11.3 Programming for Generalization and Maintenance: Stokes and Baer (1977) Tactics

Foundations of Generalization and Maintenance

In their historic 1968 paper defining the dimensions of applied behavior analysis, Donald Baer, Montrose Wolf, and Todd Risley established that a behavioral intervention has little practical value if the acquired behavior change occurs only in a sterile training clinic, in the presence of a single therapist, and disappears as soon as contrived reinforcement contingencies are terminated. The dimension of Generality dictates that behavior change must prove durable over time, appear in a wide variety of possible environments, and spread to a wide variety of related behaviors.

Foundational Definitions and Conceptual Distinctions

Precise behavioral terminology is essential for the BCaBA examination. Candidates must clearly distinguish between three interrelated yet functionally distinct constructs: Stimulus Generalization, Response Generalization, and Maintenance.

+-------------------------------------------------------------------------+
|                 THE THREE PILLARS OF BEHAVIORAL GENERALITY               |
+-------------------------------------------------------------------------+

  1. STIMULUS GENERALIZATION (Untrained Stimuli --> Trained Response)
     - Same Response Topography emitted across novel antecedent stimuli
     - Example: Taught to say "Dog" to a Golden Retriever picture -->
       Says "Dog" to a real Chihuahua, a Poodle, and a cartoon Dalmatian.

  2. RESPONSE GENERALIZATION / INDUCTION (Trained Stimulus --> Untrained Responses)
     - Novel, untrained response topographies functionally equivalent to target
     - Example: Taught to say "Hello" to greet peers -->
       Spontaneously waves, says "What's up?", and gives fist bumps.

  3. MAINTENANCE (Durability Across Time)
     - Trained response persists over time after intervention contingencies fade
     - Example: Client continues to brush teeth independently 6 months after
       token economy and therapist prompting have been completely removed.
+-------------------------------------------------------------------------+

1. Stimulus Generalization

Stimulus Generalization occurs when an operant behavior that has been reinforced in the presence of a specific discriminative stimulus ($S^D$) is emitted in the presence of novel, untrained stimulus conditions that share formal, physical, or functional properties with the instructional $S^D$.

• Key Rule: The response topography remains constant, while the antecedent stimulus conditions vary (untrained instructors, novel physical settings, varied materials, different vocal phrasing).
• Clinical Example: A child is taught in a clinic therapy room by a female behavior technician to point to a red cup when given the vocal instruction, "Touch red cup." If the child immediately points to a red ceramic mug at home when their father says, "Show me the red cup," stimulus generalization has occurred across people, settings, materials, and vocal stimuli.
• Contrast with Stimulus Discrimination: Stimulus discrimination represents the opposite behavioral process—tight stimulus control where the response is emitted only in the presence of a specific $S^D$ and is completely absent in the presence of all $S^\Delta$ conditions.

2. Response Generalization (Response Induction)

Response Generalization—historically termed response induction by B.F. Skinner—occurs when a learner emits untrained, topographically novel responses that are functionally equivalent to the trained target behavior.

• Key Rule: The antecedent stimulus condition remains constant, while the response topography varies to produce the same functional consequence.
• Clinical Example: A student is taught to open a cardboard packing box using a box-cutter utility knife. When presented with a sealed packing box without a box cutter, the student uses a brass house key, the edge of a pen, or firmly peels the packing tape with their fingers to open the box. Because the key, pen, and finger-peeling topographies were never directly taught or reinforced, their emergence represents response generalization.
• Contrast with Response Differentiation: Response differentiation is the narrowing of response topographies through differential reinforcement, where only one specific form of the response produces reinforcement while all other topographies are extinguished.

3. Maintenance (Durability Over Time)

Maintenance refers to the extent to which a learner continues to perform the target behavior after the formal behavioral intervention package, contrived prompting, and artificial reinforcement schedules have been systematically faded or entirely withdrawn.

• Key Rule: Maintenance measures temporal durability—does the behavior persist weeks, months, or years after clinical discharge?
• Clinical Mechanism: Long-term maintenance relies upon successfully transferring stimulus control from contrived, artificial contingencies (e.g., token boards, therapist praise, continuous edible delivery) to natural contingencies of reinforcement existing within the client's everyday social and physical environment.

---

The Stokes and Baer (1977) Generalization Framework

In their seminal paper, "An implicit technology of generalization" (Journal of Applied Behavior Analysis, 1977), Trevor Stokes and Donald Baer reviewed the first decade of published ABA research. They arrived at a profound and challenging conclusion: applied behavior analysts were frequently failing to actively program for generalization.

