The purpose of this study is to evaluate cardiac and respiratory modulations in the signals of arterial spin labeling (ASL) perfusion magnetic resonance imaging (MRI) using RETROICOR, an image domain based retrospective correction method. Systematic comparisons were conducted for tagging schemes, pulsed (PASL) versus frequency-modulated continuous (CASL) methods, and the use of background suppression (BGS). Results showed that cardiac pulsation accounted for more signal fluctuation in PASL than in CASL (two-tailed paired Student's t-test, p < 10(-3)), whereas no significant difference was found in the effect of respiratory motion ( p=0.55) on the two tagging schemes studied. For PASL, significantly more improvement was achieved by the inclusion of cardiac pulsation than respiratory motion in RETROICOR ( p < 0.01). On the contrary, the inclusion of respiratory motion offers more improvement for CASL ( p < 0.02). BGS effectively improved the temporal signal-to-noise ratio (tSNR) as previous studies reported, but no significant difference was measured in the spectral power of physiological modulations relative to the entire spectrum of PASL signals before and after the superimposition of BGS ( p=0.63 for cardiac component, p=0.67 for respiratory component). Thus, we conclude that BGS reduces noise without spectral selectivity, and the improvements of tSNR from RETROICOR and BGS are additive. CASL with a labeling duration at a multiple of an R-R interval can be used to minimize signal fluctuation originating from cardiac pulsation.