In cardiac tissue, functional or structural current-to-load mismatches can induce local slow conduction or conduction block, which are important determinants of reentrant arrhythmias. This study tested whether spatially repetitive mismatches result in a steady-state slowing of conduction. Patterned growth of neonatal rat heart cells in culture was used to design unbranched cell strands or strands releasing branches from either a single point or multiple points at periodic intervals. Electrical activation was followed optically using voltage-sensitive dyes under control conditions and in elevated [K+]o (5.8 and 14.8 mmol/L, respectively; in the latter case, propagation was carried by the L-type Ca2+ current). Preparations with multiple branch points exhibited discontinuous and slow conduction that became slower with increasing branch length and/or decreasing inter-branch distance. Compared with unbranched strands, conduction was maximally slowed by 63% under control conditions (from 44.9+/-3.4 to 16.7+/-1.0 cm/s) and by 93% in elevated [K+]o (from 15.7+/-2.3 to 1.1+/-0.2 cm/s). Local activation delays induced at a single branch point were significantly larger than the delays per branch point in multiple branching structures. Also, selective inactivation of inward currents in the branches induced conduction blocks. These 2 observations pointed to a dual role of the branches in propagation: whereas they acted as current sinks for the approaching activation thus slowing conduction ("pull" effect), they supplied, once excited, depolarizing current supporting downstream activation ("push" effect). This "pull and push" action resulted in a slowing of conduction in which the safety was largely preserved by the "push" effect. Thus, branching microarchitectures might contribute to slow conduction in tissue with discontinuous geometry, such as infarct scars and the atrioventricular node.