We studied the relation between a new quantitative index of concealed atrioventricular nodal (AVN) conduction and the variability of ventricular response during atrial fibrillation in 12 patients without preexcitation. The second atrial extrastimulus (A3) was introduced following the first extrastimulus (A2), which was fixed at a coupling interval 20 to 40 msec longer than the AVN effective refractory period (ERP) during a basic atrial drive (A1) cycle length of 400 to 750 msec. The AVNERP of conducted A2 defined as the longest A2A3 interval at which A3 was not conducted to the His bundle was determined. This pacing sequence was repeated, whereas A2 was fixed at a coupling interval 20 to 40 msec shorter than the AVNERP, which means A2 was concealed within the AVN. Thus AVNERP of blocked A2, defined as the longest A2A3 interval at which A3 was not conducted to the His bundle, was measured. Concealment index (AVNERP of blocked A2/AVNERP of conducted A2) was developed to quantitate the magnitude of concealed penetration into the AVN by A2. During atrial fibrillation induced by premature or rapid atrial stimulation, the coefficient of variation (SD/mean) of R-R intervals and the maximum R-R/minimum R-R interval were significantly correlated with the concealment index (r = 0.838, p < 0.001; r = 0.678, p < 0.05). However, neither of these parameters was correlated with AVNERP. Both the minimum R-R and the mean R-R interval were related to the AVNERP (r = 0.946, r = 0.823, p < 0.001). The maximum R-R interval was not related to the AVNERP (r = 0.567, p = NS). We conclude that the concealment index may be a useful predictor of the magnitude of concealed AVN conduction, which may have an important role in determining irregularity of the ventricular response to atrial fibrillation. © 1990.