The Critique of "Train and Hope"

Stokes and Baer identified that researchers routinely taught a target behavior in a single, controlled environment, administered no generalization programming, and merely administered probes in natural environments while "hoping" the behavior would spontaneously generalize.

Stokes and Baer sorted the generalization research into nine categories: train and hope, sequential modification, introduce to natural maintaining contingencies, train sufficient exemplars, train loosely, use indiscriminable contingencies, program common stimuli, mediate generalization, and train "to generalize." Train and hope is the non-strategy; the other eight are active tactics. Seven are analyzed below, and introducing natural maintaining contingencies is covered with behavior traps later in this section. These tactics operate by manipulating instructional stimuli, response variations, and reinforcement contingencies.

+-------------------------------------------------------------------------+
|               STOKES & BAER (1977) ACTIVE GENERALIZATION TACTICS        |
+-------------------------------------------------------------------------+
  1. Sequential Modification    --> Train in Setting A; probe Setting B; if
                                    it fails, systematically train in B.
  2. Train Sufficient Exemplars --> Teach across multiple stimuli & responses
                                    (General Case Analysis).
  3. Program Common Stimuli     --> Bring salient natural stimuli into the
                                    training setting (cash register, noise).
  4. Train Loosely              --> Systematically vary non-critical stimulus
                                    dimensions (lighting, tone, seating).
  5. Indiscriminable            --> Thin to variable (VR/VI) schedules or
     Contingencies                  delayed rewards to prevent predictability.
  6. Mediate Generalization     --> Teach a portable stimulus (checklist, cue
                                    card) client carries into natural settings.
  7. Train to Generalize        --> Directly reinforce response variability
                                    (Lag Reinforcement Schedules).
+-------------------------------------------------------------------------+

Detailed Analysis of Seven Active Tactics

1. Sequential Modification

• Operational Protocol: The clinician teaches the target behavior in the primary instructional setting (e.g., clinic therapy room). Generalization probes are conducted in secondary natural settings (e.g., general education classroom, home). If the behavior fails to generalize, the clinician sequentially introduces the formal intervention package into the second setting, continues instruction until criterion is reached, probes a third setting, and repeats this process until the behavior occurs reliably across all desired environments.
• Behavioral Mechanism: Applying multiple baseline design logic as a clinical programming technique. While effective, sequential modification is resource-intensive because it essentially represents repeated direct training across multiple environments rather than true generative generalization.

2. Train Sufficient Exemplars (and General Case Analysis)

• Operational Protocol: Rather than teaching with a single flashcard, a single instructor, or a single prompt phrasing, the clinician trains the behavior across multiple stimulus exemplars (different people, materials, settings) and teaches multiple response variations until general case responding emerges.
• General Case Analysis (Sprague & Horner, 1984): A systematic, empirical method for selecting teaching exemplars. The behavior analyst conducts a comprehensive environmental survey to identify the full universe of stimulus variations and response requirements present in the natural generalization setting. The clinician then selects a minimal subset of teaching exemplars that sample the entire range of variation.
  • Applied Example: Vending Machines: To teach a client to use community vending machines, the clinician does not train on a single machine. The clinician identifies that vending machines vary across coin slots, dollar bill acceptors, credit card touchpads, physical pull-knobs, and digital keypads. The clinician selects 3 specific machines that collectively encompass all these variations for direct instruction. Once the client masters these sufficient exemplars, they can independently operate any untrained vending machine in the community.

3. Program Common Stimuli

• Operational Protocol: The behavior analyst identifies salient, distinctive physical, social, or auditory stimuli that naturally exist in the generalization environment and deliberately incorporates those exact stimuli into the instructional training setting.
• Behavioral Mechanism: By establishing stimulus control in the presence of stimuli that are guaranteed to be present in the generalization setting, stimulus control transfers seamlessly from the clinic to the community.
• Applied Clinical Examples:
  • Vocational Preparation: Bringing an authentic supermarket barcode scanner, conveyor belt, shopping cart, and grocery items into a high school special education classroom rather than practicing sorting plastic toy fruit.
  • Pediatric Social Skills: Playing ambient, pre-recorded school cafeteria background noise (70 dB chatter, clattering trays) in the clinic therapy room while teaching peer conversational initiation.
  • Community Mobility: Bringing authentic municipal bus schedules, actual token fare boxes, and authentic street crossing signs into the training facility.

