Behavior Assessment and Functional Analysis
This domain tests the assessment hierarchy: which method answers which question, how to interpret a differentiated versus undifferentiated outcome, and how to weigh risk during experimental manipulation. Items often present a data pattern and ask for the functional hypothesis, or ask which assessment is appropriate given setting constraints.
The assessment hierarchy
- Indirect: FAI, QABF, MAS, FAST. Fast and broad, but subjective and vulnerable to rater bias. They generate hypotheses only.
- Descriptive: ABC narrative and structured recording, scatterplots, conditional and lag probabilities from sequential data. Correlational - high conditional probability is not proof of function.
- Experimental: functional analysis manipulates antecedents and consequences, usually in a multielement design. Only this level establishes a functional relation.
Standard FA conditions
- Play/control (attention freely available, demands removed), attention (contingent attention), escape/demand (contingent break), tangible (contingent item access), alone/ignore (tests automatic reinforcement).
- Sessions are commonly 5–15 minutes with fixed reinforcer-delivery intervals (often 30 s); differentiation means consistently higher rates in one test condition with clear separation from control.
- Masking and carryover are design threats. Undifferentiated results suggest latency-based FA, brief FA, trial-based FA in classrooms, or a synthesized contingency (IISCA).
From data to plan
- The identified function selects the intervention, and the FA condition becomes the teaching context for the alternative response.
- Screen medical and biological variables (sleep, pain, medication change) before concluding "automatic."
- Pre-specify risk management: protective equipment, medical clearance, stopping criteria, and consent when escape extinction may occasion severe behavior.
Worked example. A student screams. ABC data suggest loud noise precedes screaming, but scream-contingent staff attention follows. A trial-based FA with 2-minute classroom trials shows screaming in 0% of control trials and 80% of attention trials. Interpretation: the noise correlation was incidental; the function is attention. Plan: FCT teaching a mand for attention, extinction of scream-contingent attention, then schedule thinning.
Traps that cost marks. Reporting an undifferentiated FA as "automatic." Treating a high conditional probability as proof. Running escape extinction with no mand available. Forgetting that the alone condition needs a control comparison before you can claim automatic reinforcement.