Step 3 Thinning: Natural Intermittent Contingencies (VR8-10)      │
  │   Behavior trapped in natural community contingencies and social validation.│
  └─────────────────────────────────────────────────────────────────────────────┘

Avoiding Ratio Strain

Ratio Strain is a behavioral phenomenon characterized by avoidance, aggression, fatigue, or significant response deceleration that occurs when the schedule of reinforcement is thinned too rapidly or when the behavioral demands become too challenging relative to the reinforcement available. If a technician transitions a learner directly from an FR1 schedule to an FR10 schedule, the child will likely engage in problem behavior or task refusal due to ratio strain. Thinning must be gradual, data-driven, and systematic.

---

Systematic Probe Schedules & Remediation of Skill Regression

Archiving a mastered goal into maintenance does not mean closing the binder forever. A QASP-S must establish a standardized Maintenance Probe Schedule:

Clinical Remediation Decision Tree for Skill Regression

If a learner falls below the 80% accuracy threshold on any maintenance probe:

1. Do not ignore the failure or mark it as an anomaly.
2. Immediate Booster Sessions: Re-introduce the target into active instructional programming for 1–2 weeks.
3. Re-establish Stimulus Control: Implement Errorless Learning with Most-to-Least (MTL) prompting or Constant Time Delay (CTD) to prevent the reinforcement of incorrect response chains.
4. Re-densify Reinforcement: Temporarily shift the reinforcement schedule back down (e.g., from VR5 to FR1) to rebuild behavioral momentum before re-thinning.
5. Audit Environmental Changes: Investigate whether setting events (e.g., medication changes, sleep disruption, illness) or changes in the natural environment contributed to the performance decrement.