Motion and environment contaminate the signal
Aliases: EDA artifact · movement artifact · environmental contamination
What it is
Sensors on skin, fingers, or chest record voltage, impedance, or photoplethysmography, not “emotion.” Limb motion, speech, room temperature, light, loose electrodes, and device vibration all leave waves in the same channel that look like responses. Clicks, scrolls, head turns, and talking in interface studies are themselves motion. A physiological difference that has not had this contamination marked cannot be treated as task-evoked autonomic change.
Why it happens
Electrodermal recording needs stable skin contact and local sweat glands; a finger movement changes electrode pressure and contact area and the baseline jumps. Photoplethysmography depends on capillary volume; raising a hand or tightening a mouse grip rewrites the waveform. EEG is sensitive to blinks, jaw clench, and mains hum. Speech changes breathing and the thorax, so heart-rate variability is dragged by the cadence of talk. Air conditioning, coffee, and someone walking through the room change real physiology as well—environment-evoked, not interface-evoked. When contamination and the stimulus overlap in time, amplitude alone rarely separates them, because both can produce steep rises.
Studying it
Record a motion channel in parallel (accelerometry, video of hands and head, key and pointer events) and mark speech intervals, tagging physiological peaks that co-occur with movement as artifact candidates. Design trials so large movements do not overlap the critical window: no speech, a held posture, constant screen luminance. Analyze segments cleaned by an artifact rule and report the fraction dropped; if drops are systematically higher in one condition, that condition’s effect is not trustworthy. Control conditions should match movement—for example the same scrolling without the critical content—to isolate the physiological cost of the action itself.
Where it stops holding
A fully still laboratory lowers contamination and also lowers ecological validity. Wearable field work almost inevitably includes motion; conclusions should be limited to “the mixed signal observed during these activities,” not to pure autonomic response. When the research question is bodily load during movement, motion is the object rather than contamination, and channels should target muscle and metabolism rather than treating skin conductance as load. Medical-grade hardware is not immune to contact change; it only moves the threshold.
Applying it
- Write into the protocol that the critical window forbids changing hands, speaking, or standing; mark motion outside that window separately.
- Overlay motion/speech flags on physiological plots; peaks that coincide with flags do not enter “interface-evoked response.”
- When comparing two interfaces, match click and scroll volume or treat those actions as covariates, so a design that invites more clicking is not written up as “more exciting.”
- In the field, record a no-task interval in the same posture first, to see that environment’s typical noise, rather than starting with the stimulus.