For 2-D common-midpoint (CMP) seismic data acquisition, ground roll can be highly attenuated and, under ideal circumstances, virtually eliminated from the stacked data if care is taken to adjust the field geometry so that the long array formed by summing the traces of a CMP gather (i.e., the stackarray) is continuous and evenly weighted. In this paper, we use synthetic field data to quantify the improvement in signal-to-noise ratio that can be achieved by the stackarray and to determine if such processing steps as normal moveout (NMO) correction affect the stackarray's effectiveness. We demonstrate that, while the beneficial effects of the stackarray are mitigated somewhat by NMO correction, source-receiver coupling variations, and static corrections, acquisition geometries that result in a discontinuous or unevenly weighted stackarray have a much more pronounced negative effect. We also compare real field data recorded with geometries that do and do not satisfy the stackarray criterion. On the basis of both synthetic and real field data examples, we conclude that proper acquisition geometry is of prime importance in the effective implementation of the stackarray strategy.