Alternating-treatments design

Rapid alternation creates repeated comparisons

An alternating-treatments design, or ATD, may compare two interventions, an intervention with business as usual, or another pair of defined conditions. Conditions switch frequently across sessions, activities, or other planned opportunities. A baseline phase can precede alternation, though an ATD can also begin with rapidly alternating conditions.

The WWC Single-Case Design Technical Documentation treats structured repetition as the basis for single-case inference. Within an ATD, repeated nearby observations under each condition allow the analyst to ask whether one data path separates from another as conditions alternate.

Every comparison is its own contrast. If a design includes conditions A, B, and C, results for A versus B, A versus C, and B versus C need separate interpretation. A difference between A and B cannot answer whether B differs from C.

Plan order and exposure before data collection

Balance how often each condition appears and where it occurs in the sequence. Random assignment of condition order can reduce systematic order bias when properly planned. Counterbalancing can spread morning, afternoon, first-session, fatigue, partner, or activity effects across conditions. Neither method eliminates all confounding.

Avoid long runs of one condition inside the alternating phase. The current WWC Procedures and Standards Handbook, Version 5.0, sets research-rating rules for condition observations and contiguous points. Its highest rating requires at least five data points per condition; its with-reservations rating requires at least four. It permits no more than two consecutive points from the same condition without interruption by another condition.

Those counts define WWC review eligibility. Clinical sufficiency also depends on the question, data pattern, measurement quality, risk, feasibility, and predetermined decision rules. Five points cannot rescue biased ordering, poor fidelity, or an unsuitable outcome.

Make the active condition easy to discriminate

The participant, implementer, and observer should know which condition is active when their role requires it. Use condition-correlated cues such as a clearly labeled material, setting, partner, or schedule entry, while holding irrelevant features as constant as feasible. Explain cues in an accessible form and confirm that the person can use them.

Condition discrimination carries a tradeoff. Distinct cues may help prevent blending, yet a room, partner, color, or device can also become part of the intervention package. Report those features. A claim about a procedure alone is too narrow when the procedure always arrives with a different partner or setting.

Carryover can erase or imitate differentiation

The original Barlow and Hayes paper identifies sequential confounding, carryover effects, and alternation effects as central ATD concerns. Carryover occurs when one condition changes responding during a later condition. Alternation effects occur when frequent switching itself affects behavior.

Use ATD when effects are expected to occur quickly and dissipate before the next comparison. Consider spacing, washout periods, separate stimulus sets, or another design when effects last. Skill acquisition, medication, fatigue, emotional distress, satiation, and durable environmental changes can carry forward. If washout cannot be established safely, rapid alternation may produce an uninterpretable or unacceptable test.

An adapted ATD can assign separate but equivalent stimulus sets to different teaching methods. That reduces direct transfer across conditions, while set difficulty becomes a new threat. Match sets before assignment, counterbalance assignments across cases when possible, and state that results apply to each method-and-set package unless further evidence separates them.

Measure outcome and implementation separately

Define the outcome, eligible opportunity, scoring window, session length, prompts, exclusions, and missing-data rule before the comparison. Graph every observation by condition in time order. Inspect each path's level, trend, variability, overlap, immediacy, and consistency across blocks.

A review of ATD quantitative and graphical methods describes randomization, counterbalancing, time-series graphs, and supplemental analyses. A condition mean or percentage can summarize data, but the sequence and within-condition variability remain essential.

Measure implementation for every condition. Report whether the correct procedure, cue, dose, partner action, and timing occurred. Train and calibrate observers and sample interobserver agreement when scores drive clinical conclusions. Keep agreement, treatment integrity, and the outcome as separate measures.

The BACB BCBA Test Content Outline, 6th edition covers single-case design features, interpretation, and multielement designs. It is examination content rather than a procedure manual.

Fit, choice, and safety govern the comparison

Compare only conditions that are lawful, clinically appropriate, and acceptable for exposure. Preserve AAC, food, water, bathroom access, mobility, prescribed care, pain care, rest, emergency help, and effective safety protections across every condition. Avoid arranging distress merely to create separated data paths.

The current BACB Ethics Code applies to BCBA and BCaBA certificants and people who completed an application. It addresses competence, understandable communication, client involvement, informed consent and assent when applicable, assessment, intervention, risk, data, and evaluation. Pause or end comparison when consent is withdrawn, assent is withdrawn when applicable, risk changes, an adverse effect appears, or one option is clearly unacceptable.

A fictional ATD example

Maya is a fictional adult who already completes a five-step packing routine and wants to compare two accessible supports: a paper checklist and a digital checklist. She selects both formats as acceptable. Twelve matched opportunities are scheduled in six two-session blocks. Order within each block is randomized, so each format appears once per block. The routine, materials, partner, session length, and scoring definition stay constant.

With paper, Maya independently completes 3, 4, 4, 3, 4, and 3 steps. With digital, she completes 5, 5, 4, 5, 5, and 5. Digital produces at least four correct steps in 6 of 6 opportunities; paper does so in 3 of 6. Both conditions have six scheduled observations, and none disappears from the denominator.

The separated paths support choosing digital for this routine if fidelity, observer agreement, carryover, and Maya's preference also support it. The comparison establishes no universal superiority of digital checklists. It also cannot isolate the device from its layout, feedback, or other package features. A follow-up preference and everyday-use check addresses fit beyond the experimental outcome.