We analyzed patterns of left ventricular perfusion in arrested hearts without coronary tone and in the same hearts while beating with and without coronary tone. We used microspheres in anesthetized dogs to measure blood flow in 384 regions (averaging 140 mg wet wt) from the subendocardium, midwall, and subepicardium before and during intracoronary infusions of adenosine (beating without tone) or lidocaine and adenosine (arrest without tone). Mean coronary pressure was held constant at 80 mmHg. Changes in regional flow with arrest (vs. beating without tone) were surprisingly variable (range -28 to +124%) and exhibited substantial within-layer heterogeneity, suggesting that local differences in contractility, stresses, or strains limit maximum coronary flow. Regional flows in beating hearts with tone did not correlate with flows in the same hearts without tone, beating or not (r2 < or = 0.03; not significant). Flow patterns during beating with tone also demonstrated significantly shorter (i.e., the distance at which autocorrelation has decreased to 0.5) within-layer spatial autocorrelations as well as a complete loss of radial flow correlation (e.g., between corresponding subendocardial and subepicardial regions; r2 = 0.01). Thus neither coronary anatomy (assessed during arrest without tone) nor the mechanical effects of contraction (beating without tone) appear to influence myocardial perfusion when vasomotor tone is present.