BJC common mistakes
BJC · Understand the traps and how to avoid them.
Confusing negative reinforcement with punishment because both involve something being removed; selecting 'punishment' whenever a consequence is described as taking something away.
The learner memorizes the labels by surface wording ('add' vs. 'remove') rather than by the effect on response frequency, and the exam deliberately writes consequences that sound aversive but function as reinforcement.
What to do instead
Before answering, write the contingency in the form 'behavior -> consequence -> future frequency.' Classify only by the frequency arrow: increase = reinforcement, decrease = punishment; then use add/remove to label positive/negative. Practice sorting 20 mixed scenarios by arrow first, label second.
Treating motivating operations as identical to discriminative stimuli, so a question about deprivation or satiation is answered with 'SD' or vice versa.
Both are antecedent events that precede behavior, and both are described as 'setting the occasion' in casual study materials, so the learner never builds a functional distinction.
What to do instead
Memorize the two separate effects: an MO alters the value of a consequence (value-altering) and the momentary frequency of the behavior (behavior-altering); an SD signals the availability of a consequence based on a history of differential reinforcement. For each practice item, ask 'does this change how much the consequence is wanted, or does it signal that it is available?'
Reporting a mean or percentage of intervals as the sole measure of a behavior that occurs in bursts, then drawing a trend conclusion from it.
Summary statistics are easy to compute and feel rigorous, but they mask variability, and candidates under time pressure skip the graph.
What to do instead
Always inspect the data path first: level, trend, variability. Report a measure appropriate to the dimension of interest (rate for count-per-time, duration for temporal extent, latency for time to onset, IRT for temporal locus). State explicitly that a mean cannot show trend, and pair any summary number with a statement about variability.
Interpreting a high or low interobserver agreement percentage as proof the data are accurate, or accepting IOA collected during a condition in which the observers were not independent.
IOA is taught as the gold standard for believability, so candidates overgeneralize it to validity and miss the procedural requirements for how agreement data must be collected.
What to do instead
State the limits in your own words: IOA estimates consistency between observers, not correctness of the definition. Check that observers recorded independently, used the same response definition and measurement system, and that IOA was sampled across conditions and observers. If any of these fail, flag the data as questionable rather than 'reliable.'
Choosing a multiple-baseline design and then failing to verify the question's description of baseline stability, number of tiers, or the absence of concurrent changes in untrained tiers.
Candidates pattern-match 'multiple baselines are used when withdrawal is inappropriate' and stop there, without checking the design's internal validity requirements.
What to do instead
For any design question, run a checklist: (1) is there a stable baseline in each tier, (2) does the independent variable change at a different point in each tier, (3) do untreated tiers remain unchanged, (4) are there at least three tiers or participants. If a criterion fails, the design cannot support the experimental claim, regardless of how the results look.
Selecting a punishment or extinction procedure as the default for a severe behavior without first confirming a function-based reinforcement procedure has been considered.
Aversive and restrictive procedures are memorable and appear decisive, and candidates equate 'effective' with 'fastest,' overlooking the least-restrictive and reinforcement-first standards.
What to do instead
Before choosing any reductive procedure, state the hypothesized function and the reinforcement-based alternative that addresses it (DRA, FCT, NCR, extinction of the maintaining contingency). Justify a punishment component only as an adjunct when reinforcement alone is insufficient, and note the required safeguards: informed consent, monitoring, and a plan to fade.
Concluding a function from a single elevated category on an indirect rating scale without corroborating it with direct observation or a functional analysis.
Indirect measures are quick and produce tidy numbers, and candidates treat the highest bar as a diagnosis rather than as a hypothesis to be tested.
What to do instead
Treat indirect assessment as hypothesis generation only. Triangulate with direct descriptive data (ABC recording, conditional probability) and, where feasible and safe, a functional analysis or trial-based test. In your answer, name the convergent evidence and explicitly label the function as tentative when only indirect data exist.
Writing a behavior goal that names a label or a broad construct ('reduce aggression,' 'improve social skills') instead of an observable, measurable response definition.
Clinical language is habitual, and candidates draft goals that sound professional without checking whether two observers could independently score the same behavior.
What to do instead
Apply the dead-man's test and the observability test to every goal: could a dead person do it (if yes, it is not a behavior), and can it be seen and counted? Rewrite each goal to include the response topography, the context, and the criterion (e.g., 'hits a peer with an open hand during unstructured recess'). Then verify with a stranger-reader check.
Continuing an intervention or accepting consent without checking whether the treatment is producing socially significant change, or failing to specify who can consent for a particular client.
Momentum and rapport make it uncomfortable to question an ongoing plan, and candidates conflate assent, consent, and administrative approval.
What to do instead
Build a standing review checkpoint into every plan: predetermined decision rules for continuing, modifying, or discontinuing, tied to data and stakeholder feedback. For consent, identify the legally authorized representative, document capacity and assent where applicable, and verify that the plan is the least restrictive effective option. State the rule, not just the intention to 'monitor.'
Giving supervisees feedback that evaluates them personally or only after a summative failure, rather than delivering behavior-specific, timely feedback tied to observed performance and documented competencies.
Supervision is often learned by imitation of a mentor, and candidates reproduce vague praise or criticism without a measurement structure.
What to do instead
Use a written supervision contract with defined competencies, direct observation of the supervisee's performance, and feedback delivered as soon as possible after the observation: describe the specific behavior, its effect, and the alternative. Document the observation, the feedback, and the follow-up check. Track cumulative hours and competency sign-offs against the contract rather than relying on recall.