Double-blind Multiple Baseline Design

The core replication logic is identical: the intervention is introduced sequentially across independent baselines to demonstrate a causal effect. The difference is that in the double-blind variant, outcome assessors are kept unaware of which baseline is currently receiving the intervention, eliminating observer expectancy bias from the measurements. The double-blind procedure strengthens measurement validity without altering the design's causal logic.

A minimum of two baselines is required, but three or four are recommended. Each additional baseline provides an additional replication of the treatment effect, strengthening causal inference. With only two baselines the evidence for causality is weaker — a chance coincidence cannot be ruled out as confidently as with three or more replications.

Baselines are independent when a change in one does not cause a change in another. In practice, this means selecting behaviors that are functionally separate (not in the same response class), individuals who do not share the same immediate environment, or settings that do not share common antecedents. Pilot observation or logical analysis of the behaviors/settings should confirm independence before the study begins.

No — this is precisely what the double-blind procedure prevents. Treatment administrators must be separate from outcome assessors. If this separation is impossible logistically, the design loses its double-blind property and reverts to a standard multiple baseline design, which should be reported accurately.

Visual analysis of graphed time-series data is the primary method in the single-subject tradition. Supplementary effect-size statistics designed for single-case data — such as Tau-U, nonoverlap of all pairs (NAP), or the percentage of nonoverlapping data (PND) — can quantify effect magnitude and support meta-analytic aggregation across studies, but they do not replace visual analysis.