Baselines and within-person comparison are required

What it is

Skin-conductance level, resting heart rate, and pupil diameter can differ across people by as much as a typical task effect. Absolute readings mainly reflect skin, vasculature, sensors, and that day’s state, not what this interface did to “people in general.” The usable comparison is how far the same person moved relative to their own baseline, and how that person changed across conditions. Averaging raw microsiemens across people compares whose hands were damper, not which interface was more activating.

Why it happens

Peripheral signals have a slow tonic level and phasic waves on top. Tonic level is dominated by temperature, hydration, time of day, drugs, and electrode placement; it differs widely across people and can drift within a day. Tasks usually show up in the phasic component. Subtracting a pre-task rest, or referencing that person’s median for the session, is what lifts the phasic change out of tonic level. A between-subjects design that uses raw values confounds treatment with who was assigned which constitution. Even with random assignment, individual difference inflates variance, and small-sample interface studies readily treat constitution as a result.

Studying it

Give every session a rest or a neutral task as baseline, long enough to estimate tonic level, with posture and luminance matched to the later task. The primary dependent measure should be change from baseline, a standardized score, or a within-person condition contrast. Counterbalance order, because the baseline itself drifts. Report baseline length, exclusion rules, and how many people were dropped for an unstable baseline. If a between-subjects contrast is unavoidable, correct for baseline first and check that groups already differed on it.

Where it stops holding

Settings with no stable rest—continuous driving, surgery, field patrol—need a sliding reference or brief pauses between tasks, not a fictional laboratory rest. Some diagnostic questions care about absolute level; that is not interface evaluation. Baseline correction does not repair motion artifact already in the signal, and it does not give arousal a valence. For very brief stimuli a phasic wave may never form, and the corrected series is still noise. Tonic fluctuation is larger in children and older adults, who need longer or more frequent references.

Applying it

• Fix the protocol: one to two minutes sitting in the same posture before the task, with no stimulus and no chat.
• Default plots to change-from-baseline; if raw values appear, show each person’s baseline too.
• Do not rank two products on cross-person mean heart rate or skin conductance.
• On repeated measures in one day, insert short references between conditions; if drift appears, stop comparing later conditions to the opening baseline.

Related

• Same group: Q3.10.1 Physiological signals index arousal, not valence · Q3.10.2 Motion and environment contaminate the signal
• Adjacent: Q3.18 Experimental design and controls · Q3.09 Eye-tracking experiments
• Search terms: within-subject baseline · tonic phasic EDA · physiological individual differences