Respiratory sinus arrhythmia (RSA) is generally known as the autonomically-mediated modulation of the sinus node pacemaker frequency in synchrony with respiration. A respiratory modulation of the atrial interval has been observed even during atrial flutter (AFL), a common supraventricular arrhythmia in which atrial activation is set by a reentrant excitation wave in the atria. Previous studies suggested an intrinsic, mechanical mechanism underling AFL interval modulation (Waxman et al., 1991; Ravelli et al., 2008). The aim of this study is to provide a quantitative and thorough characterization of the cardio-respiratory interactions at the atrial and ventricular level during atrial flutter by application of spectral and cross-spectral analysis. The respiratory spectral powers of atrial (P_AA) and ventricular (P_RR) interval spectra, the coherence and phase of respiratory and atrial (γ²_AA, Φ_AA) and respiratory and ventricular (γ²_RR, Φ_RR) cross-spectra were estimated in 20 patients (68.0±8.8 years, mean ± SD) with typical atrial flutter and some degree of atrioventricular (AV) conduction block. In all patients an oscillation synchronous with respiration (frequency 0.31±0.07 Hz) was observed in both atrial and ventricular interval spectra. The respiratory spectral power was P_AA=3.75±3.31 ms² in atrial spectra and resulted amplified in ventricular spectra (P_RR=50.5±89.0 ms² and P_RR=2418.32±1990.66 ms² in patients with fixed and variable AV conduction, respectively). In all patients strong coupling and in-phase oscillations, corresponding to a paradoxical lengthening of intervals in inspiration with respect to expiration, were observed between respiration and atrial activity (γ²_AA=0.97±0.05, Φ_AA=0.71±0.30 rad). As concerns the interactions between respiration and ventricular activity, high coherence and in-phase oscillations were observed in patients with fixed AV conduction (γ²_RR=0.97±0.03, Φ_RR=1.08±0.80 rad), while lower coherence and mainly antiphase oscillations characterized patients with variable AV conduction (γ²_RR=0.50±0.23, Φ_RR=-1.24±0.97 rad). In all patients no significant frequency-dependent behaviour was observed when controlled respiration was performed in the frequency range 0.1-0.4 Hz, with paradoxical modulation of atrial intervals at all frequencies. This study demonstrates the existence of a respiratory modulation of atrial and ventricular intervals during atrial flutter and characterizes it as paradoxical and frequency-independent, which is consistent with the mechanical origin of the phenomenon. If RSA functions to optimize pulmonary gas exchange (Yasuma & Hayano, 2004), the inversion of normal RSA could exacerbate the already hemodinamically compromised conditions of atrial flutter patients.