Spatial integration characteristics were assessed with drifting gratings for both detection and direction-identification contrast thresholds. Thresholds were measured while stimulus width, length or both were varied. It was found that: (1) the shape of the size/sensitivity functions changes with spatial, but not with temporal, frequency; (2) direction-identification thresholds diverge from the detection thresholds below 1 cycle but can be reliably measured for stimulus widths as small as 0.1275 cycles; (3) the integration characteristics are slightly anisotropic for the identification but not for the detection process, and (4) the two-dimensional spatial integration cannot be directly predicted from its one-dimensional characteristics. Width/sensitivity detection functions are well fitted by predictions of Wilson and Bergen's four-channel model. Predictions from a temporal covariance model provide a poor fit to the identification data. It is argued that classes of detection and direction-identification models must involve identical nonlinearities prior to their respective thresholds. It is concluded that the hypothesis according to which both performances are determined by the same spatial integration stage cannot be rejected.