4. Train Loosely

• Operational Protocol: The clinician deliberately, unsystematically varies non-critical, incidental stimulus dimensions in the instructional environment during skill acquisition.
• Behavioral Mechanism: When instruction is conducted under hyper-rigid, invariant conditions (e.g., same quiet room, same therapist sitting in the same chair at a 90-degree angle, same desk, same lighting, exact same vocal intonation), the learner's behavior acquires overly narrow, hyper-specific stimulus control. The behavior becomes tightly bound to those incidental environmental features and collapses when any feature changes. Training loosely prevents this rigidity.
• Variables to Alter During Loose Training:
  • Vary the instructor's vocal tone, volume, and phrasing (e.g., "Touch blue", "Point to blue", "Where is blue?", "Show me blue").
  • Vary instructor seating position (standing, sitting across, sitting beside, behind).
  • Vary the physical setting (carpet, desk, hallway, cafeteria, outdoors).
  • Vary incidental stimuli (different table colors, background music playing, doors open vs. closed, daylight vs. fluorescent lights).

5. Indiscriminable Contingencies

• Operational Protocol: The clinician structures reinforcement contingencies such that the learner cannot discriminate or predict whether any specific response will produce reinforcement.
• Behavioral Mechanism: Under continuous reinforcement (FR1), the schedule is highly discriminable: every single response produces a reinforcer. When an individual transitioning to the natural environment encounters zero reinforcement for 3 or 4 responses, the abrupt contrast is instantly discriminated as extinction, causing responding to halt. By fading reinforcement to intermittent, unpredictable schedules (Variable Ratio [VR] and Variable Interval [VI]) or delayed reinforcement, the learner cannot distinguish between instructional trials and natural extinction, resulting in remarkably high, steady rates of durable responding.
• Applied Clinical Examples:
  • Variable Schedules: Thinning conversational initiation reinforcement from FR1 to VR3, then VR7.
  • Intermittent Prize Lotteries / Secret Agent Games: Informing a classroom that a timer will ring at unpredictable variable intervals throughout the day, and whoever is actively on-task when the timer rings earns a token (VI schedule).
  • Delayed Reinforcement: Delivering reinforcers at the end of the day or week based on unpredictable sampling of behavioral records.

6. Mediate Generalization

• Operational Protocol: The clinician establishes a portable stimulus or mediator response during training that the learner physically carries or cognitively transports into the natural generalization environment to prompt the target behavior.
• Behavioral Mechanism: The portable stimulus serves as an easily transportable discriminative stimulus ($S^D$) or conditioned prompt, bridging the gap between the training clinic and novel community settings.
• Applied Clinical Examples:
  • Visual Checklists & Graphic Organizers: A laminated visual task analysis checklist of morning vocational hygiene (deodorant, teeth, hair, clean shirt) that the client checks off in their home bathroom.
  • Covert Verbal Self-Instruction (Self-Talk): Teaching an individual encountering a difficult social conflict to silently repeat the vocal script: "Stop, take a deep breath, hands in pockets, ask for help."
  • Wearable Prompt Technology: A vibrating smart-watch app that delivers a tactile prompt every 15 minutes, prompting the client to check their posture, self-monitor work engagement, or initiate social greetings.

7. Train to Generalize (and Lag Schedules of Reinforcement)

• Operational Protocol: The clinician treats generalization and response variability directly as an operant response class and explicitly reinforces novel, generalized, or varied responses.
• Behavioral Mechanism: Reinforcing behavioral variability demonstrates that variability itself is an operant governed by reinforcement contingencies (Page & Neuringer, 1985).
• Lag Schedules of Reinforcement:
  • A Lag $n$ Schedule delivers reinforcement only if the emitted response is topographically different from the previous $n$ responses emitted by the client.
  • Lag 1 Schedule: The client is reinforced only if their response differs from the immediately preceding response.
    • Example (Block Building / Play): If a child stacks 3 red blocks, repeating that action produces no reinforcement. If the child builds a bridge with blue blocks (different from the last action), reinforcement is delivered.
  • Lag 2 Schedule: The response must differ from the previous two responses.
    • Example (Intraverbal Responses): When asked "What are some animals?", saying "Dog" is reinforced. Saying "Cat" is reinforced. Saying "Dog" again is not reinforced because it matches a response emitted within the last 2 trials.
  • Lag schedules are exceptionally powerful for building complex verbal repertoires, creative toy play, conversational flexibility, and eliminating repetitive, perseverative responding.

