Many groups in both basic and clinical studies [1][2][3][4][5][6][7][8][9] have long investigated basic statistical characteristics, including the frequency distribution of arrhythmic beat interval (RR), during atrial fibrillation (AF). Mean RR has been used to evaluate the average RR during AF [9][10][11]. However, the minimal number of beats from which the mean RR has been reliably obtained varied widely between only 4 beats to as many as the number of beats for 2 min among investigators and their purposes [9][10][11]. In some of these studies, mean RR was obtained to characterize the average RR even either in case of the nonnormal distribution or without testing its normality [1,5,6,[8][9][10][11].We have recently studied the frequency distributions not only of RR, but also of ventricular contractility (E max ) [12] and effective arterial elastance (E a ) [13] in in situ ejecting left ventricles during AF [14]. We have found that not only RR, but also E max , E a , and ventriculo-arterial coupling (E a /E max ) distributed nonnormally with considerable skewness (lack of symmetry of a frequency distribution) and kurtosis (peakedness or flatness of a frequency distribution) [14]. Their meanϮSD values could not uniquely characterize their nonnormal frequency distributions [14], unlike a normal frequency distribution [15,16].We recognized in the present study the necessity of isovolumic contractions to scrutinize the frequency distributions of E max during AF for the following reason. Although E a and E a /E max require ejecting contractions whether in situ or ex vivo [13,14,17] Key words: arrhythmia, statistics, histogram, normality, variance.Abstract: Mean levels of left ventricular rhythm and contractility averaged over arrhythmic beats would characterize the average cardiac performance during atrial fibrillation (AF). However, no consensus exists on the minimal number of beats for their reliable mean values. We analyzed their basic statistics to find out such a minimal beat number in canine hearts. We produced AF by electrically stimulating the atrium and measured left ventricular arrhythmic beat interval (RR) and peak isovolumic pressure (LVP). From these, we calculated instantaneous heart rate (HRϭ60,000/RR), contractility (E max ϭLVP/isovolumic volume above unstressed volume), and beat interval ratio (RR1/RR2). We found that all their frequency distributions during AF were variably nonnormal with skewness and kurtosis. Their meansϮstandard deviations alone cannot represent their nonnormal distributions. A 90% reduction of variances of E max and RR1/RR2 required a moving average of 15 and 24, respectively, arrhythmic beats on the average, whereas that of RR and HR required 60 beats on the average. These results indicate that a statistical characterization of arrhythmic cardiodynamic variables facilitates better understanding of cardiac performance during AF.