Respiratory sinus arrhythmia (RSA) is the heart period variability observed in synchrony with respiration. RSA amplitude is widely used in psychophysiological research to non-invasively index cardiac vagal activity. However, RSA measures are significantly affected by respiratory behavior, even in the absence of changes in cardiac vagal activity. Fifty-to-sixty percent of the variation in RSA can be attributed to respiration rate and tidal volume. This poses a notable challenge for ambulatory RSA measurement where respiratory behavior cannot be experimentally controlled and can show substantial variation. This pre-registered two-day ambulatory study (https://osf.io/57es4) compared four approaches to control for respiratory influences on RSA, to make an empirical recommendation on how to best capture cardiac vagal activity in daily life. We evaluated how well the RSA metric of each approach predicted (1) Minute-to-minute heart period, assumed to be predominantly governed by cardiac vagal activity, and (2) Perceived stress, positive affect, negative affect, and safety—states expected to elicit fluctuations in cardiac vagal activity— at smartphone prompts. The tidal volume-normalized RSA approach was optimal, explaining 1.47 times as much within-individual variance in heart period as that explained by uncontrolled-RSA. The need to use respiratory-controlled RSA was further highlighted by results on safety. Perceived safety was associated with uncontrolled-RSA (p = .033) but not with any of the controlled-RSA metrics. This relationship was driven by higher respiration rate co-occurring with lower safety. We recommend using tidal volume-normalized RSA in ambulatory research to avoid reporting spurious within-individual correlations between psychological states and cardiac vagal activity.