---

Natural Contingencies of Reinforcement and Behavioral Traps

Natural Maintaining Contingencies and Behavior Traps (Baer & Wolf, 1970)

The ultimate goal of programming for maintenance is to transfer behavioral control from contrived reinforcement contingencies to naturally occurring contingencies of reinforcement. When a behavior is successfully transferred to natural reinforcers, the clinician can completely fade clinical intervention while the behavior continues indefinitely.

Baer and Wolf (1970) introduced the behavior trap to describe an exceptionally potent, self-sustaining maintenance technology, and it is the clearest example of Stokes and Baer's tactic of introducing behavior to natural maintaining contingencies. A behavioral trap is a behavioral contingency that is easy to enter but difficult to leave, naturally sustaining and expanding a behavioral repertoire.

+-------------------------------------------------------------------------+
|                     THE FOUR STEPS OF A BEHAVIORAL TRAP                 |
+-------------------------------------------------------------------------+
  Step 1: Bait the Trap
          --> Identify virtually irresistible natural reinforcers that
              naturally motivate the client (e.g., peer games, trading cards).
                                    |
                                    v
  Step 2: Ensure a Low-Effort Entry Response (Already Learned)
          --> Teach simple, foundational entry responses that require
              minimal physical effort (e.g., rolling a ball, passing a card).
                                    |
                                    v
  Step 3: Spring the Trap
          --> Guide the learner to emit the low-effort response in the
              natural environment where dense natural contingencies operate.
                                    |
                                    v
  Step 4: Natural Reinforcement Traps & Sustains the Behavior
          --> Peers deliver natural social praise, laughter, and inclusion.
          --> The behavior expands into complex repertoires WITHOUT contrived
              therapist reinforcement! The learner is 'trapped' in learning.
+-------------------------------------------------------------------------+

Cooper, Heron, and Heward (2020) summarize four essential features of a behavior trap: (a) virtually irresistible reinforcers "bait" the learner into the trap; (b) only a low-effort response already in the learner's repertoire is needed to enter; (c) interrelated contingencies inside the trap motivate the learner to acquire, extend, and maintain targeted skills; and (d) traps stay effective for a long time because they show few satiation effects.

Applied Example: Playground Social Inclusion

A 6-year-old child with autism is isolated on the school playground. Contrived edible reinforcement delivered by an adult paraprofessional stigmatizes the child and does not maintain. The behavior analyst identifies that peers on the playground play a simple four-square ball game (the "trap"). The analyst teaches the child the foundational entry response of catching and bouncing a rubber ball. The analyst then prompts the child into the four-square line. Once the child bounces the ball into a square, peer cheering, turn-taking, and active play take over. Contrived adult reinforcement is faded to zero. The natural social reinforcers generated by peers maintain the child's playground engagement indefinitely.

---

Stokes and Baer (1977) Generalization Tactics Matrix

The following structured matrix compares the core behavioral mechanisms, implementation protocols, clinical examples, and failure modes across the Stokes and Baer generalization framework:

---

Common BCaBA Exam Traps: Generalization & Maintenance

• Trap 1: Confusing Stimulus Generalization with Response Generalization: This is among the most frequently missed conceptual distinctions on the certification exam. When reading a scenario, immediately ask: Did the stimulus change, or did the behavior change?
  • If the client emits the exact same behavior across novel people, settings, or materials, it is Stimulus Generalization.
  • If the antecedent stimulus remains the same, but the client emits novel, untrained topographies of behavior that achieve the same functional outcome, it is Response Generalization (Response Induction).
• Trap 2: Endorsing "Train and Hope" as an Acceptable Clinical Option: Multiple-choice distractors often describe training a skill in an analog therapy room and administering post-tests in a classroom without any proactive programming, labeling this "standard practice." The correct behavioral answer will always identify "Train and Hope" as an unscientific failure to program actively for generalization.
• Trap 3: Conflating "Program Common Stimuli" with "Train Loosely":
  • In Program Common Stimuli, the clinician intentionally brings specific, salient, critical stimuli from the natural generalization environment into the training clinic (e.g., bringing the real store register into the classroom).
  • In Train Loosely, the clinician intentionally varies incidental, non-critical stimulus features during training (e.g., changing teacher voice volume, desk position, lighting) so the student does not get "stuck" on rigid cues.
• Trap 4: Maintaining Continuous (FR1) Reinforcement During Maintenance Phases: When an exam question asks how to program for long-term maintenance of a mastered skill in a natural environment, options proposing continuous FR1 schedules are incorrect. A continuous schedule makes extinction immediately discriminable in the natural world. Maintenance demands systematic thinning to intermittent variable schedules (VR/VI) and reliance on natural contingencies.