1. We have tested the hypothesis that the non-respiratory, low-frequency (around 0.1 Hz) fluctuations of heart rate variability are generated by the baroreflexes, but with a delay caused by the slower response of the efferent sympathetic arm, (compared with the vagus), in 11 healthy subjects (mean age +/- SD 27 +/- 5 years). 2. In random order, at the onset of 20 s of apnoea starting at end expiration, we applied either 600 ms neck suction (-40 mmHg) to the carotid sinus region, or no stimulus (anticipation control), or a loud whistle (alerting control), every 60 s, for 30 min. (i.e. 10 of each 'stimulus'). We recorded neck pressure, blood pressure (Finapres), R-R interval (ECG), infra-red plethysmographic skin blood flow and respiration (impedance). By subtracting the alerting response from the neck suction response we obtained the responses caused purely by baroreceptor stimulation. 3. The initial reflex bradycardia and hypotension was followed by arteriolar vasoconstriction, presumably due to recompensation by the baroreflex, and then by a further reflex bradycardia-producing a decaying oscillation of the R-R interval about the control R-R. The period of this damped oscillation was 0.103 +/- 0.024 Hz, similar to the frequency of the low-frequency peak obtained by power spectral analysis of heart rate variability (0.093 +/- 0.016 Hz, not significant) at rest. These two values were significantly correlated in individual subjects (r = 0.715, P < 0.025). 4. These findings support the hypothesis that the low-frequency waves of heart rate variability can be generated from baroreceptor sensed blood pressure fluctuations.