BCBA Domain G: Behavior-Change Procedures

Differential reinforcement is the most heavily sampled procedure in Domain G, and the exam leans on you confusing its variants. Every variant reinforces something so that the target behavior contacts less reinforcement — the difference is what you reinforce. DRA reinforces a specific alternative (often a functional communication response) that earns the same reinforcer the problem behavior used to. DRI reinforces an alternative that is physically incompatible with the target. DRO reinforces the absence of the target across an interval. DRL reinforces responding at or below a set low rate. DRH does the opposite — it builds a behavior up to a high rate.

Two numbers-driven traps recur. First, when a behavior is severe and maintained by automatic reinforcement, the exam often prefers DRI — reinforcing a topographically incompatible response (hands flat on the table while the target is head-punching) buys immediate physical safety that teaching a new mand does not. Second, DRO intervals are set from the data, not by feel: if baseline shows a behavior 15 times in a 60-minute session, the mean inter-response time is about 4 minutes, so the initial DRO interval should sit just below that mean — start it above the baseline rate and the learner almost never earns the reinforcer, and the behavior holds or climbs.

To build behavior the exam wants three procedures kept distinct: shaping, chaining, and prompting with fading. Shaping reinforces successive approximations of a single response you cannot yet get. Chaining links an already-analyzed sequence of discrete steps — a task analysis — into one routine. The classic swap is a hand-washing item: a learner turns on the water but struggles with the ordered steps of soap, rub, rinse, dry. That is a sequence of distinct behaviors, so the answer is to switch from shaping to chaining; reinforcing closer approximations of the whole routine is the wrong tool for a task with a fixed step order.

Prompting and fading is where prompt dependence lives. A plan that says “use partial physical prompts and fade as tolerated” will strand a learner on prompts for weeks, because “as tolerated” is not a criterion — it is a feeling. The fix the exam rewards is objective, data-based criteria for moving to the next fading level, not a more powerful reinforcer or an abrupt switch of prompt hierarchy. Whatever you build, plan for generalization and maintenance from the start — program common stimuli, train loosely, and thin reinforcement toward natural contingencies so the skill survives outside the teaching setting.

Token economies are a favorite because they fail in instructive ways, and the exam tests whether you can name the exact failure. Tokens are generalized conditioned reinforcers — their strength is entirely borrowed from the backup reinforcers they buy. Two failure modes recur. When the backup store stocks only things the client already gets for free, the tokens stop working even though the client still likes holding them: the backup reinforcers lack value, so the tokens have nothing to be conditioned against. When residents accumulate hundreds of unspent tokens while prices stay flat, responding drops because of satiation on the generalized conditioned reinforcer itself — too easy to earn, too little reason to spend.

Read those two apart carefully, because the exam offers them as near-neighbors. A dead backup store is a backup-value problem; a pile of unspent tokens is a satiation problem; and neither is “the tokens lost their conditioned properties” or “ratio strain,” which are the distractors written to catch a fast reader.

On the reduction side, extinction and punishment carry the domain, and both come with a signature trap. Extinction means withholding the reinforcer that maintained the behavior — so it is function-specific. When you stop letting toy-throwing earn a parent finishing the chore and instead guide the child through it, that is escape extinction, a form of extinction, not punishment. Expect an extinction burst early: a client's attention-maintained screaming spikes in frequency, duration and volume in the first days before dropping toward zero. A separate, brief return of the behavior weeks later, after it had stopped, is spontaneous recovery — not a second burst. The exam pairs those two to see if you tie each to its cause and its timing.

Punishment is scored under a least-restrictive-alternative rule, and this is where technically-correct answers get eliminated. Removing an earned token contingent on spitting is response cost — negative punishment (you take a stimulus away), not positive punishment; distinguishing the two is a routine item. And when a caregiver pushes for an exclusionary time-out to suppress severe head-banging whose function is still ambiguous, the exam's answer is not to run the restrictive procedure faster — it is to clarify the function first (a brief functional analysis, done safely), because a least-restrictive, function-based plan cannot be built on an unknown function.